# The AI Record: every passage > 854 passages from 78 organizations, 827 checked against the source. Each is > reproduced word for word. The rules for using them are at https://the-ai-record.com/llms.txt: quote verbatim > or not at all, give the verification status and grade, say when a source sells AI, show a > context note outside the quotation marks, and link the passage's own page. The archive's own grades, notes and verification records are https://creativecommons.org/licenses/by/4.0/ (CC BY 4.0). Quotations remain the property of their rights holders. ## How widely do employees use AI? Shares of staff using AI tools, how often, and how adoption was driven. Topic page: https://the-ai-record.com/topics/adoption-and-usage > We are seeing at the moment near 100% adoption with our engineers on AI-based coding tools. - Who: Balaji Krishnamurthy, Chief Financial Officer, Uber - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/UBER-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-seeing-moment-near-100-adoption-engineers-ai-2026 > Our associates are generating more than 400,000 prompts a day, and as of last week, we had over 300 AI use cases approved, all of which have good economics, of which 114 are live generative AI use cases. - Who: Brian Moynihan, Chair and Chief Executive Officer, Bank of America - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15267023092982784 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-associates-generating-more-than-400-000-prompts-2026 > As of early April, 100% of Block employees are using AI tools to do their work. One of these tools is Builderbot, which serves as both a multiplier for employees and an end to end automation layer, embedded across our engineering workflows, automating support, reviewing code changes, and initiating production-ready builds. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at Business Highlights. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: What does AI do across the whole company?; Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/block-100-percent-employees-using-ai-tools-builderbot-2026 > Generative AI (“GenAI”) adoption at MercadoLibre has continued to expand across the organization. Approximately 95% of our employees have adopted GenAI tools in their work. Internal measurements show that about 30% of the code that reaches production is written with the use of AI. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How much code does AI write? - Passage page: https://the-ai-record.com/passages/mercado-libre-95-employees-adopted-genai-30-code-2025 > More than 80% of engineers are now using AI tools. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), February 12, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-4Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-more-than-80-engineers-now-using-ai-2026 > from February to Aug 2026, weekly active users across all agentic offerings across all our employees (engineers & non-engineers) grew 7x, and weekly agentic requests grew 9.4x. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Running a Software Factory Efficiently at Uber Scale, Uber engineering blog, August 27, 2026. https://www.uber.com/us/en/blog/efficient-software-factory/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-february-aug-2026-weekly-active-users-across-2026 > Today, 83% are weekly active users, including nearly all of GTM (marketing, sales, customer success managers, and technical account managers). - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/stripe-today-83-weekly-active-users-including-nearly-2026 > Today, 99% of our engineers use AI tools. More than 70% of pull requests are attributed to local or cloud agents. And our engineers have built 2,500+ agent skills across the software development lifecycle. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), July 7, 2026. https://x.com/praveenTweets/status/2074605343439810922 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do organizations train people to use AI? - Passage page: https://the-ai-record.com/passages/uber-today-99-engineers-use-ai-tools-more-2026 > After a couple of months, about 90% of our engineers were now using tools like … regularly to generate code. So on paper, it looked like we were all in on AI. But our CEO was convinced that engineering wasn't using AI at all, like they couldn't be, right? … he was concerned because we weren't shipping any faster. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 0:45. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Jones is describing where Block's engineering organization stood before her agent engineering work began. - Passage page: https://the-ai-record.com/passages/block-90-percent-engineers-using-agents-ceo-not-shipping-faster-2026 > 97% of active engineers inside of Airbnb are using Agentic AI on a weekly basis so far, 90% daily. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/airbnb-97-active-engineers-inside-airbnb-using-agentic-2026 > Glass gives every employee, from engineering to legal, a fully configured AI workspace, helping the company reach 99.5% AI adoption. - Who: Ramp - Source: Ramp's Series F press release, as hosted by Ontario Teachers' Pension Plan (an investor in the round), June 4, 2026. https://otpp.com/en-ca/about-us/news-and-insights/2026/ramp-raises-series-f-at-44-billion-valuation/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-glass-gives-every-employee-engineering-legal-fully-2026 > more than 99% of our engineers use AI coding tools every week, 94% report that AI has made them more productive - Who: Spotify - Source: Coding Is No Longer the Constraint: Scaling Developer Experience to Teams and Agents at Spotify, Spotify engineering blog, June 3, 2026. https://engineering.atspotify.com/2026/6/code-with-claude-coding-is-no-longer-the-constraint - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/spotify-more-than-99-engineers-use-ai-coding-2026 > In the last 30 days, 93% of Cloudflare’s R&D organization used AI coding tools powered by infrastructure we built on our own platform. - Who: Cloudflare (the organization sells AI tools) - Source: The AI engineering stack we built internally, on the platform we ship, Cloudflare blog, April 20, 2026. https://blog.cloudflare.com/internal-ai-engineering-stack/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/cloudflare-last-30-days-93-cloudflares-r-d-2026 > 84% of AI users are now working with agent-style workflows, not just tab completion. Claude Code usage nearly doubled in 2 months (32% → 63%), while IDE-based tools have largely plateaued. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), March 16, 2026. https://x.com/praveenTweets/status/2033627282418655711 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-84-ai-users-now-working-agent-style-2026 > We've tried a top-downs approach, … directives from leaders to say, you must do X, Y, and Z, you must adopt. It's had some impact as we track, as you folks know. You track a metric, it's going to go up, or it's going to improve. The more successful technique that we've applied is actually just sharing wins. … as we share examples between different engineers, cool things that they've tried that have resulted in wins, adoption of that technology has erupted. … that's been the tactic that we're pushing on now is key promoters pushing techniques to their peers, because those promoters are typically engineers, and engineers trust other engineers as opposed to directors like me. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 33:25. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do leaders push AI adoption?; How do organizations train people to use AI? - Context (the archive's note, not their words): Chadha is comparing the ways Uber has tried to raise adoption of its agentic tools among engineers. - Passage page: https://the-ai-record.com/passages/uber-top-downs-approach-more-successful-technique-sharing-wins-2026 > 9 out of 10 Klarna employees (87%) now harness the power of Generative AI in their daily work. - Who: Klarna - Source: Klarna press releases, May 14, 2024. https://www.klarna.com/international/press/90-of-klarna-staff-are-using-ai-daily-game-changer-for-productivity/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-9-out-10-klarna-employees-87-now-2024 > Effectively, 100% of our employees are using AI on a daily basis here. - Who: Rob Witoff, Head of platform, Coinbase - Source: Over 95% of Coinbase’s code is now written with help of AI, Cointelegraph (crypto news site), July 14, 2026. https://cointelegraph.com/news/over-95-of-coinbases-code-is-now-written-with-ai - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/coinbase-effectively-100-employees-using-ai-daily-basis-2026 > More broadly, we have rolled out Claude Cowork to 31,000 employees, making Mercado Libre one of the earliest, large-scale enterprise adopters globally. - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To our Shareholders, Artificial Intelligence. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/mercado-libre-rolled-out-claude-cowork-31-000-2026 > So it's really not even a question of where are we using AI, but where aren't we using AI? Because it's been an extensive usage in pretty much all departments within the company. - Who: Jeff Hoffmeister, Chief Financial Officer, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-so-its-really-not-even-question-where-2026 > Expanded the adoption of Generative AI tools with the rollout of new features that enable faster and easier access and increase overall productivity … Adoption continues to accelerate, with approximately 7 million utilizations of Citi’s two enterprise-wide tools year-to-date (approximately three times increase versus previous quarter) - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended September 30, 2025, SEC EDGAR, November 6, 2025, at Management's Discussion and Analysis, Transformation progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000154/c-20250930.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/citigroup-expanded-adoption-generative-ai-tools-rollout-2025 > Logged 385,000 utilizations of two enterprise-wide tools (document intelligence and virtual assistant) - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended March 31, 2025, SEC EDGAR, May 8, 2025, at Management’s Discussion and Analysis, Transformation, Progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000086/c-20250331.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: What does AI do across the whole company? - Context (the archive's note, not their words): Listed under the 10-Q bullet "Significantly expanding adoption of Generative AI tools, increasing efficiency and productivity across Citi". - Passage page: https://the-ai-record.com/passages/citigroup-385-000-utilizations-document-intelligence-virtual-assistant-2025 > We've run almost three million sessions with it. To give an idea of the growth curve this has resulted in, in December, before the holidays, Inspect was writing around 10% of merged … requests across all of Ramp. That jumps to roughly 30% around the start of January, and then around 50% at the end of January, and today it's about 75%. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 05:28. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How much code does AI write? - Context (the archive's note, not their words): Bruggeman describes Inspect's growth from a side project. Years are inferred from his references to 'last year' and December/January; the talk was published 2026-11-02. - Passage page: https://the-ai-record.com/passages/ramp-ive-run-almost-three-million-sessions-inspect-2026 > But what was really crazy about this is that at no point did we ever have a mandate for people to use Inspect. This was just a natural growth curve in the company, built from people watching others use Inspect successfully. Even today, there's still no mandate. That's the reason it's not 100%, and frankly, we aren't aiming for that. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 05:59. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Follows his statement that Inspect writes about 75% of merged pull requests at Ramp. - Passage page: https://the-ai-record.com/passages/ramp-no-point-did-we-ever-mandate-people-use-inspect-2026 > You can give it any prompt directly by just reacting with the little emoji to the message and it will pick it up as prompt. So, that evolved … and that became an always available teammate inside our Slack and whole environment. - Who: Francesco Galletta, Leads analytics engineering for financial crime; co-leads AI strategy for the data discipline, Monzo, Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 1:57. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Galletta is describing how Monzo's in-house agent (Agent Chip) began as a Slack emoji reaction after it spread inside the company in January 2026. - Passage page: https://the-ai-record.com/passages/monzo-reacting-little-emoji-always-available-teammate-inside-slack-2026 > 90% of these teams used Kiro, among other internal tools that we have, and what they found was it wasn't about the tools, it was about the way that they worked. The teams that achieved step function improvements intentionally changed the way that they worked, and the other simply kind of sprinkled Kiro and some of the other tools that we have on top of their existing way of working. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 7:08. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; Why do some teams gain far more from AI than others? - Context (the archive's note, not their words): Liguori is explaining what separated the Amazon Stores pilot teams that saw a median 4.5x improvement from the rest. - Passage page: https://the-ai-record.com/passages/amazon-90-percent-used-kiro-not-the-tools-the-way-they-worked-2026 > the adoption is uneven. We have teams that are very AI-forward and have already transformed their entire workflows and then we have teams that are still experimenting in the earlier acts or have lost confidence - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 02:11. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company?; Why do some teams gain far more from AI than others?; How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/figma-adoption-uneven-teams-ai-forward-others-lost-confidence-2026 > It's actually our best engineer. … they know where all the pitfalls are. They are … together with like their mental duct tape, all the places that agents are not working well and they're preventing all the really bad stuff from coming in … slower to adapt because they see all the problem … - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 03:27. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/figma-best-engineer-know-pitfalls-mental-duct-tape-slower-adapt-2026 > because they're seeing … where you're lacking validation, where your tools fail. So, and their feedback is basically the roadmap of how to improve your agent … - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 12:02. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/figma-skeptics-feedback-roadmap-improve-agent-2026 > Today, we have 800 plus projects internally going through this, cumulatively handling more than 100 million model requests per day. This includes both the frontier models and also open source models. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 03:30. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations choose and switch AI models? - Context (the archive's note, not their words): Medisetty is describing Uber's model gateway, the single endpoint all internal model use goes through. - Passage page: https://the-ai-record.com/passages/uber-800-projects-100-million-model-requests-per-day-2026 > So we have a … assistant called Cortana, all of the things that I mentioned so far, whether it's skills, MCPs and context graph, they're all plugged into that in every surface possible, whether it's on Slack, CLI, web. So anyone in the company, they can ask a simple question, it can look up the context graph, invoke any skill, check any code, check any code in any code base, and give an answer across any of these surfaces. … in the last one month, 300 unique … created and more than 20,000 sessions per day. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 10:19. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Context (the archive's note, not their words): Medisetty is describing the assistant that packages Uber's skills, tool connections and context graph for every employee. - Passage page: https://the-ai-record.com/passages/uber-cortana-assistant-every-surface-20000-sessions-per-day-2026 > nearly all of our engineers now using at least one coding agent regularly - Who: Monzo - Source: Building Agent Chip: Monzo's in-house agentic tool, Monzo blog, August 13, 2026. https://monzo.com/blog/building-agent-chip - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/monzo-nearly-all-engineers-now-using-least-one-2026 > we haven't done any mandates inside of Airbnb. It's all been organic adoption. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/airbnb-havent-done-any-mandates-inside-airbnb-its-2026 > So as of now, we have about double the amount of users of agentic AI tools as we do active developers - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-so-now-about-double-amount-users-agentic-2026 > Today, 100% of Block's engineers regularly use AI in their work. - Who: Block - Source: Block rolls out BuilderBot, a new suite of AI-native tools that changes the way we ship, Block (own site), June 17, 2026. https://block.xyz/inside/block-rolls-out-builderbot-a-new-suite-of-ai-native-tools-that-changes-the-way-we-ship - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/block-today-100-blocks-engineers-regularly-use-ai-2026 > 600+ users including engineers, PMs, and designers, with 335 active and 197 power users. 5,068 merged PRs across 461 repositories and 10 orgs. 3.5x more merged PRs per engineer for Mux users compared to baseline: 39.6 vs 11.4. - Who: Coinbase - Source: Coding Had a Concurrency Problem: How Mux Helped Solve It, Coinbase engineering blog (byline: Alessandro Rodi, Adam Wolf), May 11, 2026, at The Evidence: Mux. https://coinbase.com/blog/coding-had-a-concurrency-problem-how-mux-helped-solve-it - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Does AI make developers more productive?; Can people outside engineering ship code with agents? - Passage page: https://the-ai-record.com/passages/coinbase-600-users-5-068-merged-prs-mux-2026 > the good part was that we were ready with all the background work that we were doing with respect to the building of the context and building of the right sort of MCP infrastructure internally. So when that became a reality, we started seeing a lot more adoption and because it started helping in the real work that people were doing. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 6:16. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; Why do some teams gain far more from AI than others? - Context (the archive's note, not their words): Answering how Cloudflare reached 93% of R&D using AI tools; he has just said adoption was slow at first because the models were not good enough. - Passage page: https://the-ai-record.com/passages/cloudflare-ready-background-work-context-mcp-infrastructure-2026 > In the first 30 days, the system completed 131,246 review runs across 48,095 merge requests in 5,169 repositories. The average merge request gets reviewed 2.7 times (the initial review, plus re-reviews as the engineer pushes fixes), and the median review completes in 3 minutes and 39 seconds. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at Show me the numbers! / The overview. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/cloudflare-first-30-days-131-246-review-2026 > As soon as we made our agentic … available to developers, what we saw is 70 percent of the workloads that developers are pushing into the system were toil tasks. There's a couple of reasons for that. One is the accuracy of these tasks was much higher compared to the more ambiguous workloads. And it makes sense, right? Like the start and end state of some of these tasks, if you think about a library upgrade or a migration, is much more straightforward versus, say, building a brand new feature or an experience … that requires an experiment. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 06:47. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations?; How do coding agents open and merge changes? - Context (the archive's note, not their words): Chadha is describing what developers first sent to Uber's agentic system and why. - Passage page: https://the-ai-record.com/passages/uber-agentic-workflows-available-70-percent-workloads-toil-tasks-2026 > I ran a demo session with some of my VPs, and in 24 minutes, I had four VPs land code for the first time in years. It was a pretty amazing experience. They were pretty satisfied by it, too. But our adoption for this technology has been relatively slow. It's been slower than I've expected, and part of it is because we're trying to have developers do something that they're so not used to. They're used to looking at code and generating from scratch, operating in their IDE, and we're telling them to take a risk by operating in a very different way. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 32:45. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Can people outside engineering ship code with agents?; How do leaders push AI adoption? - Context (the archive's note, not their words): Chadha is describing a demo of Uber's agentic tooling to VPs and why developer adoption has been slower than he expected. - Passage page: https://the-ai-record.com/passages/uber-ran-demo-vps-24-minutes-four-vps-land-code-adoption-slow-2026 > In March 2025, we had roughly 2.5x as many Tab users as agent users. Now, that is flipped: we now have 2x as many agent users as Tab users. - Who: Michael Truell, Author of the Cursor blog post; CEO of Cursor per the lead record (the page gives no title), Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026, at From Tab to agents. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cursor-2-5x-tab-users-agent-users-flipped-2026 > Agent usage in Cursor has grown over 15x in the last year. - Who: Michael Truell, Author of the Cursor blog post; CEO of Cursor per the lead record (the page gives no title), Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026, at From Tab to agents. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cursor-agent-usage-grown-over-15x-year-2026 > Dropbox saw AI usage among our engineering teams jump to over 90% this year - Who: Dropbox - Source: Dropbox engineering blog, 2026. https://dropbox.tech/culture/ai-adoption-productivity-dropbox-cto-ali-dasdan - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/dropbox-dropbox-saw-ai-usage-among-engineering-teams-2026 > … you have the same kind of, on the vertical axis, the productivity increase, and then on the horizontal one, you have the token usage per engineer per month on a logarithmic scale. And what you can see is that the correlation is quite loose, point 20 or so linearly, and there is a bit of a death valley effect around the 10 million token mark whereby teams that were using that amount of … seem to be doing worse than teams that were using a bit less tokens. It's very directional, but interesting nevertheless. The conclusion here might be that AI usage quality matters more than AI usage volume. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 03:24. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Denisov-Blanch is describing a chart of teams' productivity gain from AI against token use. - Passage page: https://the-ai-record.com/passages/stanford-correlation-quite-loose-death-valley-10-million-token-2025 > We applied this benchmark to one of the companies in our research data set and we saw this. This company had two business units with equal access to AI tools, right? Same licenses, … same tools, same everything. But the adoption rate and the usage rate was very different by business unit. On the left, the first business unit, as you can see in the area in the blue, … to be using AI a lot more for almost 40% of their work. Whereas on the right, the second business unit seemed to struggle behind a bit more. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 07:35. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Denisov-Blanch is describing one company in Stanford's data set, which he does not name, and how two of its business units used AI. - Passage page: https://the-ai-record.com/passages/stanford-two-business-units-equal-access-different-adoption-2025 > This AI Assistant started as an internal experiment and now has scaled to support over 90% of PRs across the company impacting more than 600K pull requests per month. - Who: Sneha Tuli, Principal Product Manager, Microsoft (the organization sells AI tools) - Source: Enhancing Code Quality at Scale with AI-Powered Code Reviews, Engineering@Microsoft blog, July 14, 2025. https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/microsoft-ai-assistant-started-internal-experiment-scaled-2025 > And so we got a lot of feedback. And then we had a public channel. And so we started getting like all of the engineering managers started using it and asking us if we could like come help them talk about how to do knowledge work for engineering. Our data scientists have one that is like essentially a fork of this. And so we were getting a lot of feedback about what it was good at and what it wasn't good at. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 13:25. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: What does AI do across the whole company?; Who builds an organization's agents: every team, or a central group? - Context (the archive's note, not their words): Mechaber is describing who beyond product managers picked up Figma's PM OS. - Passage page: https://the-ai-record.com/passages/figma-pm-os-engineering-managers-data-scientists-fork-2026 > we did an off site with every PM at Figma this spring where we had everyone install it, use it, hack on it, come up with like what was good, what was bad, just like roast it. And so some of the skills that are in here came out of that. Some of the skills either got whittled down or whittled away. … I would say as the models have gotten better, I've been trying to basically take stuff out that's no longer useful. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 12:32. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): Mechaber is describing how the skills in Figma's PM OS were chosen and trimmed. - Passage page: https://the-ai-record.com/passages/figma-pm-os-offsite-every-pm-install-hack-roast-whittled-2026 > I walked into Figma a couple of years ago and I thought everyone was going to be using AI. It was going to be crazy. I was already feeling behind. … And I walked in feeling really behind and I realized basically no one was using these tools outside of engineering to like get real stuff done. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 02:58. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in product and design? - Context (the archive's note, not their words): Mechaber is answering the host's question about why he built Figma's PM OS. - Passage page: https://the-ai-record.com/passages/figma-walked-in-no-one-using-ai-outside-engineering-pm-os-2026 > So we started out the year with 3% of deployments on … we're coming from coding agents. Now 60% are agentic. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 14:52. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Rauch answers the host's question about how many companies still read the code their coding agents write, and is describing deployments on the Vercel platform. - Passage page: https://the-ai-record.com/passages/vercel-3-percent-deployments-coding-agents-now-60-agentic-2026 > For a long time, we had a 'capability overhang': the models were capable but the products didn't fully bring that out. Now, we're seeing an awareness gap. Some people have figured out they can use Codex to monitor Slack, update Airtable, or create onboarding materials. But many others still use it for one task and then discover more uses from teammates via word-of-mouth. - Who: Akshay Nathan, Engineering Lead, Productivity team, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 1. Codex takes over at OpenAI. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Also filed under: What does AI do across the whole company?; Why do some teams gain far more from AI than others? - Passage page: https://the-ai-record.com/passages/openai-long-time-we-had-capability-overhang-models-2026 > The number one thing that is changing is that people are starting to use threads for much longer, and this longer usage has been a breakthrough. It's surprising to see the sheer length of time that people spend on a thread – even days! They often set a goal and then have the model crank. - Who: Andrew Ambrosino, Lead, Desktop, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 1. Codex takes over at OpenAI. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/openai-number-one-thing-changing-people-starting-use-2026 > And I think … when I look at most companies, there's still somewhere between step one and two. … And when I look at Anthropic, we're probably somewhere around step three on average. So most engineers are running dozens or hundreds of agents. And the way you do this is you have the agent run another agent. And we have a lot of tools for this. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 20:38. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How many agents can one person run at once? - Context (the archive's note, not their words): Cherny has just described stages companies go through, from every engineer running one agent to 10, 100 and then 1,000. - Passage page: https://the-ai-record.com/passages/anthropic-most-engineers-running-dozens-hundreds-agents-2026 > don't promise an engineer the world when you give them a platform, give them a valuable use case because you have to go to where the pain lives. If I can prove to an engineer that the core value proposition of taking certain painful toil tasks away from you is beneficial, they will naturally understand that AI is not here necessarily to take away all of the work, just the parts of the work that are tedious, they require a … and tedium to go execute. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 05:59. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Ramakrishnan is describing how Uber rolls Minion out to engineers. - Passage page: https://the-ai-record.com/passages/uber-dont-promise-engineer-the-world-go-to-where-pain-lives-2026 > People had to realize for themselves, it had to reach a point of capability before people adopted it. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 11:02. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Ramakrishnan is answering a host question on whether Uber's builder-oriented engineering culture drives how fast it adopts AI. The sentence before it, on mandating AI use, is left out because the two passes hear it differently. - Passage page: https://the-ai-record.com/passages/uber-mandating-did-not-work-people-realize-capability-2026 > pre that week we had roughly 10% of the employee base was using AI with some frequency. After that week it was over 50%. - Who: Wade Foster, Co-founder and Chief Executive Officer, Zapier - Source: Madrona (venture capital firm) podcast, April 2, 2026. https://www.madrona.com/zapier-has-more-ai-agents-than-employees-heres-how-that-happened/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/zapier-pre-week-had-roughly-10-employee-base-2026 > What we are seeing among early customers of Moneybot and Managerbot is that agency matters. When AI helps a customer take an action, they return at dramatically higher rates than when it merely informs. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To Our Shareholders. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/block-ai-helps-customer-take-action-return-higher-rates-2026 > We expect this leverage to be partially offset by the ramp up in our use of third-party AI models. This reflects a deliberate choice to prioritize company-wide adoption over cost optimization at this stage. This investment is contributing to cost efficiency across many other expense lines (customer service, for example) as well as helping us to generate incremental revenue (in Acquiring and Advertising, for example). - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at Costs and expenses, Product Development. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: What does AI cost organizations?; How do organizations keep AI spending in check?; Does AI customer service cut the cost of support? - Passage page: https://the-ai-record.com/passages/mercado-libre-ramp-up-our-use-third-party-ai-2026 > We also reduced or canceled contracts with over 1,700 suppliers following our AI adoption and standardization - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in finance, legal and operations? - Passage page: https://the-ai-record.com/passages/klarna-also-reduced-canceled-contracts-over-1-700-2026 > To drive usage, the Company has trained over 23,000 people, including leaders and developers, on solving business challenges with AI tools, prompting, and coding. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business, AI-Driven Innovation at MercadoLibre. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations train people to use AI? - Passage page: https://the-ai-record.com/passages/mercado-libre-trained-over-23-000-people-including-leaders-2025 > Since the beginning of 2025, we've seen a 30% increase in output per engineer, with the majority of that growth coming from the adoption of agentic coding, which saw a big jump in Q4. - Who: Meta - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), January 28, 2026. https://transcripts.platformaeronaut.com/transcripts/META-4Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/meta-since-beginning-2025-weve-seen-30-increase-2026 > The big companies I’m talking to (e.g. banks) tell me they are still struggling with AI in software development. They see promising but inconsistent results, non-replicability across teams, ballooning costs, internal cultural pushback, etc. - Who: Steve Yegge, Independent; builds and runs Wheelhouse, a solo multi-agent software factory (2026), Steve Yegge (independent) (the organization sells AI tools) - Source: The Counting House, Medium (Steve Yegge), October 27, 2026. https://steve-yegge.medium.com/the-counting-house-steve-yegge-83f5145cc35b - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/steve-yegge-big-companies-banks-still-struggling-ai-2026 > River initially created only 6% of merged mobile pull requests. This is how the number evolved over time, reversing the trend of local development and evolving to 61% of merged mobile pull requests - Who: Shopify - Source: We gave AI agents a mobile dev environment, Shopify Engineering, October 22, 2026, at Helping people use it. https://shopify.engineering/remote-dev - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/shopify-river-initially-created-only-6-merged-2026 > It's increasingly common now at SpaceX where we see bots joining meetings for employees … but they want their bot to go and just take notes and come back to them, which has gotten funny. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 11:26. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Context (the archive's note, not their words): Robinson is listing the events that wake an always-on agent, such as a Slack message or a meeting. The cut skips 'where they could make it to the meeting', which both passes transcribed the same way but which reads as a mis-hearing. - Passage page: https://the-ai-record.com/passages/spacex-bots-joining-meetings-take-notes-2026 > a 5% decline in case volume —and 12% fewer cases than forecasted—even while product adoption grew - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site, webinar listing), October 2026. https://www.salesforce.com/events/webinars/how-salesforce-uses-agentforce-to-transform-support-handoffs/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How much customer support does AI resolve? - Passage page: https://the-ai-record.com/passages/salesforce-5-decline-case-volume-12-fewer-cases-2026 > about a quarter of all the pull requests … in any given month were completely or at least largely generated by AI. Then I backtracked this data to the last 12 months. And it turns out that this number is growing really fast. In fact, early last year, fewer than 1% of pull requests had any evidence of being completely … generated. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 04:36. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How much code does AI write? - Passage page: https://the-ai-record.com/passages/greptile-quarter-prs-largely-ai-generated-under-1-percent-2026 > The second one is slowing down to speed up. In almost every team that was interviewed, they reported that their productivity actually went down, as they intentionally adopted a new way of working. That's counterintuitive, right? You have to do intentional engineering work before you're gonna see that hockey stick curve in productivity improvement. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 9:39. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; Why do some teams gain far more from AI than others? - Context (the archive's note, not their words): Liguori is describing the second of five habits that Amazon teams in the pilots and sprints built, according to the interviews she cites. - Passage page: https://the-ai-record.com/passages/amazon-slowing-down-to-speed-up-productivity-went-down-2026 > They watched 50 teams that were totally … distribution of early career folks, mid-career senior engineers, and that worked on existing systems. … And they watched them for the better part of last year, and they found something super interesting. They found that there was a big difference in the productivity gains that they saw between half of the teams and the other half. And in this case, they used a productivity metric of deployment velocity to production. … And they saw that for half of the teams, they achieved less than 3X increase. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 5:59. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Does AI make developers more productive?; Why do some teams gain far more from AI than others? - Context (the archive's note, not their words): Liguori is describing a pilot run by Amazon Stores, which she says covers Amazon.com, its retail websites and its physical stores. - Passage page: https://the-ai-record.com/passages/amazon-stores-pilot-50-teams-half-under-3x-deployment-velocity-2026 > Figma has pivoted very strongly from being a traditional tool to an AI-first tool. But in this talk, I'm not … talk about our product, I'm … talk more about our internal organization and how our engineering org has been adopting AI agents. - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 00:44. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/figma-pivoted-traditional-tool-ai-first-engineering-org-adopting-agents-2026 > And we have 2,500 skills and … more than 20,000 skill executions per day across our fleet. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 08:27. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Medisetty is summarizing how much the managed skills are used. - Passage page: https://the-ai-record.com/passages/uber-2500-skills-20000-skill-executions-per-day-2026 > So we put out Composer 2.5 in May. And it's now the most popular model in Cursor, which is exciting. And we scaled up training here quite a bit by generating more RL environments, trying out some new methods for learning. And also just making more ambitious problems for the models to solve. And the results have been pretty promising so far. - Who: Lee Robinson, Machine Learning Engineer, Model Behavior, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Recursive Model Improvement, AI Engineer (YouTube), July 15, 2026, at 2:41. https://www.youtube.com/watch?v=q4Tr-DknG2M - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Robinson is describing Cursor's own model, Composer, which Cursor trains in-house. - Passage page: https://the-ai-record.com/passages/cursor-composer-2-5-most-popular-model-2026 > And in reality, the vast majority of our revenue today comes from agent usage. And that means that all of the data inside of Cursor is also coming from agent usage, and we can use that to train better models. - Who: Lee Robinson, Machine Learning Engineer, Model Behavior, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Recursive Model Improvement, AI Engineer (YouTube), July 15, 2026, at 4:41. https://www.youtube.com/watch?v=q4Tr-DknG2M - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Robinson is explaining where the data for training Cursor's models comes from. The talk gives no revenue figure. - Passage page: https://the-ai-record.com/passages/cursor-vast-majority-revenue-agent-usage-2026 > We were merging on the order of forty PRs a day across our GitHub organizations, well over a thousand a month, and climbing. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at Opening. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/rewind-merging-forty-prs-day-well-over-thousand-month-2026 > we've been trying to build … platform layer around AI tooling at Cloudflare, not just coding … but the stack that makes the tools useful and safe inside a real engineering organization. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 2:10. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Describing what his team built, in answer to a request to summarize the internal work. - Passage page: https://the-ai-record.com/passages/cloudflare-build-platform-layer-around-ai-tooling-2026 > sometime in November, we started looking and seeing a lot more benefit for some of the models that we were trying to leverage. I think Opus 4.5 maybe was the inflection point where that then generally made it so that it was very easy to see the kind of results we were seeing and expecting. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 7:54. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Asked which month or model made AI coding tools start to work at Cloudflare; he had said the company began piloting coding assistants in mid-2025. - Passage page: https://the-ai-record.com/passages/cloudflare-opus-4-5-maybe-was-inflection-point-2026 > you can scale this to like 100 or 1,000 agents. I'm not sure how many we are running today, but I bet that's like multiple thousands a day, just agents running in the same or like copies of the code base. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 21:54. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Zakariasson is talking about running cloud agents, each in its own virtual machine, and guesses how many Cursor itself runs; he says he does not know the number. - Passage page: https://the-ai-record.com/passages/cursor-multiple-thousands-of-agents-running-a-day-2026 > we run around 50,000 builds a week, a number that is only increasing with AI adoption among our engineering teams accelerating commit rates and productivity - Who: Aditya Anchuri, Tech lead, Ruby Infra Platform team, Stripe - Source: Stripe developer blog, April 9, 2026. https://stripe.dev/blog/selective-test-execution-at-stripe-fast-ci-for-a-50m-line-ruby-monorepo - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/stripe-run-around-50-000-builds-week-number-2026 > A powerful result is that anyone is able to contribute regardless of their technical expertise or role in the company, they don’t need to understand and set up Git or any command line tools to start contributing to our codebases. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at Core Experience. https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; Can people outside engineering ship code with agents? - Passage page: https://the-ai-record.com/passages/cognition-anyone-able-contribute-regardless-technical-expertise-role-2026 > We've found it especially useful for the kind of ad-hoc questions that used to sit in someone's queue for days. Questions like "why did signups drop Tuesday?" or "break down consumption by enterprise vs self-serve" can be answered by anyone on the team just asking in Slack, with the SQL included so we can validate the logic. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at DANA (Data Analyst Agent). https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/cognition-dana-ad-hoc-questions-sat-queue-days-2026 > When surveyed, 30% to 50% of developers told us that they were choosing not to submit some tasks because they did not want to do them without AI. This implies we are systematically missing tasks which have high expected uplift from AI. - Who: METR - Source: We are Changing our Developer Productivity Experiment Design, METR, February 24, 2026, at Wider adoption of AI has made it more difficult to measure task-level productivity. https://evals.alignment.org/blog/2026-02-24-uplift-update/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): METR is a nonprofit that studies AI agents; this post reports on its own randomized study of AI use by experienced open-source developers, not on METR's internal AI use. - Passage page: https://the-ai-record.com/passages/metr-surveyed-30-50-percent-developers-not-submit-tasks-2026 > The AI is approving over 60% of all PRs. For those PRs, engineers get feedback in minutes, not hours. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Results: Speed and cognitive relief. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Can AI approve code changes without a person? - Passage page: https://the-ai-record.com/passages/customer-io-ai-approving-over-60-percent-prs-feedback-minutes-2025 > Yeah, I think they … happening entirely … organically. No, I think this is like, I think this is like, I don’t have to enforce these kind of things. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify CEO on why using AI isn’t optional, Sources with Alex Heath (YouTube), October 28, 2026, at 49:32. https://www.youtube.com/watch?v=r9SHzCW9zxo - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Lütke is answering the host’s question whether two rules from his 2025 memo, bringing AI into performance reviews and asking whether AI can do the work before adding headcount, are still the rules at Shopify. - Passage page: https://the-ai-record.com/passages/shopify-happening-entirely-organically-dont-have-enforce-2026 > But not in Shopify, which was fun. … so like in Shopify actually started, I think more people ended up paying attention to the memo after the press made a big deal out of it … that was actually gratifying because … it was seen as controversial, but really not inside of a firm. That just means we … managed to transition already pretty early. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify CEO on why using AI isn’t optional, Sources with Alex Heath (YouTube), October 28, 2026, at 45:37. https://www.youtube.com/watch?v=r9SHzCW9zxo - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Lütke is responding to the host, who says his memo to staff about AI use was controversial and drew a lot of headlines. - Passage page: https://the-ai-record.com/passages/shopify-more-people-ended-up-paying-attention-memo-press-2026 > I actually don’t know if it’s in a current one. There was like a question about, like this sort of peer rating kind of thing is like how reflexively does so-and-so show up … a prototype from AI or something. It was, you know, sometimes you play a little bit of tricks there. … this was in a performance review questionnaire, but of course it didn’t matter. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify CEO on why using AI isn’t optional, Sources with Alex Heath (YouTube), October 28, 2026, at 49:46. https://www.youtube.com/watch?v=r9SHzCW9zxo - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Lütke is answering the host’s question whether two rules from his 2025 memo, bringing AI into performance reviews and asking whether AI can do the work before adding headcount, are still the rules at Shopify. - Passage page: https://the-ai-record.com/passages/shopify-peer-rating-how-reflexively-prototype-from-ai-2026 > It took us about 10 years to get to a billion deployments. Deployment is one of my favorite words. Every time you ship a change to your app, that's a deployment. And then in the next 10 months, the 10 months since then, until now, we've added another 1.5 billion deployments. So what took us 10 years, we've not sort of passed in just a few months. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 8:45. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Rauch is explaining what he means by the exponential, before he says why shipping has sped up. - Passage page: https://the-ai-record.com/passages/vercel-10-years-billion-deployments-added-1-5-billion-10-months-2026 > But what's happened is that in the beginning of the year … most of the traffic was going to proprietary models. So OpenAI, Gemini, XAI, et cetera. And now the situation is almost entirely flipped. 80% of tokens in our network … is open models. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 54:48. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Rauch is describing traffic through the Vercel AI Gateway, which he had introduced as the network that distributes tokens from many model providers. - Passage page: https://the-ai-record.com/passages/vercel-80-percent-tokens-network-open-models-flipped-from-proprietary-2026 > The big theme of the past months has been that everything is now a coding agent. Whether the visible code is your output or not, agents write your artifacts. - Who: Andrew Ambrosino, Lead, Desktop, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 1. Codex takes over at OpenAI. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Also filed under: What does AI do across the whole company? - Passage page: https://the-ai-record.com/passages/openai-big-theme-past-months-everything-now-coding-2026 > And honestly, [Claude Code] didn't work very well for … the first six months. … And it wasn't until, I would say, Opus 4 in May of 2025, where it really started to work. And … since then, the growth for … the product has been exponential also. It was not at first, but that's when it became exponential. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 8:05. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Cherny is explaining why Anthropic built a purely agentic product and how it developed. - Passage page: https://the-ai-record.com/passages/anthropic-claude-code-exponential-growth-after-opus-4-2026 > But most usage is actually much more delegating via tagging… And then more advanced version is really trying to think of it as a teammate that is actually … has memory and can be proactive. And that's just really changed how we operate internally. It's much more like this multiplayer async proactive way than it is most people off in their own CLIs. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 8:57. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Describing how people at Anthropic work with Claude. The recording's transcripts render the name of what is being tagged unclearly, so the passage is cut at 'tagging'. - Passage page: https://the-ai-record.com/passages/anthropic-most-usage-delegating-via-tagging-multiplayer-async-2026 > If I … give you the tool and I tell you use this or else, that's not gonna work. If I tell you, I give you the tool and that tool says, I can do X, Y and Z and then you actually put it in production and it can do a quarter of X, a tenth of Y and a 50th of Z, I will fail there as well. What I need is the engineer to be able to see that I know the pain they see. That's what gets a builder really hungry. When they know that they can take the tool and they can mold it to their outcome. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 12:08. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Ramakrishnan is describing how Uber's platform team gets engineers to keep using a tool, after saying that adoption comes from letting people fail in a safe way. - Passage page: https://the-ai-record.com/passages/uber-tool-use-this-or-else-not-gonna-work-engineer-sees-pain-2026 > You let them use whatever they want. Here's what I'll say. If they don't use a copilot, they usually get creamed by someone who does. - Who: Farhan Thawar, Head of Engineering, Shopify - Source: How AI is changing software engineering at Shopify with Farhan Thawar, The Pragmatic Engineer, July 2, 2025, at Interesting detail: Shopify’s interview process. https://newsletter.pragmaticengineer.com/p/how-ai-is-changing-software-engineering - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Context (the archive's note, not their words): From a live podcast conversation at LDX3 in London between Gergely Orosz and Farhan Thawar, about how Shopify lets candidates use AI in its interviews. - Passage page: https://the-ai-record.com/passages/shopify-let-them-use-whatever-they-want-copilot-creamed-2025 ## What are agents allowed to touch? Sandboxes, permissions, data access and approval prompts that limit what agents can reach and do. Topic page: https://the-ai-record.com/topics/agent-permissions > what we were realizing is if there's a person that's deciding yes or no every time, at some point, the person actually just stops reading, and they just say yes, yes, yes, because they're tired of pressing yes. … And our security people were also realizing this. And so … the thing that they started working on is, is there a way to route this to a classifier? So the classifier can decide. … It took many months to kind of get this to a mature state. And in the end, through EVALS, through red teaming, through pen testing, we're able to show that this is actually much safer. And so … at Anthropic, we run on Auto Mode, and we recommend this to all our customers, too. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 34:33. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How does AI help keep a product secure? - Context (the archive's note, not their words): Cherny is explaining why Anthropic's security team worked on routing permission decisions to a classifier. - Passage page: https://the-ai-record.com/passages/anthropic-run-on-auto-mode-classifier-instead-of-yes-2026 > We take a stance with the sandboxing that hurts us in terms of general adoption. However, we do not want to promote something that could be unsafe by default. As a dev, you can always go into your configuration and disable these settings if you want. - Who: Thibault Sottiaux, Head of Codex, OpenAI (the organization sells AI tools) - Source: How Codex is built, The Pragmatic Engineer, February 17, 2026, at 2. Technology and architecture choices. https://newsletter.pragmaticengineer.com/p/how-codex-is-built - Grade C: An interview or press quote. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; Where do agents run? - Passage page: https://the-ai-record.com/passages/openai-we-take-stance-sandboxing-hurts-us-2026 > Very simple philosophy. We use our screens. We look at stuff. We do work. Every single thing that a human can see in pixels on the screen, an agent in your building should be able to see via API. The analogy I use is imagine you brought the smartest people ever, the greatest investors alive, and you brought them into your company, but you gave them a computer without logins to 75% of your services. It's almost immaterial how smart … system is if it doesn't have the tools to actually action that. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 01:36. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Long is describing the inputs an agent needs to do investing work, the first part of his framework. - Passage page: https://the-ai-record.com/passages/coatue-agents-should-see-what-humans-see-via-api-2026 > There is structure: doctrine and policy, human gates, and just fourteen mechanical “fences” (programmatic denials). But my agents are first-class colleagues. They work under my own identity, signing their own names to the work, and they can see the whole system and act independently. It is YOLO mode scaled up to the size of a small company. - Who: Steve Yegge, Independent; builds and runs Wheelhouse, a solo multi-agent software factory (2026), Steve Yegge (independent) (the organization sells AI tools) - Source: The Counting House, Medium (Steve Yegge), October 27, 2026, at Wheelhouse Recap. https://steve-yegge.medium.com/the-counting-house-steve-yegge-83f5145cc35b - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/steve-yegge-fourteen-mechanical-fences-agents-first-class-2026 > Again, we are a bank. We live in a regulated environment. So, we want to make sure that the non-determinism from the agents is controlled. But at the same time, we want to have an environment where we can let it run loose and develop. - Who: Felippe Felisola Caso, Monzo (conference affiliation, 2026), Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 6:50. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run?; How autonomous are the agents, and who checks their work?; How do banks and other regulated organizations use AI within the rules? - Context (the archive's note, not their words): Felisola Caso is explaining why Monzo's agents run in a sandbox in an isolated namespace, separate from production infrastructure. - Passage page: https://the-ai-record.com/passages/monzo-we-are-bank-regulated-environment-non-determinism-agents-controlled-2026 > It has access to some production systems with guardrails, you know? So we built, we talked about specialist dots, but specialist dots is just like, operate with additional guardrails, additional monitoring, and then, you know, on their own hardware. So we actually run some on … The fundamental difference in the way that we've built dots is like, the harness does not run on the machine, you know? And then this is like, maybe people have not realized. It's like, it can connect to as many devices as you want. So it has its own computer. You can actually connect it to your own laptop as well. Over time, you might connect it to like, 10 different devices and it can control it all, a little bit like an octopus. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 21:47. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: Where do agents run?; How autonomous are the agents, and who checks their work? - Context (the archive's note, not their words): Asked whether his Dots agent has access to OpenAI's production code, Sottiaux describes the guardrails and where specialist Dots run. - Passage page: https://the-ai-record.com/passages/openai-dots-production-guardrails-own-hardware-2026 > We took inspiration from Terraform: agent definitions should live in version control and move through the same review and deployment process as the rest of the product. An agent’s instructions, model, tools, and environment are declared in code, and our deployment pipeline registers changes with Laminar. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/ramp-agent-definitions-live-version-control-move-2026 > the key infrastructure for us to spawn a sandboxed container that would run an agent - Who: Monzo - Source: Building Agent Chip: Monzo's in-house agentic tool, Monzo blog, August 13, 2026. https://monzo.com/blog/building-agent-chip - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/monzo-key-infrastructure-us-spawn-sandboxed-container-would-2026 ## What does AI do across the whole company? Organization-wide statements about AI that are not about one function. Topic page: https://the-ai-record.com/topics/ai-across-the-company > AI is transforming how we operate: Over the past three years, AI has helped us reduce cost to serve by nearly 30% and increase ARPAM by 23%, all while improving customer satisfaction. AI has also meaningfully increased internal productivity: we have boosted developer throughput, cut code review time, and more than doubled marketing creative output while reducing production costs. - Who: Chime - Source: Chime Financial, Inc. Q4 and full year 2025 results (Form 8-K, Exhibit 99.1), SEC EDGAR, February 25, 2026, at Fourth Quarter and Full Year 2025 Highlights. https://www.sec.gov/Archives/edgar/data/1795586/000179558626000004/ex991q425.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact?; What does AI cost organizations?; Does AI customer service cut the cost of support? - Passage page: https://the-ai-record.com/passages/chime-ai-reduce-cost-serve-nearly-30-2026 > Our vision for GenAI is that everyone at the Company can use it to boost their productivity and accelerate business impact. To solve the various challenges in our operations, MercadoLibre has numerous proprietary AI solutions, and our teams are much more efficient today. We are using AI to assist our employees in the production of TV commercials, handling of customer service cases, and response to legal demands. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business, AI-Driven Innovation at MercadoLibre. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; Can AI do legal work like contract review? - Passage page: https://the-ai-record.com/passages/mercado-libre-vision-genai-everyone-can-use-boost-2025 > Forward into today, … authors over 10% of all merge PRs across tech. So, that's data, machine learning, that's a back … app, et cetera. Also runs more than 1,800 tasks per day, which is a lot. And it's been actively starting to be adopted by business users as well. So, it's not just coding and tech. It's also trying to answer business questions, writing queries or self-serving anything that a business user may need. - Who: Francesco Galletta, Leads analytics engineering for financial crime; co-leads AI strategy for the data discipline, Monzo, Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 2:55. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How much code does AI write? - Context (the archive's note, not their words): Galletta is describing Monzo's in-house agent, Agent Chip, at the time of the talk. 'Merge PRs' is as spoken in both passes. - Passage page: https://the-ai-record.com/passages/monzo-forward-into-today-authors-over-10-all-merge-prs-across-tech-2026 > And so what we're helping companies figure out is one, how do you build agents so that you can streamline your operations? Almost everything we do at Vercel, we do with agents. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 27:13. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Rauch follows an example of a Slack agent that reads Salesforce for Vercel's account executives. - Passage page: https://the-ai-record.com/passages/vercel-almost-everything-we-do-at-vercel-we-do-with-agents-2026 > At Anthropic, [Claude] is at the center of everything that we do. It's every single business process, every single bit of everything everyone does throughout the day. [Claude] is just at the center of it. … And so, … [Claude] writes the code. It does the code review. It does the security review. It brainstorms and comes up with new product ideas. It aggregates user feedback. It triages incidents. … And … now this is happening not just for engineering. It's happening for product and for design and for marketing and for GTM, for every function. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 11:46. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Cherny is answering the host's question about what his typical day at Anthropic looks like. - Passage page: https://the-ai-record.com/passages/anthropic-claude-center-every-business-process-2026 > AI is a powerful productivity driver for our clients and for IBM. We are transforming our enterprise operations, driving efficiency and cost savings with our Client Zero approach… - Who: IBM (the organization sells AI tools) - Source: IBM Form 10-Q for the quarterly period ended September 30, 2026, SEC (Form 10-Q), October 22, 2026, at Management Discussion, Looking Forward. https://www.sec.gov/Archives/edgar/data/51143/000005114326000095/ibm-20260930.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/ibm-ai-powerful-productivity-driver-client-zero-2026 > In Consulting, AI is both a growth driver and a productivity engine. Our experts are helping clients design and execute AI strategies by leveraging the IBM Consulting Advantage platform, an AI delivery platform designed to implement solutions at scale, transforming how our consultants work. As agents take on more work, delivery becomes faster, more software driven, and more scalable. - Who: IBM (the organization sells AI tools) - Source: IBM Form 10-Q for the quarterly period ended September 30, 2026, SEC (Form 10-Q), October 22, 2026, at Management Discussion, Looking Forward. https://www.sec.gov/Archives/edgar/data/51143/000005114326000095/ibm-20260930.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Passage page: https://the-ai-record.com/passages/ibm-consulting-ai-growth-driver-productivity-engine-2026 > 7.0 billion Agentic Work Units (“AWUs”) delivered to date across Agentforce and Slack, with 3.2 billion in Q2 - Who: Salesforce (the organization sells AI tools) - Source: SEC filing (8-K exhibit: the company's quarterly earnings release), August 26, 2026. https://www.sec.gov/Archives/edgar/data/1108524/000110852426000187/crm-q2fy27xexhibit991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-7-0-billion-agentic-work-units-awus-2026 > We are refining our operating model using a jobs to be done framework, mapping critical work to a combination of highly focused teams and increasingly capable AI agents. Focused teams aligned to top strategic priorities are allowing us to move faster, execute more efficiently, and better serve our community, partners, and advertisers. - Who: Snap - Source: Snap Inc. Form 8-K, Exhibit 99.2: Investor Update, Snap’s Path to Net Income Profitability, SEC EDGAR, April 15, 2026, at Slide 5, Transforming How We Operate to Drive Efficiency. https://www.sec.gov/Archives/edgar/data/1564408/000119312526155861/d36756dex992.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do leaders push AI adoption?; Do organizations reorganize teams around AI? - Passage page: https://the-ai-record.com/passages/snap-refining-operating-model-using-jobs-done-framework-2026 > our engineers use an AI-assisted case log classification tool that organizes documents and categorizes over one million monthly chat conversations. - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-engineers-use-ai-assisted-case-log-classification-2026 > we operate an internal knowledge chatbot, which we call Kiki, that helps employees find information in real time across internal systems. - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-operate-internal-knowledge-chatbot-which-call-kiki-2025 > Wheelhouse is far more sophisticated, and is now doubtless one of the top solo software factories on the planet, comfortably producing 2000–4000 commits a week on a light multi-machine setup. Its agents work around the clock: building, reviewing, deploying game features, negotiating on Slack with my game admins, interacting on Discord with players, exchanging emails with my accountant and my lawyers. - Who: Steve Yegge, Independent; builds and runs Wheelhouse, a solo multi-agent software factory (2026), Steve Yegge (independent) (the organization sells AI tools) - Source: The Counting House, Medium (Steve Yegge), October 27, 2026, at Wheelhouse Recap. https://steve-yegge.medium.com/the-counting-house-steve-yegge-83f5145cc35b - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/steve-yegge-wheelhouse-top-solo-software-factories-2026 > So again, all of these are examples of how our technology team is using it, but even better than that, we've observed a lower barrier for non-specialists in the company to contribute to our systems. … Just call agent chip, get a draft that works that is already tested in staging and then and only then involve an engineer to make sure we get the changes reviewed and deployed properly. So this doesn't change the rigor to which we develop software at Monzo, but this does allow a number of non-specialists to also start contributing more closely to the team. - Who: Felippe Felisola Caso, Monzo (conference affiliation, 2026), Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 12:07. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; Can people outside engineering ship code with agents? - Context (the archive's note, not their words): Felisola Caso is describing how people outside engineering use Monzo's in-house agent, Agent Chip, in Slack. - Passage page: https://the-ai-record.com/passages/monzo-lower-barrier-non-specialists-contribute-draft-tested-staging-engineer-2026 > Anthropic uses Claude to write software far more than almost any other company - Who: Anthropic (the organization sells AI tools) - Source: Anthropic (own site), October 13, 2026. https://www.anthropic.com/engineering/redesigning-source-control - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/anthropic-anthropic-uses-claude-write-software-far-more-2026 > MCP Gateway unlocked a scalable, fast, and consistent path to building AI agents within Uber and is currently hosting over 800 MCP servers and over 5000 tools. - Who: Alok Srivastava, Principal Engineer, Business Platform, Uber - Source: Uber engineering blog, October 1, 2026. https://www.uber.com/us/en/blog/designing-mcp-gateway/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-mcp-gateway-unlocked-scalable-fast-consistent-path-2026 > Since May 2026, 3,000+ Stripes have created 12,000 prototypes, and 25% of Stripes have created at least one prototype in Harbor. - Who: Cristian Rivera, Staff engineer, Web Presence and Platform team, Stripe - Source: Stripe developer blog, September 14, 2026. https://stripe.dev/blog/harbor-stripes-ai-assisted-prototyping-tool - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-since-may-2026-3-000-stripes-created-2026 > all of a sudden all the design docs and all the Slack messages and all the emails have gotten three or four times as long and … gotten two or three times as many emails and they say basically as much as they did before. - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 04:16. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/figma-docs-slack-emails-three-four-times-as-long-say-same-2026 > routinely running more than 1800 tasks every day - Who: Monzo - Source: Building Agent Chip: Monzo's in-house agentic tool, Monzo blog, August 13, 2026. https://monzo.com/blog/building-agent-chip - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/monzo-routinely-running-more-than-1800-tasks-every-2026 > And across Stripe, more than 5,000 sessions everyday center on data analysis. - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-across-stripe-more-than-5-000-sessions-2026 > The software factory starts with signals from the outside world: bug reports, internal conversations, customer feedback, business requirements. These signals get triaged and turned into planned changes. These changes are built, tested, reviewed, secured, shipped, and monitored. Monitoring that deployed software generates more signals. The entire system is a continuous feedback loop. - Who: Factory (the organization sells AI tools) - Source: Factory 2.0: From coding agents to software factories, Factory (own site), June 15, 2026. https://factory.ai/news/software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/factory-software-factory-starts-signals-outside-world-2026 > More than 300 are active every week. - Who: Andrii Yakovenko, Author describing Intercom's internal Claude Code rollout, Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: We gave Claude Code to everyone at Intercom, Intercom Fin blog, March 23, 2026. https://ideas.fin.ai/p/we-gave-claude-code-to-everyone-at - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/intercom-more-than-300-active-every-week-2026 > GitHub Copilot populates approximately 80% of the issue's metadata automatically - Who: GitHub (the organization sells AI tools) - Source: GitHub blog, March 12, 2026. https://github.blog/ai-and-ml/github-copilot/continuous-ai-for-accessibility-how-github-transforms-feedback-into-inclusion/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/github-github-copilot-populates-approximately-80-issues-metadata-2026 > GitHub Copilot then posts a comment on the issue containing: A summary of the problem and user impact - Who: GitHub (the organization sells AI tools) - Source: GitHub blog, March 12, 2026. https://github.blog/ai-and-ml/github-copilot/continuous-ai-for-accessibility-how-github-transforms-feedback-into-inclusion/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/github-github-copilot-then-posts-comment-issue-containing-2026 > We've spent the last few months using Custom Agents internally at Notion. We now have more agents than employees. - Who: Notion (the organization sells AI tools) - Source: Notion (own site), February 24, 2026. https://www.notion.com/blog/introducing-custom-agents - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/notion-weve-spent-last-few-months-using-custom-2026 > And so for a lot of problems, a coding agent plus deployment over … has allowed us to delete four or five vendors. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 42:56. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Rauch is answering which software Vercel builds itself with agents rather than buys. - Passage page: https://the-ai-record.com/passages/vercel-coding-agent-plus-deployment-allowed-us-to-delete-four-or-five-vendors-2026 > the way that we work at Vercel is I can go into my agent that we deploy on Slack … Hey, tell me about Molly and how frequently she deploys on Vercel … I'm gonna have a meeting with her and I'm her account executive. And give me a quick report on some of the problems that she's reported. So I never actually went to Salesforce. I just went to the agent and the agent went to Salesforce. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 26:48. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Rauch is describing how Vercel's own staff use an agent, using the host as the example account. - Passage page: https://the-ai-record.com/passages/vercel-agent-on-slack-account-executive-report-never-went-to-salesforce-2026 > "This intelligence layer means I now have total information," he said. "I've cut my meetings in half." - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: Entrepreneur (business news magazine), reporting remarks at the Skift Global Forum, September 25, 2026. https://www.entrepreneur.com/business-news/how-airbnb-ceo-brian-chesky-created-ai-brain-cut-meetings-in-half - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-intelligence-layer-means-i-now-total-information-2026 > Probably the most popular agent at Vercel used by the entire company is D0. D0 is a data analyst agent that everyone at Vercel can access via Slack. Questions that used to take a week to ticket and get an answer back to from the data science team can now be answered in less than a minute. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 12:01. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Vercel's COO describes D0, an internal agent in Slack, in the part of the talk on why agents need a strong data foundation. - Passage page: https://the-ai-record.com/passages/vercel-d0-data-analyst-agent-week-to-minute-2026 > we now have automated, actually, a decent number of core functions within the company. For example, our customer support agent, and you can imagine at Vercel, a lot of our cases are complicated and highly technical, now handles 93% of total caseload. Similarly, our content agent is used heavily by the marketing organization and did 96% of our major content updates last quarter. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 00:54. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How much customer support does AI resolve? - Context (the archive's note, not their words): Opening a SaaStr AI Annual 2026 talk, Vercel's COO lists functions the company says it has automated after ten months of building agents. - Passage page: https://the-ai-record.com/passages/vercel-support-agent-handles-93-percent-content-agent-96-2026 > we now have a number of distilled models where we take a teacher model, usually a big frontier model, then teach a specific use case to a smaller model - Who: Harley Finkelstein, President, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/shopify-now-number-distilled-models-where-take-teacher-2026 > In a new world, an engineer can spin up 10 agents and those agents can work 24 seven. I mean, I'm kind of exaggerating a little bit. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 7, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/airbnb-new-world-engineer-can-spin-up-10-2026 > First, because we are remote-first, decisions, code, designs, and progress are recorded and machine readable, giving AI a useful operating model to learn from and act on. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To Our Shareholders. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/block-remote-first-machine-readable-decisions-code-designs-2026 > normally that would have taken 40 or 50 people about a year to do. And we took five really smart people, AI forward thinking people building on agentic coding tools. And those five people rebuilt it in 65 days. - Who: Amazon (the organization sells AI tools) - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), April 2026. https://transcripts.platformaeronaut.com/transcripts/AMZN-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Do organizations reorganize teams around AI?; Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/amazon-normally-would-taken-40-50-people-about-2026 > In 2026, we plan to accelerate even further. Our proprietary technology platform, ChimeCore, has significantly increased our product velocity, we’re scaling our Enterprise channel, and we’re embedding AI across the platform to drive durable growth and profitability. - Who: Chris Britt, CEO and Co-founder, Chime - Source: Chime Financial, Inc. Q4 and full year 2025 results (Form 8-K, Exhibit 99.1), SEC EDGAR, February 25, 2026, at Press release opening. https://www.sec.gov/Archives/edgar/data/1795586/000179558626000004/ex991q425.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/chime-embedding-ai-across-platform-drive-durable-2026 > We have accelerated AI and data transformation with partnerships with Open AI, AWS and Accenture. - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), February 13, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/13022026/nwg-company-announcement.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-accelerated-ai-data-transformation-partnerships-open-ai-2026 > Using Generative AI to enhance detection of unauthorized trading activity to improve FX trade surveillance in Markets - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended March 31, 2025, SEC EDGAR, May 8, 2025, at Management’s Discussion and Analysis, Transformation, Progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000086/c-20250331.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How does AI help keep a product secure? - Passage page: https://the-ai-record.com/passages/citigroup-generative-ai-detect-unauthorized-trading-fx-2025 > So it's probably no secret to you guys that, you know, news from Asia and Korean media outlets has been super important this year because semis, AI, have been a huge, huge theme in the market, huge theme. And I think this is an area where AI … poised to have a lot of impact is reading Korean language news better than our completely non-Korean speaking staff, right? I think I have a good thesis that this might be a good use case. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 05:08. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Long is walking through one example, interpreting news, of the first step in his framework. - Passage page: https://the-ai-record.com/passages/coatue-agents-reading-korean-news-better-than-staff-2026 > And just as importantly, we have a culture that is super, super rigorous and built from the ground up to capture all this insight. For decades, we filed every single note about every single trade, about every … decision we made or didn't make, about what went well and what didn't go well. And that's poised us to expose us to AI. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 03:00. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Long is describing the firm's internal context, the records an AI system can draw on. - Passage page: https://the-ai-record.com/passages/coatue-decades-of-filed-notes-poised-us-for-ai-2026 > Something that we're doing very recently that's been creating a lot of value is also triaging data quality alerts and checking for backend changes that can create problems. - Who: Felippe Felisola Caso, Monzo (conference affiliation, 2026), Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 11:26. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI fix bugs and handle incidents? - Context (the archive's note, not their words): Felisola Caso is describing a recent use of Monzo's in-house agent, Agent Chip. - Passage page: https://the-ai-record.com/passages/monzo-triaging-data-quality-alerts-checking-backend-changes-create-problems-2026 > Humans helped prioritize 38 ideas, then agents implemented the changes, trained models, debugged failed runs, and evaluated the results. - Who: Ryan Drapeau, Machine learning engineer, Intelligent Commerce, Stripe - Source: Stripe developer blog, October 19, 2026. https://stripe.dev/blog/scaling-radar-fighting-fraud-with-agent-assisted-machine-learning - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/stripe-humans-helped-prioritize-38-ideas-then-agents-2026 > 94% of client interactions are now digital. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), October 14, 2026. https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-94-client-interactions-now-digital-2026 > Very often, we found that the role of the engineers was to tell Claude “no.” - Who: Anthropic (the organization sells AI tools) - Source: Anthropic (own site), October 13, 2026. https://www.anthropic.com/engineering/redesigning-source-control - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do engineering and product jobs change when agents write the code? - Passage page: https://the-ai-record.com/passages/anthropic-very-often-found-role-engineers-tell-claude-2026 > One week in May, we burned $1.5 million in tokens. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What does AI cost organizations? - Passage page: https://the-ai-record.com/passages/ramp-one-week-may-burned-1-5-million-2026 > When we moved our own work onto it, our bill fell 30% overnight. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-when-moved-own-work-onto-bill-fell-2026 > AI-native development processes at Atlassian have cut issue cycle time by nearly half, from 11.5 days to 6.7 days. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian (own site), October 1, 2026. https://www.atlassian.com/blog/ai-at-work/ai-sdlc-transformation-playbook - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/atlassian-ai-native-development-processes-atlassian-cut-issue-2026 > In internal benchmarks, agents with access to the Teamwork Graph delivered 44% more accurate results while using 48% fewer tokens compared to agents operating without it. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian (own site), October 1, 2026. https://www.atlassian.com/blog/ai-at-work/ai-sdlc-transformation-playbook - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/atlassian-internal-benchmarks-agents-access-teamwork-graph-delivered-2026 > the vanilla setup required roughly 20 days of engineering effort, compared to 4 days when using the reusable prompts. - Who: David Dunne, Software engineer, Local Payment Methods, Stripe - Source: Stripe developer blog, September 30, 2026. https://stripe.dev/blog/stripes-payment-method-factory-orchestrating-agents-for-repeated-custom-integrations - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need?; Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/stripe-vanilla-setup-required-roughly-20-days-engineering-2026 > Assisted by AI, the team was able to go from a proof of concept to a fully rebuilt native app published in the app stores in just 12 weeks. - Who: Max Da Silva, VP of Engineering, Growth, Shopify - Source: Shopify engineering blog, September 10, 2026. https://shopify.engineering/shop-app-migration - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/shopify-assisted-ai-team-able-go-proof-concept-2026 > we've decided we always, every PR description will start with something like that, something that I wrote by hand … could be very short to describe what this is doing and what this is doing. - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 13:41. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/figma-pr-description-starts-with-hand-written-ai-description-after-2026 > We've already shipped 75 integrations this way, reducing work that took weeks or months down to hours - or minutes for the customer just trying to get unblocked. - Who: Shubhayan Srivastava, Software Engineer Intern (summer 2026), Ramp - Source: Integrations That Write Themselves, Ramp engineering blog, August 14, 2026. https://builders.ramp.com/post/integrations-that-write-themselves - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-weve-already-shipped-75-integrations-way-reducing-2026 > Anyone could build and deploy workflow-specific agents that could use tools. Over 4,000 agents were built using this system. - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Who builds an organization's agents: every team, or a central group? - Passage page: https://the-ai-record.com/passages/stripe-anyone-could-build-deploy-workflow-specific-agents-2026 > ensures human oversight, transparency, and accountability for all outcomes - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-ensures-human-oversight-transparency-accountability-all-outcomes-2026 > The last bit here is that the model is learning to train the next model. And it's a little hard to wrap your brain around. The way I like to think about it is every time you release a new version of this intelligence, then you can create or distill these derivative versions that you use to speed up other parts of the training process, … - Who: Lee Robinson, Machine Learning Engineer, Model Behavior, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Recursive Model Improvement, AI Engineer (YouTube), July 15, 2026, at 18:30. https://www.youtube.com/watch?v=q4Tr-DknG2M - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Robinson is closing a talk on how Cursor trains its models and uses earlier models to speed up later training. - Passage page: https://the-ai-record.com/passages/cursor-model-learning-to-train-next-model-2026 > Now with this, we had successfully reached stage five where engineers are delegating complete tasks to agents. … able to produce shippable results without human hand holding. In fact, this wasn't even limited to engineers anymore. Anyone at the company could … or implement a new feature. They didn't even need GitHub. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 16:12. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; Can people outside engineering ship code with agents? - Context (the archive's note, not their words): Jones is describing Block's top stage of agent use and who could use it. - Passage page: https://the-ai-record.com/passages/block-stage-five-anyone-at-company-fix-bug-feature-2026 > 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. - Who: Brian Armstrong, CEO, Coinbase - Source: X post by unnamed account, X post by Brian Armstrong, CEO, June 27, 2026. https://api.fxtwitter.com/brian_armstrong/status/2070670644577280109 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Passage page: https://the-ai-record.com/passages/coinbase-91-employees-were-never-hitting-their-usage-2026 > The agent runs our cleanup skill and one-shots it nearly every time, because the task is well scoped and always the same shape. I just review and merge. - Who: Linear - Source: Linear (own site), June 12, 2026. https://linear.app/now/linear-agent-bug-fix - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/linear-agent-runs-cleanup-skill-one-shots-nearly-2026 > Designers work, I would say like 50/50 in Figma and code at this point … of them do code. All of them … do push to production. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 1:18:40. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design?; Can people outside engineering ship code with agents?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Zakariasson is answering an audience question about how product managers, designers and engineers work together at Cursor. - Passage page: https://the-ai-record.com/passages/cursor-designers-push-to-production-2026 > I used coding agents to build agents that automated part of my job. - Who: GitHub (the organization sells AI tools) - Source: GitHub blog, March 31, 2026. https://github.blog/ai-and-ml/github-copilot/agent-driven-development-in-copilot-applied-science/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Who builds an organization's agents: every team, or a central group? - Passage page: https://the-ai-record.com/passages/github-i-used-coding-agents-build-agents-automated-2026 > Zendesk announced its intent to acquire Forethought on March 11, 2026. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (the company's own press release), March 26, 2026. https://www.zendesk.com/newsroom/press-releases/zendesk-completes-acquisition-of-forethought/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-zendesk-announced-its-intent-acquire-forethought-march-2026 > we built a central internal MCP server called Toolshed, which hosts more than 400 MCP tools - Who: Alistair Gray, Software engineer, Leverage team, Stripe - Source: Minions: Stripe’s one-shot, end-to-end coding agents, Stripe developer blog, February 9, 2026. https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/stripe-built-central-internal-mcp-server-called-toolshed-2026 > with the virtual assistant having reduced calls into the IT service desk by 50% - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), August 20, 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-virtual-assistant-having-reduced-calls-into-service-2025 > I think it's a power distribution. So, like, there's probably, like, 60% of the work … We can do really well right now, and you know, from now on, it's like a cost and latency thing. But then there's the rest of the the thing where, especially for, like, very complex analysis, we we still need more intelligence. - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 7:30. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/legora-60-percent-work-done-well-power-distribution-2026 > I think I've taken a lot of attention now to like, how are agents? Like, agent is a customer, and how are they using our platform, and are the agents happy as well? - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 27:35. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in product and design? - Context (the archive's note, not their words): Rauch is describing where he pays attention at Vercel, alongside customers and the internal employee experience. - Passage page: https://the-ai-record.com/passages/vercel-agent-is-a-customer-how-are-they-using-our-platform-happy-2026 > Inside of Shopify, the amount of people I know who really write code is vanishingly small now. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: The Knowledge Project (podcast by Shane Parrish of Farnam Street, independent), transcript on fs.blog, September 10, 2026. https://fs.blog/knowledge-project-podcast/tobi-lutke-3/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do engineering and product jobs change when agents write the code? - Passage page: https://the-ai-record.com/passages/shopify-inside-shopify-amount-people-i-know-who-2026 > basically it just means can you create sort of like a, as automated as possible, a workflow for developing debugging. Mm-hmm. Merging, reviewing, and maintaining a code base - Who: Simon Last, Co-founder and head of AI, Notion (the organization sells AI tools) - Source: Notion's Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future (Latent Space), Latent Space (independent AI engineering podcast), April 15, 2026. https://www.latent.space/p/notion - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/notion-basically-just-means-can-you-create-sort-2026 ## Can AI approve code changes without a person? Where organizations let an AI reviewer approve or merge a change on its own, which changes still need a person, and what happened after. Topic page: https://the-ai-record.com/topics/ai-approves-code > The change sets the tier, not the author. A classifier reads the change and picks its risk tier. … An agent's approval never counts as a person's, either. Automated review can clear a low-risk change, but it can't fill a slot that calls for a human. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Layer by layer: 5. Human approval, scaled by risk. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Context (the archive's note, not their words): From the section on human approval, where the post sets three risk tiers for changes to Claude Code's repos, needing no, one or two human approvals. - Passage page: https://the-ai-record.com/passages/anthropic-agent-approval-never-counts-as-persons-2026 > The builder verifies with targeted checks and runs a quick self-review. CI runs the deterministic harness. The deep review runs on request in an independent session, with the verdict returned to whoever asked. The author's watch loop handles threads and formal re-requests. Humans own the required approvals, and bot or agent verdicts do not satisfy them. - Who: TradersPost - Source: The AI Review Bot We Reverted, TradersPost blog, October 13, 2026, at What we do instead today. https://blog.traderspost.io/article/the-ai-review-bot-we-built-and-reverted - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/traderspost-builder-verifies-ci-runs-deep-review-2026 > You get a pretty high degree of confidence on merge. Today, almost a fifth of all the pull … are merged without any human review or without any human testing. … that is a number that we care a lot about and want to bring higher and higher, of course, within the guardrails of producing really high quality code. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 11:50. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How is AI used in code review? - Passage page: https://the-ai-record.com/passages/greptile-almost-fifth-prs-merged-without-human-review-2026 > Auto-approve is not auto-merge. When Diff Vader approves, it posts an APPROVE review, the same thing a human reviewer clicking “approve” does. It does not push the merge button. The engineer merges (or GitHub’s native auto-merge does, if they enabled it), and every existing gate—CI, branch protection, required-reviewer counts—still applies. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at What an engineer actually sees. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How is AI used in code review? - Passage page: https://the-ai-record.com/passages/rewind-auto-approve-not-auto-merge-bot-approve-review-gates-2026 > So we built a system that fixes the allocation. It’s called Diff Vader: an AI reviewer that looks at every PR across our orgs and, for a carefully bounded set of low-risk changes, becomes the only reviewer needed, freeing human reviewers to spend their judgment where judgment actually matters. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at Opening. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/rewind-diff-vader-ai-reviewer-only-reviewer-needed-low-risk-2026 > It declares the hard rules the verdict engine is not allowed to override no matter what the council says: paths that touch authentication, authorization, secrets, encryption, and anything that changes who is allowed to approve future changes always require a human. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at The council: many specialists, one question. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How does AI help keep a product secure? - Passage page: https://the-ai-record.com/passages/rewind-hard-rules-verdict-engine-cannot-override-auth-secrets-2026 > we had code owners in our code base and they were kind of right most of the time, like 80% of the time. But for these 20% of the time, they cost a lot of bottlenecks for us internally. … So what we started doing was building this agentic code owner thing. And what it essentially does is look in at PRs and check in like, first of all, what's the risk of this? What's the risk level? … And if it is a low risk, it can just approve the PR … - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 27:55. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Context (the archive's note, not their words): Zakariasson is describing an internal Cursor tool that replaces some human code-owner reviews of pull requests. - Passage page: https://the-ai-record.com/passages/cursor-agentic-code-owner-approves-low-risk-prs-2026 > We started with a controlled pilot of over 100 PRs through the AI approval pipeline. The results: zero reverts of AI-approved PRs, and a 6–16x improvement in time-to-approval at the 75th percentile. Since then, the system has scaled significantly. In the first four weeks of broader rollout, 497 PRs went fully autonomous, with Claude writing the code and our AI approval system reviewing, approving, and shipping to production. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at The data. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/intercom-controlled-pilot-100-prs-zero-reverts-2026 > Today, over 93% of our pull requests (PRs) across our two main codebases are Agent-driven. And over 19% are auto-approved with no human reviewer in the loop. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at Introduction. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/intercom-over-93-prs-agent-driven-19-2026 > The AI agent can approve the review, but a person always clicks the merge button. That merge is the traceable approval event for SOC 2: the agent handles the review workload, the human controls what ships. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at Governance. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How is AI used in code review?; How do banks and other regulated organizations use AI within the rules? - Passage page: https://the-ai-record.com/passages/ona-agent-approve-review-person-always-clicks-merge-2026 > First, engineers cannot self-classify a change as low-risk. An Ona Automation evaluates every PR against objective criteria automatically, which removes the temptation to game the system and makes the boundary auditable. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at The policy. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/ona-engineers-cannot-self-classify-change-low-risk-2026 > Our engineers waited a median of 2 hours 49 minutes for their first PR approval. Four weeks after we let an AI agent approve low-risk PRs, that dropped to 3.8 minutes, lead time fell 74%, and deploys tripled, all with the same size team. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at Opening. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Does AI make developers more productive?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/ona-engineers-waited-median-2-hours-49-minutes-first-approval-2026 > We run Code Review on nearly every PR at Anthropic. Before, 16% of PRs got substantive review comments. Now 54% do. It won't approve PRs — that's still a human call — but it closes the gap so reviewers can actually cover what's shipping. - Who: Anthropic (the organization sells AI tools) - Source: Bringing Code Review to Claude Code, claude.com (Anthropic), March 9, 2026, at Managing the review bottleneck. https://claude.com/resources/articles/code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/anthropic-run-code-review-nearly-every-pr-16-54-2026 > High-risk PRs that the system flagged for human review took about a day to approve. That's longer than the baseline. This is correct behavior. These are architectural changes, infrastructure modifications, and breaking changes. They need time and human judgment. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Results: Safety validation. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How is AI used in code review? - Passage page: https://the-ai-record.com/passages/customer-io-high-risk-prs-flagged-human-review-took-day-2025 > We taught AI to approve pull requests for merging. After 500 AI-approved PRs: zero production incidents, and merge time dropped from 3.5 hours to 33 minutes. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Article summary. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/customer-io-taught-ai-approve-pull-requests-500-approved-2025 > right now we still require an engineer to review every PR … goes in. … Once the more confidence we get in our risk scoring and we can also back test it with years of data, right? We have all of this data and the more confidence we have in our AI code reviews, then maybe for some class of low risk PRs, we could enable, we call it green lane like auto landing. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 59:37. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review?; Do engineers still read the code agents write? - Context (the archive's note, not their words): Answering the host's question whether human code review will always exist. - Passage page: https://the-ai-record.com/passages/uber-still-require-engineer-review-every-pr-green-lane-2026 > The reviewer agent surfaces issues and inconsistencies; the human makes the final call on quality and correctness. - Who: Warp (the organization sells AI tools) - Source: Set up your software factory, Warp docs, October 28, 2026, at 5. Human review: merge the PR. https://docs.warp.dev/guides/agent-workflows/set-up-a-software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/warp-reviewer-agent-surfaces-issues-inconsistencies-human-2026 > So the goal is that most pull requests in the Claude Code repos … merge with at most one human in the loop, who can be the author, with no increase in reverts or incidents. To measure human time, we track human approvals per PR (median and 95th percentile), human comments, and how long a PR waits on a person. Revert and incident rates are the guardrail: if they rise, we haven't saved human time, we've only moved it. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at The scarce resource: human time and attention. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/anthropic-goal-at-most-one-human-in-loop-per-pr-2026 > We don't yet have circuit breakers that tighten approval when those numbers move; engineers here want them before more code moves to zero-human merges. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Layer by layer: 7. Fast detection, fast rollback. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Context (the archive's note, not their words): "Those numbers" are the revert and incident rates: the sentence before says noticing a bad merge still depends on the author, on-call, and the revert and incident dashboards. - Passage page: https://the-ai-record.com/passages/anthropic-no-circuit-breakers-before-zero-human-merges-2026 > Currently, we still have humans approving the code, but we see a short path in the near future to a percentage of our code landing automatically, having automatic approvals, right? - Who: Will Bond, Staff Engineer, AI Foundations team, Development Platform organization, Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 11:39. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How is AI used in code review? - Context (the archive's note, not their words): Will Bond is describing where Uber expects code review to go as it moves to an agentic software development life cycle. - Passage page: https://the-ai-record.com/passages/uber-currently-we-still-have-humans-approving-2026 > cortex went all-in: the AI writes the code, the AI reviews the code. What humans keep their hands on is "tuning the prompts and guidelines themselves" -- not making decisions inside each individual PR, but watching the system from above and adjusting. - Who: Ryosuke Tsuji, CTO, airCloset - Source: Human-on-the-Loop: AI Reviewing AI PRs at cortex -- 769 PRs/month while raising the quality bar, DEV Community, May 26, 2026, at cortex's answer: move the reviewer role to AI as well. https://dev.to/ryantsuji/human-on-the-loop-ai-reviewing-ai-prs-at-cortex-769-prsmonth-while-raising-the-quality-bar-4lh5 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/aircloset-cortex-went-all-ai-writes-code-2026 > Every one of those 769 PRs had an AI reviewer as the first reviewer, with an average of 10.8 review-fix loop iterations per PR (max 56). 1 in 5 merged within 10 minutes, roughly half within 30 minutes. What humans do now is look at review outcomes and tune the review prompt and the guidelines themselves -- this is human-on-the-loop, not human-in-the-loop. Humans operate on the policy layer, not the execution layer. - Who: Ryosuke Tsuji, CTO, airCloset - Source: Human-on-the-Loop: AI Reviewing AI PRs at cortex -- 769 PRs/month while raising the quality bar, DEV Community, May 26, 2026, at Start with last month's numbers. https://dev.to/ryantsuji/human-on-the-loop-ai-reviewing-ai-prs-at-cortex-769-prsmonth-while-raising-the-quality-bar-4lh5 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/aircloset-every-769-prs-ai-reviewer-first-2026 > we are allowing the bot to merge that directly into our main branch because we have integration tests, unit tests, etc., canary tests. - Who: Xiucheng Quek, Software engineer, Genentech - Source: Background Agents for Genomics and Cloud Ops, YouTube video of a talk or interview (speaker in notes), May 7, 2026. https://www.youtube.com/watch?v=GXDtw0EmoX0 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/genentech-allowing-bot-merge-directly-into-main-branch-2026 > Our Agent is strict. It won’t approve large PRs. If a change is too big, too complex, or too broad in scope, it flags it and requires it to be broken down. Creating a direct, positive incentive for engineers to ship smaller, more incremental, well-scoped changes. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at Not just faster; safer. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/intercom-agent-strict-wont-approve-large-prs-2026 > The bias is explicitly toward approval, meaning a single warning in an otherwise clean MR still gets approved_with_comments rather than a block. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at The coordinator helps keep things focused. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/cloudflare-bias-explicitly-toward-approval-meaning-single-2026 > That is fast enough that most engineers see the review comment before they have finished context-switching to another task. The metric we’re the proudest about, though, is that engineers have only needed to “break glass” 288 times (0.6% of merge requests). - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at Show me the numbers! / The overview. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/cloudflare-metric-we-re-proudest-though-engineers-2026 > We've been running this system internally across tens of thousands of merge requests. It approves clean code, flags real bugs with impressive accuracy, and actively blocks merges when it finds genuine, serious problems or security vulnerabilities. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at Introduction. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/cloudflare-running-tens-thousands-merge-requests-approves-clean-2026 > Most AI code review tools stop at comments. They suggest improvements, flag issues, and point out problems. We built a system that approves PRs for merging. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Problem: The cognitive weight of waiting. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/customer-io-most-ai-code-review-tools-stop-comments-approves-2025 > we've merged more than 2.5 million automated maintenance PRs, the vast majority auto-merged with no human in the loop. - Who: Spotify - Source: Coding Is No Longer the Constraint: Scaling Developer Experience to Teams and Agents at Spotify, Spotify engineering blog, June 3, 2026. https://engineering.atspotify.com/2026/6/code-with-claude-coding-is-no-longer-the-constraint - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/spotify-weve-merged-more-than-2-5-million-2026 ## What does AI cost organizations? Spending on models, tokens and tools, and how organizations budget for it. Topic page: https://the-ai-record.com/topics/ai-cost-and-spend > I am spending the equivalent of $122k/month of API token spend, or about $4,000 per day, using 21 Claude Max accounts, a number that has been growing steadily at 2 per week. - Who: Steve Yegge, Independent; builds and runs Wheelhouse, a solo multi-agent software factory (2026), Steve Yegge (independent) (the organization sells AI tools) - Source: Fences, not Sandboxes, yegge.ai (Steve Yegge), August 24, 2026. https://yegge.ai/essays/fences-not-sandboxes/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/steve-yegge-spending-equivalent-122k-month-api-token-2026 > … speaking of costs, the cost of AI is too damn high. … since 2024, our costs have gone up at least 6x. … again, no question that it's had positive impact, but it's gone from something that I can self fund using my own budget to something that I need to ask permission from … the CFO. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 35:56. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Context (the archive's note, not their words): Chadha is describing how the cost of AI tools at Uber has grown and what it changed about who approves the spending. - Passage page: https://the-ai-record.com/passages/uber-cost-ai-too-damn-high-costs-gone-up-6x-since-2024-2026 > If you haven't spent at least $1,000 on tokens today per human engineer, your software factory has room for improvement - Who: StrongDM - Source: How StrongDM's AI team build serious software without even looking at the code, Simon Willison's Weblog, February 7, 2026. https://simonwillison.net/2026/feb/7/software-factory/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/strongdm-haven-t-spent-least-1-000-tokens-2026 > On R&D, the majority of our internal AI spend is allocated here, so you've seen a modest uptick in year-over-year growth. - Who: Jeff Hoffmeister, Chief Financial Officer, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-r-d-majority-internal-ai-spend-allocated-2026 > We were able to hold gross margins at a relatively consistent level quarter over quarter while Sidekiq usage scaled, which reflects some cost efficiencies in support - Who: Jeff Hoffmeister, Chief Financial Officer, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How much customer support does AI resolve?; Does AI customer service cut the cost of support? - Passage page: https://the-ai-record.com/passages/shopify-were-able-hold-gross-margins-relatively-consistent-2026 > Inside the business, we also use AI, and that is more closer to the $10 million range internally. - Who: Duolingo - Source: earnings call transcript via Webull (third-party transcript site), August 2026. https://www.webull.com/news/15392596720296960 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/duolingo-inside-business-also-use-ai-more-closer-2026 > We decreased spending on external marketing suppliers, such as translation, production, CRM and social agencies, by $7 million in the first quarter of 2024 (as compared to the first quarter of 2023) and we attribute 37% of this decrease to the use of AI - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/klarna-decreased-spending-external-marketing-suppliers-such-translation-2026 > GLOWC, an internal framework used by engineers at Airbnb, makes it easy to fine-tune open-weight language (and vision-language) models into fast, accurate classifiers. In this blog post, we put one of these classifiers head-to-head with open-weight and proprietary models and find that it matches frontier-level quality in under 100ms at roughly 1% of the cost. - Who: Airbnb - Source: Frontier quality at 1% of the cost: a production benchmark of LLM classifiers, Airbnb Tech blog, November 3, 2026, at Introduction. https://airbnb.tech/ai-ml/frontier-quality-at-1-of-the-cost-a-production-benchmark-of-llm-classifiers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations keep AI spending in check?; How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/airbnb-glowc-fine-tuned-classifier-matches-frontier-quality-2026 > Proprietary models only win at low traffic: below roughly 15 to 30 RPM, Claude Sonnet 5.5 and GPT-6.1 Sol are cheaper than a dedicated GPU. - Who: Airbnb - Source: Frontier quality at 1% of the cost: a production benchmark of LLM classifiers, Airbnb Tech blog, November 3, 2026, at Cost at scale. https://airbnb.tech/ai-ml/frontier-quality-at-1-of-the-cost-a-production-benchmark-of-llm-classifiers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check?; How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/airbnb-proprietary-models-only-win-low-traffic-2026 > Running several agents in parallel burns tokens quickly, so set budgets and keep an eye on spending from day one. - Who: Noah Blauensteiner, Author, Xebia articles, Xebia (the organization sells AI tools) - Source: Shining Some Light in Dark Agentic Software Factories, Xebia articles, August 12, 2026, at What is an Agentic Software Factory?. https://xebia.com/articles/shining-some-light-in-dark-agentic-software-factories/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Passage page: https://the-ai-record.com/passages/xebia-running-several-agents-parallel-burns-tokens-2026 > That expertise on use cases is also very valuable to Frontier Labs. We find that our evals and our Early Access Program partnerships actually help us a lot with Frontier Labs and it's something that we can exchange instead of extraordinary large commits. And I don't think the discount is ever worth the lost in optionality. - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 14:39. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Context (the archive's note, not their words): Sachs is describing what Notion offers AI model labs in place of large spending commitments; 'that expertise' refers to Notion's evaluations of models on its own use cases. - Passage page: https://the-ai-record.com/passages/notion-evals-early-access-exchange-instead-large-commits-2026 > So the concrete example here is we do security scanning after on our PRs in local and after our PRs even land because they will always find some things that are real that we have overlooked. And they sort of beat humans on the … code review. And it's expensive. It … us, I think like five bucks a PR or something like that to run all the checks that we want. But that's where we made an explicit decision that it's worth it. - Who: Greg Pstrucha, Developer, Sentry, Sentry (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 36:59. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How does AI help keep a product secure?; Does AI code review catch real problems?; What does an AI code review cost? - Context (the archive's note, not their words): Pstrucha is answering the moderator's question: 'when does a loop pay for itself, and how often is this actually the case?' - Passage page: https://the-ai-record.com/passages/sentry-security-scanning-five-dollars-a-pr-explicit-decision-worth-it-2026 > Before I trusted this workflow, I back tested the scoring against months of clean up history. Then I ran it live. Seven for seven PRs with green CI. Now it runs daily on my laptop and hands work back for me to review. … $1.26 a pull request. A backlog of around 520 flags a year runs for under $700. If you had engineers working on this by hand, at least $26,000 that you're looking at. - Who: Andrew Orobator, Senior Software Engineer, Reddit - Source: Scale the Judgment, Not the Model, AI Engineer (YouTube), June 30, 2026, at 11:59. https://www.youtube.com/watch?v=6MudaeKdBSk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do organizations measure AI's impact? - Context (the archive's note, not their words): The speaker is a Reddit Android engineer. In the talk he describes this agent as running in his local workflow and on his laptop, and says some of his other practices are from his personal side projects. The $26,000 is his own estimate of manual cost, and the PR count is a small sample. - Passage page: https://the-ai-record.com/passages/reddit-flag-agent-seven-for-seven-green-ci-1-26-per-pull-request-2026 > For the curious: reviews average around $0.73 per PR, with significant variance. Small dependency bumps cost cents; large multi-file PRs with many substantive findings can push past twelve dollars. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at The Death Star. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; What does an AI code review cost? - Passage page: https://the-ai-record.com/passages/rewind-reviews-average-73-cents-per-pr-twelve-dollars-2026 > On the cost side, the average review costs $1.19 and the median is $0.98. The distribution has a long tail of expensive reviews – massive refactors that trigger full-tier orchestration. The P99 review costs $4.45, which means 99% of reviews come in under five dollars. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at Show me the numbers! / The overview. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; What does an AI code review cost? - Passage page: https://the-ai-record.com/passages/cloudflare-cost-side-average-review-costs-1-19-2026 > we have a security agent that processes over 7 billion tokens per day on Kimi. That would cost an estimated $2.4M per year on a mid-tier proprietary model. But on Workers AI, it’s 77% cheaper. - Who: Cloudflare (the organization sells AI tools) - Source: The AI engineering stack we built internally, on the platform we ship, Cloudflare blog, April 20, 2026, at Act 1, "How we increasingly leverage Workers AI". https://blog.cloudflare.com/internal-ai-engineering-stack/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How does AI help keep a product secure? - Context (the archive's note, not their words): Cloudflare sells Workers AI, the product it describes running this agent on. - Passage page: https://the-ai-record.com/passages/cloudflare-security-agent-processes-over-7-billion-tokens-2026 > After 6 months in production: 70% reduction in trivial review comments (style, formatting) 40% faster PR turnaround time Zero security issues missed in production (caught during review) $0.12 average cost per PR review (using cost-optimized model routing) - Who: Juspay (the organization sells AI tools) - Source: How We Built an AI Code Reviewer with NeuroLink and Bitbucket, dev.to (NeuroLink AI organization account, Juspay), April 6, 2026, at Results & Lessons Learned. https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact?; What does an AI code review cost? - Context (the archive's note, not their words): List markers dropped; the four bullets under "After 6 months in production:" are joined with spaces. - Passage page: https://the-ai-record.com/passages/juspay-70-reduction-trivial-review-comments-40-faster-turnaround-2026 > Multi-model strategy works: Using cheaper models for simple checks and expensive ones for complex reasoning cut costs by 80% - Who: Juspay (the organization sells AI tools) - Source: How We Built an AI Code Reviewer with NeuroLink and Bitbucket, dev.to (NeuroLink AI organization account, Juspay), April 6, 2026, at Results & Lessons Learned > Key Lessons. https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; What does an AI code review cost?; How do organizations keep AI spending in check? - Passage page: https://the-ai-record.com/passages/juspay-multi-model-strategy-cheaper-models-simple-checks-cut-costs-2026 > And we have this notion of spend tiers. So by default, you get certain … tier, and then we have many levels of tiers. Once you're close to that tier, then you can raise a request to go to the next tier. And you just need a simple manager approval. … But we don't want to restrict engineers … they should be able to spend whatever they need because ultimately we want to … maximize the gain from this. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 26:54. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Context (the archive's note, not their words): Answering the host's question whether an engineer at Uber needs to be aware of AI spend. - Passage page: https://the-ai-record.com/passages/uber-spend-tiers-manager-approval-roi-conversation-2026 > To give you some perspective, the lead qualification agent that I opened with costs us under $5,000 a year between infrastructure and tokens to run. It takes about 20% of one engineer to maintain it. That's a 32x ROI. … It's now $5,000 and it operates 24/7 with faster speed to lead and all the quality of human equivalent metrics. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 14:05. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Vercel's COO, making the case that agents change the build-versus-buy decision, gives the running cost of the lead qualification agent she described at the start. - Passage page: https://the-ai-record.com/passages/vercel-lead-agent-under-5000-year-32x-roi-2026 > supported by AI-enabled productivity gains and continued cost discipline. - Who: Klarna - Source: SEC filing (the company's own regulatory report), furnished on Form 6-K, August 18, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000162828026057573/q22026earningsrelease.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-supported-ai-enabled-productivity-gains-continued-cost-2026-2 > our customer support related cost per booking declined approximately 16% year-over-year - Who: Airbnb - Source: SEC filing (8-K exhibit: the company's own shareholder letter), August 6, 2026. https://www.sec.gov/Archives/edgar/data/0001559720/000119312526337928/d70413dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI customer service cut the cost of support? - Passage page: https://the-ai-record.com/passages/airbnb-customer-support-related-cost-per-booking-declined-2026 > it's not about how many tokens you use, it's about the throughput of your product and the quality of your product designs and how much you're shipping. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-its-not-about-how-many-tokens-you-2026 > it obviously does assume a material increase in terms of the AI spend over the course of the year - Who: Ellie Mertz, Chief Financial Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-obviously-does-assume-material-increase-terms-ai-2026 > Customer support is a clear area where we should be able to up the quality of our customer support interactions as well as reduce the cost of our effort there. - Who: Balaji Krishnamurthy, Chief Financial Officer, Uber - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/UBER-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How much customer support does AI resolve? - Passage page: https://the-ai-record.com/passages/uber-customer-support-clear-area-where-should-able-2026 > This is supported by AI-enabled productivity gains and continued cost discipline across the business. - Who: Klarna - Source: SEC filing (the company's own regulatory report), furnished on Form 6-K, May 14, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000162828026034877/exhibitno992q126earnings.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/klarna-supported-ai-enabled-productivity-gains-continued-cost-2026 > Between the year ended December 31, 2022 and the year ended March 31, 2025, we reduced our support costs per Active Member by 60% and fraud loss rates by 29%. These improvements did not come at the expense of the member experience, either, as we doubled our member support satisfaction scores from the first quarter of 2022 to the first quarter of 2025. - Who: Chime - Source: Chime Financial, Inc. prospectus (Form 424B4), SEC EDGAR, June 12, 2025, at Business, Member Support. https://www.sec.gov/Archives/edgar/data/1795586/000162828025030855/chimefinancialinc-finalpro.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Are customers happy with AI customer service?; Does AI customer service cut the cost of support? - Passage page: https://the-ai-record.com/passages/chime-reduced-support-costs-per-active-member-60-2025 > One trick that we do, which I think is really helpful, is while you're still within the warm cache period, you should probably do the compaction right now, because you have this long conversation … going to be much cheaper. So just go ahead, do the compaction, and now when the user comes back at 30 minutes or whatever, they're starting from this summary, and the summary now is going to be much cheaper for them to continue adding messages on. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 32:20. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Context (the archive's note, not their words): Robinson is describing how his team keeps the cost of long agent conversations down by summarizing the context before the model provider's prompt cache expires. - Passage page: https://the-ai-record.com/passages/spacex-grok-bot-compact-while-cache-warm-2026 > we saw that against like a very naive implementation, our costs were down by 60%, and our quality and our accuracy was up by around 70%. - Who: Ameya Ketkar, Speaker, AI Engineer World's Fair 2026 (job title not stated), Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 10:50. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact?; What does an AI code review cost?; How do organizations test whether their AI works? - Context (the archive's note, not their words): Ameya Ketkar is summarizing the effect of the observability and evaluation work on uReview, Uber's code-review system. - Passage page: https://the-ai-record.com/passages/uber-we-saw-against-like-very-naive-2026 > The planning agent is responsible for the actual thinking, so it should run on your most powerful model. I initially used Fable for this purpose when it was available, as it excels in UI/UX design. When Fable was no longer accessible, I switched to Opus. Since the planner writes the specifications, the implementation can be handled by a more cost-effective and faster model like Sonnet. - Who: Noah Blauensteiner, Author, Xebia articles, Xebia (the organization sells AI tools) - Source: Shining Some Light in Dark Agentic Software Factories, Xebia articles, August 12, 2026, at What is an Agentic Software Factory?. https://xebia.com/articles/shining-some-light-in-dark-agentic-software-factories/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/xebia-planning-agent-runs-your-most-powerful-2026 > You could figure out which ones took less tool calls, … which one made less errors, which one produced the correct outcome, and deterministically know what's going on. The point is you can start building data-driven systems without ever writing a single line of code. … in order to build a programming language, it wouldn't have taken eight people. It wouldn't have taken less than two years. It would have taken hundreds and thousands and tens of thousands of man hours, and then you would still have a broken system. And today, we can just spend billions of tokens and make it work, and we can make it stable. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 6:00. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Does AI make developers more productive? - Context (the archive's note, not their words): Gupta is comparing what building a programming language would have taken before with what it takes the team now. - Passage page: https://the-ai-record.com/passages/boundary-ab-test-tool-calls-billions-of-tokens-programming-language-2026 > Notion's auto model does this really well. We have state-of-the-art models available always. But we also have an auto model there at the top that handles about 75% of our traffic, right? We have the ability to switch between models and our product, and we also offer it to our customers … they have access to these models without vendor lock-in. - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 13:15. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design?; How do organizations keep AI spending in check?; How do organizations choose and switch AI models? - Context (the archive's note, not their words): Sachs is describing how Notion avoids depending on a single AI model provider. - Passage page: https://the-ai-record.com/passages/notion-auto-model-handles-about-75-traffic-2026 > And the last thing is CPUs over GPUs. We've recently launched something at Notion called Workers. I don't think that the GPU is necessary for every job. A lot of the jobs that we have are actually serving discrete pieces of code. Like you don't need an LLM to turn a CSV into a PDF. You don't need an LLM to talk to Notion tool calls if we have a CLI. You definitely don't need an LLM to do deterministic SQL queries. This is where people become token poor very quick. - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 16:57. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run?; How do organizations keep AI spending in check? - Context (the archive's note, not their words): Sachs is closing a list of ways to lower AI costs. - Passage page: https://the-ai-record.com/passages/notion-launched-workers-cpus-over-gpus-token-poor-2026 > And not all traffic is equal. It is a huge miss to send all of these to the latest Opus model. Some of these absolutely. Large scale data analysis, when you do it on Notion, will recommend Opus, right? When you triage an email inbox, if we're charging you to do that on Opus, we're ripping you off, and ourselves. - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 10:24. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design?; How do organizations keep AI spending in check?; How do organizations choose and switch AI models? - Context (the archive's note, not their words): Sachs is arguing that an AI product should route different kinds of requests to different models. - Passage page: https://the-ai-record.com/passages/notion-not-all-traffic-equal-triage-email-opus-2026 > We posted this review when we announced our partnership with parallel as our web search provider. If you were to look at just latency of a single call or just cost, parallel might not be the cheapest. But if you have expertise in entire web search trajectories, you'll see how it differs. The granularity of this eval is what lets us make the best decisions for our customers, because we understand all of the trade-offs on entire trajectories - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 14:07. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Sachs is explaining how Notion compares model and tool providers, using its choice of a web search provider as the example. - Passage page: https://the-ai-record.com/passages/notion-posted-review-parallel-web-search-trajectories-2026 > … the issue is that these are activity metrics … These are not necessarily business outcomes. … when we start talking about the costs of this technology, … our CFO has asked me, what is the impact of this … I can't point him to diffs. I need to show him what's the impact on revenue. I'm sure you folks are dealing with the same problem. This is not necessarily a solved problem for us. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 35:07. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Chadha is describing the gap between the activity metrics Uber tracks for AI use and the business outcomes its CFO asks about. - Passage page: https://the-ai-record.com/passages/uber-activity-metrics-not-business-outcomes-cfo-asked-impact-revenue-2026 > The AI agent helped contribute to a 40% reduction in the HR team’s operational costs over the past four years. - Who: IBM (the organization sells AI tools) - Source: IBM (own site), 2025. https://www.ibm.com/case-studies/ibm-askhr - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/ibm-ai-agent-helped-contribute-40-reduction-hr-2025 > I’ve given him unlimited tokens because he’s going to refactor code… So I’m like, you can go, you can go nuts. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Cisco chief found a 27-year-old ex-founder ‘buried away in the organization doing something unstrategic’—now he’s upgrading 98% of its code, Fortune (Orianna Rosa Royle), October 30, 2026. https://fortune.com/2026/10/30/cisco-president-jeetu-patel-gen-z-27-year-old-unlimited-ai-tokens-rewrite-legacy-code-entry-level-jobs/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: Can AI do large code migrations?; How do leaders push AI adoption?; How do organizations keep AI spending in check? - Passage page: https://the-ai-record.com/passages/cisco-given-him-unlimited-tokens-because-he-s-going-2026 > the reason … we can build such amazing software for new businesses is because we also have a lot of businesses that are grown to be billion-dollar businesses … we can amortize … all of our … engineers and … tokens over, you know, a large pool of people - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify CEO on why using AI isn’t optional, Sources with Alex Heath (YouTube), October 28, 2026, at 62:46. https://www.youtube.com/watch?v=r9SHzCW9zxo - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-amortize-all-our-engineers-and-tokens-large-pool-2026 ## How do organizations test whether their AI works? Evaluation sets, benchmarks and reviews of agent transcripts that organizations use to judge their own AI systems. Topic page: https://the-ai-record.com/topics/ai-evals > Any incidents are now an eval. The team collected past PRs that caused incidents, with their fix-forwards and reverts, and had engineers who know that code confirm each bug and where it lived. The reviewer runs against that set daily. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at How we make the reviewer better. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/anthropic-past-incidents-become-daily-eval-for-reviewer-2026 > What we did was we built a system that actually has agents constantly running and creating BAML programs. … We then look at the entire … transcripts, see what tools they use, see what happened, and obviously we as humans can inspect them. But more importantly, we can have agents go inspect them. And agents find what was good, what was bad. … And we can have humans collaborate with these issues to figure out which ones are real, which ones are hallucinations, which ones don't have taste as much as I hate to use that word. And then we can have agents go ahead and create fixes to these problems and go to address them. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 5:05. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; Can AI fix bugs and handle incidents?; How is AI used for testing? - Context (the archive's note, not their words): Gupta is describing how the team tests its programming language when it does not rely on reading every line of code. - Passage page: https://the-ai-record.com/passages/boundary-agents-run-baml-programs-inspect-transcripts-humans-triage-2026 > And that's why we have Cursor Bench. We have this private eval set that is mostly made up of things that happen in our code base, which is held out from the … So we ensure that the models aren't trained on it. And it's based on those real world engineering tasks. - Who: Lee Robinson, Machine Learning Engineer, Model Behavior, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Recursive Model Improvement, AI Engineer (YouTube), July 15, 2026, at 8:12. https://www.youtube.com/watch?v=q4Tr-DknG2M - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How do organizations choose and switch AI models? - Context (the archive's note, not their words): Robinson has just said public benchmarks are an imperfect test of how models are used in practice, and is describing Cursor's private evaluation set for its own models. - Passage page: https://the-ai-record.com/passages/cursor-bench-private-eval-own-codebase-2026 > Regression test. Those live in ci, those have to pass a certain percent, you know, within some stochastic error rate. Then we have, as you’re building a product, evals of these aren’t passing right now, and this is launch quality. So we have a report card and we need to, on these categories, you know, be it 80 or 90% of all of these user journeys to launch, and then what we have what we call frontier or headroom evals, where we actively wanna be at 30% pass rate. - Who: Sarah Sachs, Manages Notion's core AI capabilities and infrastructure team, Notion (the organization sells AI tools) - Source: Notion's Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future (Latent Space), Latent Space, April 15, 2026, at 00:24:23. https://www.latent.space/p/notion - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/notion-regression-launch-quality-frontier-evals-30-2026 > what can I actually compartmentalize in a way that we can understand and evaluate? Because that's what I'm hill climbing against, right? Otherwise, I'm just tuning these things. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 11:47. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Long is describing how to break an investing process into parts that can each be tested. - Passage page: https://the-ai-record.com/passages/coatue-compartmentalize-evaluate-what-im-hill-climbing-against-2026 > Fine-tuning Gemma4-E2B on hundreds of thousands of domain-specific examples lifts its F1 from 0.563 to 0.902, on par with the best proprietary models, while median latency stays at 35ms. - Who: Airbnb - Source: Frontier quality at 1% of the cost: a production benchmark of LLM classifiers, Airbnb Tech blog, November 3, 2026, at Quality vs. latency. https://airbnb.tech/ai-ml/frontier-quality-at-1-of-the-cost-a-production-benchmark-of-llm-classifiers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Do organizations build their own AI tools or buy them?; How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/airbnb-fine-tuning-gemma4-e2b-lifts-f1-2026 > We built a benchmark of 200 credit analyst-reviewed income statements and balance sheets, with the analyst's figures recorded for every metric and time period. - Who: Ramp - Source: Agentic Credit Analysis, Ramp Builders blog, October 22, 2026, at Evaluation dataset. https://builders.ramp.com/post/credit-analysis-using-semantics - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-built-benchmark-200-credit-analyst-2026 > In a later randomized split, Opus 5.5 at medium showed no measurable change in review outcomes or critical findings compared with Opus 5, so we switched. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026, at Changing models and effort. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations choose and switch AI models? - Passage page: https://the-ai-record.com/passages/cloudflare-later-randomized-split-opus-55-medium-2026 > Before switching all reviews over from v1 to v2, we ran every review in every canary repo on both v1 and v2 in a comparative shadow mode. The two agreed on the approval decision in 98% of the merge requests we tested, so we moved everyone over in one go in early October and removed the v1 code. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026, at What else has changed since our last post?. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/cloudflare-switching-all-reviews-v1-v2-ran-2026 > once you thought you were going to save time by writing actually your context instead of all your code, but if you take this rigorously, you're going to spend time … writing the right evals. And that's kind of like a lot of work … because now you don't only have one prompt that you're trying to get right, it's like all the prompts of the evals. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Context Is the New Code, AI Engineer (YouTube), May 3, 2026, at 26:07. https://www.youtube.com/watch?v=bSG9wUYaHWU - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; Does AI make developers more productive? - Context (the archive's note, not their words): Debois answers an audience question after the talk, about the effort that testing context takes. - Passage page: https://the-ai-record.com/passages/tessl-save-time-writing-context-spend-time-writing-evals-2026 > Think about this, like, you run it five times, and out of five, how many times does it succeed? And … maybe in several cases it hits 100% all the time, which is great, but in others not. And depending on how you change your context, it will influence which tests actually work or not. I find it personally helpful to think about this as error budgets, I give a set of tests … that I really care about, so it's only allowed … to fail minimally, and other pieces are okay. So that's how you have to think about testing context, you cannot do, like, exact testing all the time. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Context Is the New Code, AI Engineer (YouTube), May 3, 2026, at 13:04. https://www.youtube.com/watch?v=bSG9wUYaHWU - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; How do organizations measure AI's impact? - Context (the archive's note, not their words): Debois explains how he tests the instructions and skills given to a coding agent, because the same test run twice can give different results. - Passage page: https://the-ai-record.com/passages/tessl-testing-context-run-five-times-error-budgets-2026 > we can't look at any traces basically. We can't look at any data because it's all super confidential. And that's, like, extremely frustrating to work with because you you basically don't know what people are doing. - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 25:30. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How do banks and other regulated organizations use AI within the rules? - Passage page: https://the-ai-record.com/passages/legora-traces-no-access-customer-data-evals-2026 > So we have different, basically, like, worlds, like RL environments, essentially, which is like, here's a… Because because lawyers work in projects always, more or less, so it's like, here's 10,000 file data room, you're doing your due diligence - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 9:00. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/legora-worlds-rl-environments-10000-file-data-room-2026 > we’ll all sit down every Friday and look at the P 99 of like the most token exhaustive custom agent transcript and just look at why it didn’t do well and cut a bunch of tasks. - Who: Sarah Sachs, Manages Notion's core AI capabilities and infrastructure team, Notion (the organization sells AI tools) - Source: Notion's Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future (Latent Space), Latent Space, April 15, 2026, at 00:09:31. https://www.latent.space/p/notion - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/notion-friday-review-p99-token-heavy-agent-2026 > Notions last exam is a big thing inside the company and we have people, full-time staff to it exclusively. Mm. We have a data scientist, a model behavior engineer, and an full-time, um, evals engineer just dedicated to the evals that we pass 30% of the time. - Who: Sarah Sachs, Manages Notion's core AI capabilities and infrastructure team, Notion (the organization sells AI tools) - Source: Notion's Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future (Latent Space), Latent Space, April 15, 2026, at 00:24:23. https://www.latent.space/p/notion - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/notion-notions-last-exam-full-time-staff-2026 > We had a GTM engineer, as I mentioned, shadow one of our best SDRs for a few days. … Watching every single tab she opened, … The engineer converted each of those data sources into a step in a tool calling workflow. … After building our first version, … ran in production in shadow mode for six weeks. Our best SDR reviewed every single output to create a feedback flywheel. And then finally, she couldn't improve it anymore. The architecture reflects exactly what she was doing manually. But now, it performs like the 90th percentile rep, 100% of the time. We applied the same framework to 30 different SDR workflows, … SDR quotas increased by 30% this quarter. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 15:00. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used in sales and marketing? - Context (the archive's note, not their words): Vercel's COO describes how her team built the lead qualification agent, step by step, and what followed. - Passage page: https://the-ai-record.com/passages/vercel-shadow-best-sdr-six-weeks-90th-percentile-2026 > Early on we realized that none of the existing tracing solutions are 100% matching what we want to see and code is cheap, so we’ve build a custom admin panel that evolved in a full solution. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/ramp-none-existing-tracing-solutions-100-matching-2026 ## What does an AI code review cost? The price of one AI review of a pull request, what drives it up, and what organizations did to bring it down. Topic page: https://the-ai-record.com/topics/ai-review-cost > With code quality at high effort, the average review now costs just $0.77, down 68% from the $2.43 it averaged over our two most expensive weeks before we started. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations keep AI spending in check? - Passage page: https://the-ai-record.com/passages/cloudflare-code-quality-high-effort-average-review-2026 > In its first week in production the roster brought the median cost per review from about thirty cents to eleven while publishing at roughly the same rate. - Who: Dana Dzik, Engineer, Coval (author of the Coval blog post on its code reviewer), Coval (the organization sells AI tools) - Source: Notes on Building a Code Reviewer, Coval blog, September 10, 2026, at The roster. https://www.coval.ai/blog/how-we-built-our-code-reviewer/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/coval-roster-brought-median-cost-per-review-2026 > Reviews are billed on token usage and generally average $15–25, scaling with PR size and complexity. - Who: Anthropic (the organization sells AI tools) - Source: Bringing Code Review to Claude Code, claude.com (Anthropic), March 9, 2026, at Cost and control. https://claude.com/resources/articles/code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/anthropic-code-review-billed-token-usage-15-25-per-review-2026 > The configuration that catches the most of what our human reviewers catch is also the cheapest one we have run, and most of its budget goes to models that would be dismissed on sight as too weak for code review—which, individually, they are. - Who: Dana Dzik, Engineer, Coval (author of the Coval blog post on its code reviewer), Coval (the organization sells AI tools) - Source: Notes on Building a Code Reviewer, Coval blog, September 10, 2026, at Where it goes. https://www.coval.ai/blog/how-we-built-our-code-reviewer/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/coval-configuration-catches-most-what-our-2026 > Medium effort did lose real findings. In the randomized split, code quality at medium led to roughly a quarter as many acted-on findings as at high, which is why we're moving it to high. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026, at Review quality. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/cloudflare-medium-effort-did-lose-real-findings-2026 > Review has to converge. An agent author and an agent reviewer can keep each other busy indefinitely. One CLI PR went through more than fifty rounds of review and fixes overnight before its author stopped it. When we ran git blame on what the reviewer kept flagging, about 59% of it was code that had been there since the reviewer's first pass. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at How we make the reviewer better. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/anthropic-one-pr-fifty-review-rounds-overnight-2026 > Before and after the convergence work: pull requests passing after one review and fix round rose from 65% to 78%; pull requests needing five or more rounds fell from 8.9% to 1.5%; mean reviews to pass fell from 2.08 to 1.32. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at How we make the reviewer better: Figure D, Review that converges (the chart's text description). https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Context (the archive's note, not their words): The text description of a chart in the post comparing internal pull requests before and after changes to re-review, which the post lists as a cap of three rounds, re-reviewing only the new diff after the first few passes, flagging the class of bug, and not repeating findings already posted. - Passage page: https://the-ai-record.com/passages/anthropic-prs-passing-after-one-review-round-65-to-78-2026 > Noise costs more when the author is an agent. A person can wave off a nit in seconds. An agent fixes it, which means another commit and another round of review. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at How we make the reviewer better. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI code review catch real problems? - Passage page: https://the-ai-record.com/passages/anthropic-review-noise-costs-more-when-author-agent-2026 ## Does AI code review catch real problems? How often AI review comments lead to a change, what kinds of problems they find, and where reviewers say the comments are noise. Topic page: https://the-ai-record.com/topics/ai-review-quality > Before the changes, authors acted on 34% of the code quality reviewer's serious findings. Now it's 59%. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026, at Review quality. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cloudflare-changes-authors-acted-34-code-quality-2026 > We measure noise by watching what people do with the comments. It's easy to catch more bugs if you don't mind flagging a lot of non-bugs, so we need a check on that. We sort every comment the reviewer leaves by how the person on the PR responded: they fixed it, said it was wrong, said the behavior was intentional, or decided it didn't matter. We also run the reviewer on PRs we know are clean and count how often it reports something that would have led someone to make a harmful change. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at How we make the reviewer better. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/anthropic-measure-reviewer-noise-by-author-responses-2026 > Findings get verified by other agents before humans review them. Several agents look for bugs at the same time. Then a separate group votes on each finding, and it only gets posted if enough of them agree it's real. Findings that are only about style, diagnostics, or cleanup get dropped. We A/B tested those filters on our own PRs before keeping them. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Layer by layer: 2. AI review on every pull request. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/anthropic-separate-agents-vote-on-each-review-finding-2026 > Reviews are capped at three findings, a partly arbitrary number kept because it measurably changes whether people read the output; and silence is the usual outcome, 922 of the 1,732 recorded runs of this design having published nothing, a mean of 0.70 findings per review. - Who: Dana Dzik, Engineer, Coval (author of the Coval blog post on its code reviewer), Coval (the organization sells AI tools) - Source: Notes on Building a Code Reviewer, Coval blog, September 10, 2026, at The rebuild. https://www.coval.ai/blog/how-we-built-our-code-reviewer/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/coval-reviews-capped-three-findings-silence-2026 > [uReview] does … around 25,000 comments a week, and we get 10% of them actually get some feedback … We also saw that the overall address rate was around 67%, and almost three quarters of the high severity issues were usually addressed by the developers - Who: Ameya Ketkar, Speaker, AI Engineer World's Fair 2026 (job title not stated), Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 10:13. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact? - Context (the archive's note, not their words): Ameya Ketkar is reporting results for uReview, Uber's code-review system. The address rate is how often a developer acts on a comment, as he defined it earlier in the talk. - Passage page: https://the-ai-record.com/passages/uber-ureview-does-around-25-000-comments-2026 > We already had Ona's code review automation running on every PR. The agent reviewed diffs, ran the code in a full environment, cross-referenced tickets, and left inline comments with an 88% acceptance rate and a 50% reduction in time spent on pull requests. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at The experiment. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/ona-already-code-review-automation-running-every-pr-2026 > We've been running Code Review internally for months: on large PRs (over 1,000 lines changed), 84% get findings, averaging 7.5 issues. On small PRs under 50 lines, that drops to 31%, averaging 0.5 issues. Engineers largely agree with what it surfaces: less than 1% of findings are marked incorrect. - Who: Anthropic (the organization sells AI tools) - Source: Bringing Code Review to Claude Code, claude.com (Anthropic), March 9, 2026, at Code Review in action. https://claude.com/resources/articles/code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/anthropic-large-prs-84-get-findings-under-1-percent-incorrect-2026 > we conclude that RovoDev Code Reviewer is effective in generating code review comments that could lead to code resolution for 38.70% (i.e., comments that triggered code changes in the subsequent commits) - Who: Atlassian (the organization sells AI tools) - Source: RovoDev Code Reviewer: A Large-Scale Online Evaluation of LLM-based Code Review Automation at Atlassian, arXiv (Atlassian researchers; accepted at ICSE 2026, SEIP track), January 5, 2026, at Abstract. https://arxiv.org/abs/2601.01129 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/atlassian-rovodev-code-reviewer-effective-generating-comments-resolution-2026 > developers are most likely to resolve comments about readability, bugs, and maintainability. These comment types are found to be more specific and actionable, i.e., code changes were made to resolve the comments, with resolution rates between 36% and 43%. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?, Inside Atlassian (Atlassian engineering blog), November 10, 2025, at Which Comments Actually Get Resolved?. https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/atlassian-developers-most-likely-resolve-comments-readability-2025 > uReview today analyzes over 90% of the weekly ~65,000 diffs (equivalent of pull requests) landed at Uber. Engineers who interact with the tool mark 75% of its comments as useful - Who: Sonal Mahajan, Staff Engineer, Programming Systems team, Uber - Source: uReview: Scalable, Trustworthy GenAI for Code Review at Uber, Uber engineering blog, August 12, 2025. https://www.uber.com/us/en/blog/ureview/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/uber-ureview-today-analyzes-over-90-weekly-65-2025-2 > Important findings per review went from 0.28 before to 0.30 now, and the intervals overlap. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Cutting our inference bill by 68% by being obsessive about optimization, Cloudflare blog, October 19, 2026, at Review quality. https://blog.cloudflare.com/obsessive-inference-optimization/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/cloudflare-important-findings-per-review-went-028-2026 > When we launched the research preview, the share of internal PRs getting substantive review comments went from 16% to 54%. Most PRs over 1,000 lines get findings, and most small ones come back clean. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Layer by layer: 2. AI review on every pull request. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): About Code Review for Claude Code, Anthropic's AI pull-request reviewer, which the post says is now generally available on its Team and Enterprise plans. - Passage page: https://the-ai-record.com/passages/anthropic-substantive-review-comments-16-to-54-percent-2026 > A review bot that posts on every pull request can waste more attention than it saves. Humans have to read the comments, decide which are right, and resolve threads. A reviewer who has learned that the bot is often wrong learns to skim, and then skims the good findings too. - Who: TradersPost - Source: The AI Review Bot We Reverted, TradersPost blog, October 13, 2026, at A noisy bot costs attention. https://blog.traderspost.io/article/the-ai-review-bot-we-built-and-reverted - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/traderspost-review-bot-posts-every-pull-request-waste-2026 > Asked to name sensitive risk domains in prose, the model claimed authorization in 38% of reviews, while about 18% of the changes touched anything authorization-shaped; required to quote the specific changed line instead, it produced 570 claims, of which 98% resolved to real code—the same model over the same pull requests, with only the form of the question changed. - Who: Dana Dzik, Engineer, Coval (author of the Coval blog post on its code reviewer), Coval (the organization sells AI tools) - Source: Notes on Building a Code Reviewer, Coval blog, September 10, 2026, at The rebuild. https://www.coval.ai/blog/how-we-built-our-code-reviewer/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/coval-asked-name-sensitive-risk-domains-2026 > One of the biggest learnings in this process was like the model doesn't know that it's wrong. It always confidently says 100% sure that, yeah, this is the review for your code. Go ahead. But we saw that, no, it actually needs a lot of guidance from the teams, because each team has its own style guide, its own patterns or … anti-patterns that they want to look for. - Who: Ameya Ketkar, Speaker, AI Engineer World's Fair 2026 (job title not stated), Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 06:35. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): Ameya Ketkar is describing what Uber's team learned while adding observability and evaluation to its uReview code-review agent. - Passage page: https://the-ai-record.com/passages/uber-one-biggest-learnings-process-like-model-2026 > engineers don't trust AI opinions. They just trust clear findings tied to standards that we … agreed upon. … it stopped sounding more like, maybe this is a problem and started sounding more like this violates a standard that we actually care about. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 26:40. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Context (the archive's note, not their words): Describing early misses with Cloudflare's internal AI code reviewer, before the team tied its findings to machine-readable engineering standards. - Passage page: https://the-ai-record.com/passages/cloudflare-engineers-dont-trust-ai-opinions-clear-findings-2026 > Memory matters: Per-PR conversation context dramatically improved review quality over stateless approaches - Who: Juspay (the organization sells AI tools) - Source: How We Built an AI Code Reviewer with NeuroLink and Bitbucket, dev.to (NeuroLink AI organization account, Juspay), April 6, 2026, at Results & Lessons Learned > Key Lessons. https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/juspay-memory-matters-per-pr-conversation-context-review-quality-2026 > In addition, Rovo Dev is focusing only on actionable comments with only 0.6% classified as “No Issue”. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?, Inside Atlassian (Atlassian engineering blog), November 10, 2025, at Results. https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/atlassian-rovo-dev-focusing-actionable-comments-only-2025 > Atlassian’s internal projects, Rovo Dev generated a higher proportion of bug (2.8x) and maintainability (1.4x) comments, while human reviewers focused more on readability and slightly more design. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?, Inside Atlassian (Atlassian engineering blog), November 10, 2025, at Results. https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/atlassian-rovo-dev-generated-higher-proportion-bug-2025 > so when Devin fixes his own comments, you might be scared that, oh, maybe I’ll infinite loop. But we’ve put a lot of work into making sure it doesn’t, both by making sure that the comments are high signal, but also that the agent is thoughtful about what comments it immediately goes and tries to fix, and what comments it’s like, “Wait a second, I think you’re wrong.” - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:24:37. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/cognition-devin-fixes-own-comments-infinite-loop-high-signal-2026 > We absolutely had to get the bots to help us review PR's. … But now, with the repo readiness work that the champions had done, coupled with better models and tools, … we were getting much better results here. So we enabled … on all the repos, and we also created an auto fix loop where if … issues, another agent will automatically fix those issues and commit them to the PR. So now, at least people were no longer complaining about not wanting to review sloppy PR's from bots, right? By time they got to them, they were in pretty good shape. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 13:31. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Context (the archive's note, not their words): Jones is describing how Block handled code review once engineers were running several agents at once. - Passage page: https://the-ai-record.com/passages/block-bots-review-prs-auto-fix-loop-2026 > the best example of a rule that I can think of internally is for Bugbot. So when we're doing database migrations, we're not really using foreign keys on a database for performance reasons. And the models like the right way to do this is use foreign keys, right? So they will always add a foreign key. But when it … hits … and there's a PR created, we have Bugbot … it's like, oh, I have this rule saying like we should never use foreign keys. So then it flags this. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 45:06. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): Zakariasson is giving an example of a rule Cursor keeps for its Bugbot code-review tool, in an answer about writing rules for agents. - Passage page: https://the-ai-record.com/passages/cursor-bugbot-rule-never-use-foreign-keys-2026 > Off-the-shelf AI code review tools: Great for generic checks, but couldn't understand our Haskell payment systems or custom conventions - Who: Juspay (the organization sells AI tools) - Source: How We Built an AI Code Reviewer with NeuroLink and Bitbucket, dev.to (NeuroLink AI organization account, Juspay), April 6, 2026, at The Architecture Decision. https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/juspay-off-the-shelf-ai-code-review-tools-generic-checks-haskell-2026 > Both utilize a semantic search pipeline based on retrieval-augmented generation to assemble relevant contextual information for the review, thereby tackling the uncovered challenges. Developers evaluated both variations in real-world settings: AI-led reviews are overall more preferred, while still being conditional on the reviewers’ familiarity with the code base, as well as on the severity of the pull request. - Who: Fannar Steinn Aðalsteinsson, Author, affiliated with WirelessCar Sweden AB and Chalmers University of Technology, WirelessCar - Source: Rethinking Code Review Workflows with LLM Assistance: An Empirical Study, arXiv, May 22, 2025, at Abstract. https://arxiv.org/abs/2505.16339 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/wirelesscar-field-experiment-two-llm-assisted-code-review-2025 ## How much customer support does AI resolve? Shares and volumes of customer conversations AI handles, and what each organization means by resolved. Topic page: https://the-ai-record.com/topics/ai-support-resolution > Nearly 45% of issues that start with our AI assistant are now resolved without a human agent while delivering much faster resolution times. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/airbnb-nearly-45-issues-start-ai-assistant-now-2026 > Customer agent now resolves 72% of support tickets without human escalation. - Who: HubSpot (the organization sells AI tools) - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 2026. https://transcripts.platformaeronaut.com/transcripts/HUBS-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/hubspot-customer-agent-now-resolves-72-support-tickets-2026 > 7.1 million conversations handled by our digital assistant Cora in H1 2026, of which 3.8 million were fulfilled entirely digitally, up by 23% compared with H1 2025. - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), July 31, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/31072026/nwg-h1-2026-results.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-7-1-million-conversations-handled-digital-assistant-2026 > when guests contact us through our AI aid assistant, over 40% of issues are now resolved without a human agent. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 7, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/airbnb-when-guests-contact-us-through-ai-aid-2026 > Our AI assistant has handled 80% of customer service chats in the year ended December 31, 2025 (according to our service chat log data) - Who: Klarna - Source: SEC filing (the company's own regulatory report), furnished on Form 6-K, February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-ai-assistant-handled-80-customer-service-chats-2026 > Since launch, our AI assistant has had 31 million conversations, handling 80% of customer service chats in the year ended December 31, 2025 - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-since-launch-ai-assistant-had-31-million-2026 > nearly 30% of tickets in North America that are English-based are handled by an AI agent - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), February 12, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-4Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/airbnb-nearly-30-tickets-north-america-english-based-2026 > our AI assistant has had 25.3 million conversations, solving 69% of customer service chats in the last twelve months ended June 30, 2025 - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-ai-assistant-had-25-3-million-conversations-2025 > Our AI assistant handled 69% of customer service chats in the last twelve months ended June 30, 2025, according to our service chat log data - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-ai-assistant-handled-69-customer-service-chats-2025 > Our ability to scale efficiently and improve our members’ support experience is supported by our investments in AI and automation, which allow these tools to do the work of over 1,000 human member support agents. These innovations enabled us to deliver 68% of our member support interactions without the need for human intervention in the first quarter of 2025. - Who: Chime - Source: Chime Financial, Inc. prospectus (Form 424B4), SEC EDGAR, June 12, 2025, at Business, Member Support. https://www.sec.gov/Archives/edgar/data/1795586/000162828025030855/chimefinancialinc-finalpro.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): The prospectus footnote for the 1,000-agent figure gives its date as December 31, 2024. - Passage page: https://the-ai-record.com/passages/chime-ai-automation-work-over-1-000-human-support-agents-2025 > In 2025, Cora handled 12.9m retail banking customer conversations, 50% of which needed no human intervention. - Who: NatWest Group - Source: NatWest Group website (the company's own sustainability page), October 29, 2026. https://www.natwestgroup.com/sustainability/society/innovation-and-digitisation.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-2025-cora-handled-12-9m-retail-banking-2026 > Fin has industry leading resolution rates, averaging 76% across 12,000+ customers, with many seeing over 85%. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: Intercom (own site), October 2026. https://www.intercom.com/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/intercom-fin-industry-leading-resolution-rates-averaging-76-2026 > Agentforce on Help has now surpassed five million customer conversations, resolving more than 63% of them instantly. - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-agentforce-help-now-surpassed-five-million-customer-2026 > In 2025, Cora handled 12.9 million conversations, with 50% of these interactions requiring no human intervention from our teams - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-2025-cora-handled-12-9-million-conversations-2026 > we have seen around a 20-percentage-point increase in queries resolved without any human intervention, compared to the equivalent non-Gen AI supported Cora journey. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/natwest-seen-around-20-percentage-point-increase-queries-2026 > Each quarter, we resolve over 60,000 support requests with AI agents that carry out full workflows, execute backend actions, and adapt in real time to what our customers need. - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-each-quarter-resolve-over-60-000-support-2025 > In 2024, 21 percent of requests were fully automated and resolved without any agent intervention - Who: Zendesk (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), June 30, 2025. https://www.zendesk.com/customer/zeal/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-2024-21-percent-requests-were-fully-automated-2025 > In 2024, Cora, our AI virtual assistant, handled 11.2 millionLA Retail Banking conversations, almost half of which required no human input. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2025. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2024/strategic-report-2024.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-2024-cora-ai-virtual-assistant-handled-11-2025 > Around half of Cora conversations are now entirely contained and we achieved a Cora CSAT of 66%. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2025. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2024/strategic-report-2024.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Are customers happy with AI customer service? - Passage page: https://the-ai-record.com/passages/natwest-around-half-cora-conversations-now-entirely-contained-2025 > In 2023, Cora, our AI virtual assistant, handled 10.8 million(*) Retail Banking conversations, almost half of which required no human input. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2024. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2023/strategic-report-2023.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-2023-cora-ai-virtual-assistant-handled-10-2024 > Cora, fully supported 49% of customer queries through 10.8 million(*) conversations handled without colleague intervention. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2024. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2023/strategic-report-2023.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-cora-fully-supported-49-customer-queries-through-2024 > So while we solve 50% of the tickets, we're actually deliberate about the tickets we don't choose to solve yet [with agents]. - Who: Ahmad Al-Dahle, Chief Technology Officer, Airbnb - Source: Latent Space (independent AI engineering publication), October 2, 2026. https://www.latent.space/p/airbnb - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-so-while-solve-50-tickets-were-actually-2026 > Our AI assistant handled 69% of customer service chats in the last twelve months ended June 30, 2025 - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-ai-assistant-handled-69-customer-service-chats-2025-2 > We estimate that our AI assistant does the equivalent work of over 700 full-time agents - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/klarna-estimate-ai-assistant-does-equivalent-work-over-2025 > Currently at 2 million weekly resolutions and growing fast - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: Fin website (Intercom's rebranded own site), October 29, 2026. https://fin.ai/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/intercom-currently-2-million-weekly-resolutions-growing-fast-2026 > A resolution is a type of outcome that is counted when, following Fin's last answer in a conversation, the customer either confirms the answer was satisfactory (confirmed resolution), or exits the conversation without requesting further assistance (assumed resolution). - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: Intercom help center (the company's own documentation), October 29, 2026. https://www.intercom.com/help/en/articles/8205718-fin-resolutions - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/intercom-resolution-type-outcome-counted-when-following-fins-2026 > We've been seeing around 560,000 resolutions a month with Fin, at about a 79% resolution rate, - Who: Isabel Larrow, Head of AI Support (title as shown on Intercom's homepage), Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: Intercom website (the vendor's own homepage, quoting a customer), October 29, 2026. https://www.intercom.com/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/intercom-weve-been-seeing-around-560-000-resolutions-2026 > Clients now interact with Erica nearly 200 million times per quarter. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), October 14, 2026. https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-clients-now-interact-erica-nearly-200-million-2026 > has surpassed 3.8 billion client interactions since its 2018 launch. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), October 14, 2026. https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-surpassed-3-8-billion-client-interactions-since-2026 > 94% Automated resolution (during tax season) 72% Verified resolution (during tax season) - Who: Zendesk (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), October 8, 2026. https://www.zendesk.com/customer/liberty-tax/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-94-automated-resolution-during-tax-season-72-2026 > we achieved a 4.6 CSAT rating and a 74% resolution rate by empowering AI agents and humans to work as one unified team - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site, webinar listing), October 2026. https://www.salesforce.com/events/webinars/how-salesforce-uses-agentforce-to-transform-support-handoffs/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Are customers happy with AI customer service? - Passage page: https://the-ai-record.com/passages/salesforce-achieved-4-6-csat-rating-74-resolution-2026 > answering more than 3 million support requests in the first year - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site, webinar listing), October 2026. https://www.salesforce.com/events/webinars/how-salesforce-uses-agentforce-to-transform-support-handoffs/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-answering-more-than-3-million-support-requests-2026 > automated resolutions for its highest-performing product increased from 39 to 46 percent - Who: Zendesk (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), September 17, 2026. https://www.zendesk.com/customer/bgl/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-automated-resolutions-its-highest-performing-product-increased-2026 > Stripe’s user-facing AI agent that holds 100k+ conversations per week - Who: Stripe - Source: Stripe job posting, hosted on the General Catalyst job board (General Catalyst is a Stripe investor), September 13, 2026. https://jobs.generalcatalyst.com/companies/stripe/jobs/93096758-product-manager-support-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-stripes-user-facing-ai-agent-holds-100k-2026 > Support workflow creation: 9,000 manual workflows → self-service automation. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), July 7, 2026. https://x.com/praveenTweets/status/2074605343439810922 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-support-workflow-creation-9-000-manual-workflows-2026 > We had an initial goal of about 30 to 35 percent automated resolutions when we first launched. We went up to 60 percent within the first two weeks—and then we had peaks of 85 percent. - Who: Luca Giaccardi, Senior Operations Excellence, E-Commerce, Qantas Loyalty - Source: Zendesk customer story (the vendor writing about its own customer), June 25, 2026. https://www.zendesk.com/customer/qantas-loyalty/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-had-initial-goal-about-30-35-percent-2026 > automates 91% of support tickets and saves 5,000 engineer-hours a month - Who: Vercel (the organization sells AI tools) - Source: Vercel Ship 2026 recap, Vercel blog, June 2026. https://vercel.com/blog/vercel-ship-2026-recap - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-automates-91-support-tickets-saves-5-000-2026 > now automates 91% of support tickets - Who: Vercel (the organization sells AI tools) - Source: Vercel Ship 2026 recap, Vercel blog, June 2026. https://vercel.com/blog/vercel-ship-2026-recap - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-now-automates-91-support-tickets-2026 > 67% Automated resolution rate 86% AI understood conversations 19% Escalations to a human agent - Who: Zendesk (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), April 6, 2026. https://www.zendesk.com/customer/digidentity/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-67-automated-resolution-rate-86-ai-understood-2026 > Zendesk expects AI to manage more customer service interactions than humans in 2026. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (the company's own press release), March 11, 2026. https://www.zendesk.com/newsroom/press-releases/zendesk-secures-key-industry-recognition-as-its-ai-first-strategy-gains-momentum/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-zendesk-expects-ai-manage-more-customer-service-2026 > Last year, 20.6 million users interacted with Erica nearly 700 million times. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), March 10, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/03/bofa-ai-and-digital-innovations-fuel-30-billion-client-interacti.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-last-year-20-6-million-users-interacted-2026 > Since Erica's launch in 2018, total client interactions have surpassed 3.2 billion. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), March 10, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/03/bofa-ai-and-digital-innovations-fuel-30-billion-client-interacti.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-since-ericas-launch-2018-total-client-interactions-2026 > We're seeing 71 percent automated resolution—far exceeding the other use cases that we have on average, which sit at around 50 percent. - Who: Emily Dowse, Head of Support, Arbor Education (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), December 9, 2025. https://www.zendesk.com/customer/arbor-education/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-were-seeing-71-percent-automated-resolution-far-2025 > 70% of inquiries have been resolved with a digital assistant - Who: IBM (the organization sells AI tools) - Source: IBM (own site), August 26, 2025. https://www.ibm.com/think/insights/enterprise-transformation-extreme-productivity-ai - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ibm-70-inquiries-been-resolved-digital-assistant-2025 > CashPro Chat is used by 65% of clients, with Erica handling greater than 40% of client interactions. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), August 20, 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-cashpro-chat-used-65-clients-erica-handling-2025 > More than 98% of users find the information they need, significantly decreasing call center volume - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), August 20, 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/bank-of-america-more-than-98-users-find-information-they-2025 > 2,000+ workflow-heavy service requests automated per quarter - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-2-000-workflow-heavy-service-requests-automated-2025 > With 7% of incoming tickets handled by AI agents, Jaycar has saved around AUD85,000. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk customer story (the vendor writing about its own customer), November 27, 2024. https://www.zendesk.com/customer/jaycar/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-7-incoming-tickets-handled-ai-agents-jaycar-2024 > To date, Erica has responded to 800 million inquiries from over 42 million clients - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), April 8, 2024. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2024/04/bofa-s-erica-surpasses-2-billion-interactions--helping-42-millio.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-date-erica-responded-800-million-inquiries-over-2024 > More than 2 billion since 2018. That's how many interactions Bank of America clients have had with Erica - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), April 8, 2024. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2024/04/bofa-s-erica-surpasses-2-billion-interactions--helping-42-millio.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-more-than-2-billion-since-2018-thats-2024 > More than 98% of clients get answers they need from Erica within 44 seconds, on average. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), April 8, 2024. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2024/04/bofa-s-erica-surpasses-2-billion-interactions--helping-42-millio.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/bank-of-america-more-than-98-clients-get-answers-they-2024 > Ultimate’s AI agents automate up to 80 percent of support requests. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (the company's own press release), March 13, 2024. https://www.zendesk.com/newsroom/press-releases/ultimate-acquisition24/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-ultimates-ai-agents-automate-up-80-percent-2024 > an autonomous support agent that Zendesk believes will solve 80% of support issues without human intervention. - Who: Russell Brandom, AI Editor, Zendesk (the organization sells AI tools) - Source: TechCrunch (news site), October 8, 2025. https://techcrunch.com/2025/10/08/zendesk-says-its-new-ai-agent-can-solve-80-of-support-issues - Grade C: An interview or press quote. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/zendesk-autonomous-support-agent-zendesk-believes-will-solve-2025 > we have now done about a million and a half conversations with customers, and at the same time, that’s the agentic layer speaking to the customer - Who: Marc Benioff, Chair, CEO & Co-Founder, Salesforce (the organization sells AI tools) - Source: Salesforce Ben (independent Salesforce news site), September 3, 2025. https://www.salesforceben.com/ai-agents-drive-4000-job-cuts-in-salesforce-support-division/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-now-done-about-million-half-conversations-customers-2025 > 50% are with agents, 50% are with humans. - Who: Marc Benioff, Chair, CEO & Co-Founder, Salesforce (the organization sells AI tools) - Source: Fortune (news), September 2, 2025. https://dc.fortune.com/2025/09/02/salesforce-ceo-billionaire-marc-benioff-ai-agents-jobs-layoffs-customer-service-sales - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-today-those-same-interactions-happening-50-agents-2025 > Our AI assistant is now available in more than 50 languages. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-ai-assistant-now-available-more-than-50-2026 > We also expanded AI-enabled capabilities across onboarding, operations and customer servicing, with five agents now live supporting our colleagues in core processes and our first customer-facing GenAI capability launched to help customers with Bankline queries. - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), July 31, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/31072026/nwg-h1-2026-results.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-also-expanded-ai-enabled-capabilities-across-onboarding-2026 > New AI capabilities have improved service, increased efficiency, and allowed teammates to focus on higher-value client interactions. - Who: Alastair Borthwick, Executive Vice President and Chief Financial Officer, Bank of America - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15267023092982784 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-new-ai-capabilities-improved-service-increased-efficiency-2026 > Agentforce can resolve customer issues swiftly and accurately with answers grounded in our trusted, unstructured content unified in Data 360 - Who: Salesforce (the organization sells AI tools) - Source: SEC filing (the company's own regulatory report), March 2, 2026. https://www.sec.gov/Archives/edgar/data/1108524/000110852426000060/crm-20260131.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/salesforce-agentforce-can-resolve-customer-issues-swiftly-accurately-2026 > At the same time, appreciating that certain consumers may nevertheless prefer to interact with human representatives, we continue to offer all of our customers that option. - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-same-time-appreciating-certain-consumers-may-nevertheless-2026 > For Employee Support, a proprietary GenAI conversational bot handles over 1.4 million annual inquiries, increasing the self-service rate from 70% to 93%. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business, AI-Driven Innovation at MercadoLibre. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/mercado-libre-proprietary-genai-conversational-bot-handles-over-2025 > New AI capabilities enabled quicker responses to customer complaints, saving around 90,000 hours per annum through automated summarisation and AI-generated complaint responses. - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), February 13, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/13022026/nwg-company-announcement.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make human support agents faster? - Passage page: https://the-ai-record.com/passages/natwest-new-ai-capabilities-enabled-quicker-responses-customer-2026 > Consistent with prior years, we relied on a global network of approximately 13,000 third-party workers as of December 31, 2025, to support the majority of our community support contacts. - Who: Airbnb - Source: SEC filing (the company's own regulatory report), February 12, 2026. https://www.sec.gov/Archives/edgar/data/1559720/000155972026000004/abnb-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-consistent-prior-years-relied-global-network-approximately-2026 > During 2025, we introduced new artificial intelligence (“AI”) features to help deliver customer support in certain countries and languages. - Who: Airbnb - Source: SEC filing (the company's own regulatory report), February 12, 2026. https://www.sec.gov/Archives/edgar/data/1559720/000155972026000004/abnb-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-during-2025-introduced-new-artificial-intelligence-ai-2026 > We built a custom agent built on 13 different models that have been tuned off of tens of thousands of conversations. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 2025. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-built-custom-agent-built-13-different-models-2025 > By leveraging GenAI and other ML tools, we deliver high-quality, efficient member support at scale, including our GenAI chatbot and voicebot that we launched in 2025. Our proprietary, AI-powered disputes management platform has contributed to improved agent productivity and reduced dispute investigation times, creating better experiences for our members. - Who: Chime - Source: Chime Financial, Inc. prospectus (Form 424B4), SEC EDGAR, June 12, 2025, at Business, Data and AI, Member Support. https://www.sec.gov/Archives/edgar/data/1795586/000162828025030855/chimefinancialinc-finalpro.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; Does AI make human support agents faster? - Passage page: https://the-ai-record.com/passages/chime-genai-chatbot-voicebot-launched-2025-disputes-platform-2025 > Launched Agent Assist, Citi’s first USPB Generative AI customer service tool - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended March 31, 2025, SEC EDGAR, May 8, 2025, at Management’s Discussion and Analysis, Transformation, Progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000086/c-20250331.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Context (the archive's note, not their words): USPB is U.S. Personal Banking, one of Citi’s five operating segments. - Passage page: https://the-ai-record.com/passages/citigroup-launched-agent-assist-first-uspb-generative-ai-2025 > We can teach you what to do because we have fully transformed our own customer experience - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: Fin website (Intercom's rebranded own site), October 29, 2026. https://fin.ai/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/intercom-can-teach-you-what-do-because-fully-2026 > 830M AI interactions - Who: Zendesk (the organization sells AI tools) - Source: Zendesk website (the company's own marketing page), October 24, 2026. https://www.zendesk.com/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-830m-ai-interactions-2026 > Automate up to 80% of interactions with Zendesk AI agents. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk website (the company's own marketing page), October 24, 2026. https://www.zendesk.com/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-automate-up-80-interactions-zendesk-ai-agents-2026 > Built to improve with every resolution. Proven across 4.8B+ resolutions. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk website (the company's own marketing page), October 24, 2026. https://www.zendesk.com/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-built-improve-every-resolution-proven-across-4-2026 > Resolve up to 80%+ of even the most complex service issues across any channel. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk website (the company's own pricing page), October 24, 2026. https://www.zendesk.com/pricing/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-resolve-up-80-even-most-complex-service-2026 > Conversations that don’t pass this verification are considered a Contained resolution. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/zendesk-conversations-dont-pass-verification-considered-contained-resolution-2026 > Conversations that pass this verification are considered a Verified resolution. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/zendesk-conversations-pass-verification-considered-verified-resolution-2026 > Messaging: By default, the conversation ends 2 hours after the first message in the conversation, though this can be increased up to 72 hours. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-messaging-default-conversation-ends-2-hours-after-2026 > Only conversations that were not escalated are evaluated for automated resolutions. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/5352026794010-About-the-automated-resolutions-platform-prior-to-resolution-tiers - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work?; What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/zendesk-only-conversations-were-not-escalated-evaluated-automated-2026 > Paying per automated resolution means you pay only for customer requests that were successfully resolved by an AI agent, without any escalation to a human agent. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/zendesk-paying-per-automated-resolution-means-you-pay-2026 > The system flags the interaction as an automated resolution because the live conversation ended with the AI agent rather than a human representative. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk help center (the company's own documentation), October 20, 2026. https://support.zendesk.com/hc/en-us/articles/11148708255770-A-conversation-displays-an-automated-resolution-but-it-was-escalated-to-an-agent - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/zendesk-system-flags-interaction-automated-resolution-because-live-2026 > Say a support agent is stuck on a ticket that needs an engineer. She shares her Glass session. The engineer jumps in, sees everything, they finish it together. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-say-support-agent-stuck-ticket-needs-engineer-2026 > Take new feature requests. They used to be copied by hand into Salesforce, about 300 a month. In May, an account manager saved a prompt to Glass that pulls them straight from customer calls. Now we capture 20,000 a month. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/ramp-take-new-feature-requests-they-used-copied-2026 > A third emails customers when something they asked for ships. More than 5,000 of those emails have gone out so far. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-third-emails-customers-when-something-they-asked-2026 > Agentforce conversations that result in a transfer to a human agent - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/salesforce-agentforce-conversations-result-transfer-human-agent-2026 > Agentforce conversations where Agentforce answered the customer's inquiry - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/salesforce-agentforce-conversations-where-agentforce-answered-customers-inquiry-2026 > Agentforce conversations where the user did not engage with Agentforce after starting the conversation - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What counts as resolved when AI answers customers? - Passage page: https://the-ai-record.com/passages/salesforce-agentforce-conversations-where-user-did-not-engage-2026 > available 24/7, in seven languages, and now that same help is live not just over chat, but on 1-800-NO-SOFTWARE too. - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-available-24-7-seven-languages-now-same-2026 > initially opening it to just 200 authenticated users within a four-week period - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-initially-opening-just-200-authenticated-users-within-2026 > Salesforce Help gets over 60 million visits annually from customers in need of assistance - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-salesforce-help-gets-over-60-million-visits-2026 > our support team still receives around two million support requests each year - Who: Salesforce (the organization sells AI tools) - Source: Salesforce (own site), October 2026. https://www.salesforce.com/agentforce/use-cases/customer-zero/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-support-team-still-receives-around-two-million-2026 > Some are seeing automated resolution rates spike 10% post-deployment. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (the company's own press release), September 14, 2026. https://www.zendesk.com/newsroom/press-releases/zendesk-introduces-specialized-ai-agents-purpose-built-for-your-business/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-some-seeing-automated-resolution-rates-spike-10-2026 > Later this year, Bank of America plans to expand EricaAssist to support additional servicing scenarios and business lines. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-later-year-bank-america-plans-expand-ericaassist-2026 > Used by more than 18,000 customer service representatives - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-used-more-than-18-000-customer-service-2026 > Cora, our digital assistant, deals with over 100 topics across mobile, online, and telephony banking. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-cora-digital-assistant-deals-over-100-topics-2026 > assisting nearly 50 million users since launch, surpassing 3 billion client interactions, and now averaging more than 58 million interactions per month. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), August 20, 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-assisting-nearly-50-million-users-since-launch-2025 > now averaging more than 58 million interactions per month - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), August 20, 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-now-averaging-more-than-58-million-interactions-2025 > we immediately activated them in our own support experience on day one. - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-immediately-activated-them-own-support-experience-day-2025 > with no human intervention required - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-no-human-intervention-required-2025 > Using our AI agent builder, we simply describe business procedures in natural language and the AI agent is ready to go. - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-using-ai-agent-builder-simply-describe-business-2025 > When hiring for our service team, we were able to rescope the skillset to deemphasize the need for deep workflow expertise. - Who: Austin Lacey, Director, Chatbots, AI & Automation, Zendesk (the organization sells AI tools) - Source: Zendesk blog (the company's own, signed by a Zendesk director), July 23, 2025. https://www.zendesk.com/blog/zendesk-insights/expertise/how-zendesk-uses-agentic-ai-to-deliver-instant-human-like-support-at-scale/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations train people to use AI?; Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/zendesk-when-hiring-service-team-were-able-rescope-2025 > When further assistance is needed, Bank of America's Mobile Servicing Chat capability connects clients with a live representative who can answer more complex servicing questions. - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), April 8, 2024. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2024/04/bofa-s-erica-surpasses-2-billion-interactions--helping-42-millio.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-when-further-assistance-needed-bank-americas-mobile-2024 > We need to rethink this and make customer service a human experience. Everyone who wants a human connection should get it. - Who: Sebastian Siemiatkowski, Chief Executive Officer and co-founder, Klarna - Source: CX Foundation (specialist customer-experience media and analyst site), February 17, 2026. https://cxfoundation.com/news/klarna-recruits-customers-as-service-reps - Grade C: An interview or press quote. Checked against the source. - Also filed under: What AI efforts failed or were rolled back?; Are customers happy with AI customer service? - Passage page: https://the-ai-record.com/passages/klarna-need-rethink-make-customer-service-human-experience-2026 ## How much code does AI write? What organizations say about the share of their code that AI writes, and how they count it. Topic page: https://the-ai-record.com/topics/ai-written-code > Nearly 60% of the code our engineers produce is now written by AI, which we estimate is about twice the industry average. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 7, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-nearly-60-code-engineers-produce-now-written-2026 > About 10% of our codes now is committed. That committed is built by agents, autonomous agents out there. Obviously, we check the code before it gets committed. - Who: Dara Khosrowshahi, Chief Executive Officer, Uber - Source: Uber Q1 2026 earnings call transcript (FactSet CallStreet corrected transcript, hosted on Uber's investor site), May 6, 2026. https://s23.q4cdn.com/407969754/files/doc_financials/2026/q1/Uber-Q1-26-Earnings-Call-Transcript.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/uber-about-10-codes-now-committed-committed-built-2026 > AI right now writes well over 50% of our code today, and that number is going up significantly, not down. - Who: Harley Finkelstein, President, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-ai-right-now-writes-well-over-50-2026 > When AI is used for Software Development, more than 40% of the new code is either AI-Generated or AI-Assisted. - Who: Walmart - Source: earnings call transcript via MarketBeat (third-party transcript site), November 20, 2025. https://www.marketbeat.com/earnings/reports/2025-11-20-walmart-inc-stock/ - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/walmart-when-ai-used-software-development-more-than-2025 > the last time I said the number was like 25% of code that's checked in. It involves people accepting AI-suggested solutions. That number is well over 30% now. - Who: Google (Alphabet) (the organization sells AI tools) - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), April 2025. https://transcripts.platformaeronaut.com/transcripts/GOOG-1Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/google-last-time-i-said-number-like-25-2025 > As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude. - Who: Anthropic (the organization sells AI tools) - Source: When AI builds itself, Anthropic (the Anthropic Institute, on anthropic.com), June 4, 2026. https://www.anthropic.com/institute/recursive-self-improvement - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/anthropic-may-2026-more-than-80-code-merge-2026 > Inspect is an internal software factory that writes more than two-thirds of Ramp's code. - Who: Ramp - Source: Ramp's Series F press release, as hosted by Ontario Teachers' Pension Plan (an investor in the round), June 4, 2026. https://otpp.com/en-ca/about-us/news-and-insights/2026/ramp-raises-series-f-at-44-billion-valuation/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-inspect-internal-software-factory-writes-more-than-2026 > Today, 75% of all new code at Google is now AI-generated and approved by engineers, up from 50% last fall. - Who: Google (Alphabet) (the organization sells AI tools) - Source: Google blog, April 22, 2026. https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/google-today-75-all-new-code-google-now-2026 > Even within traditional IDEs, ~70% of committed code is now AI-generated. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), March 16, 2026. https://x.com/praveenTweets/status/2033627282418655711 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-even-within-traditional-ides-70-committed-code-2026 > ~40% of daily code written at Coinbase is AI-generated. I want to get it to >50% by October. - Who: Brian Armstrong, CEO, Coinbase - Source: X post by Brian Armstrong, CEO, September 3, 2025. https://api.fxtwitter.com/brian_armstrong/status/1963315806248604035 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/coinbase-40-daily-code-written-coinbase-ai-generated-2025 > close to 100% of our code, probably somewhere between 95% and 100%, is written by or with LLMs today. - Who: Rob Witoff, Head of platform, Coinbase - Source: Over 95% of Coinbase’s code is now written with help of AI, Cointelegraph (crypto news site), July 14, 2026. https://cointelegraph.com/news/over-95-of-coinbases-code-is-now-written-with-ai - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/coinbase-close-100-code-probably-somewhere-between-95-2026 > I'd say maybe 20%, 30% of the code that is inside of our repos today and some of our projects are probably all written by software, - Who: Satya Nadella, Chairman and Chief Executive Officer, Microsoft (the organization sells AI tools) - Source: CNBC (news), April 29, 2025. https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-as-30percent-of-microsoft-code-is-written-by-ai.html - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/microsoft-id-say-maybe-20-30-code-inside-2025 > This blended figure reflects differences across our portfolio, where mature, large-scale systems and newer environments show varying levels of AI-generated code. All code written with AI is incorporated through our established development, review, and release processes. During the year, we also observed an approximately 40% increase in merged contributions, which we believe reflects improvements in development velocity and ongoing enhancements to our development processes. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/mercado-libre-blended-figure-reflects-differences-across-portfolio-2025 > We're around 70% AI-generated code - Who: Dropbox - Source: Dropbox CTO Ali Dasdan on moving from AI adoption to transformation, Dropbox engineering blog, September 23, 2026. https://dropbox.tech/culture/learnings-from-deploying-ai-at-company-scale - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/dropbox-were-around-70-ai-generated-code-2026 > I need to say, it has been 10 months now, I do, I minimally use any open source software, minimally. I just generate it to my requirements. And then when a supply chain attack happens, I'm like … affect me. - Who: Geoffrey Huntley, Creator of the Ralph loop; Latent Patterns, Latent Patterns (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 58:39. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How does AI help keep a product secure? - Context (the archive's note, not their words): Huntley is giving his closing statement after the moderator and another panelist raised software supply chain security. - Passage page: https://the-ai-record.com/passages/latent-patterns-minimal-open-source-generate-own-supply-chain-attack-2026 > And I don't see myself going back to writing code by hand. It's been two and a half years since I manually wrote code by hand. - Who: Geoffrey Huntley, Creator of the Ralph loop; Latent Patterns, Latent Patterns (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 5:31. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Huntley is speaking in his opening four-minute statement for the side arguing that loops work today, in an Oxford-style debate. - Passage page: https://the-ai-record.com/passages/latent-patterns-no-hand-written-code-two-and-a-half-years-2026 > So now I felt comfortable claiming that we were truly delegating work to the agents. At this point, it had been three months since launching … champions program and we were seeing great results. AI authored code was up by 69% … savings increased 37% and automated PR's increased 21 times. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 12:27. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Jones is reporting results about three months after the AI champions program began at Block. - Passage page: https://the-ai-record.com/passages/block-champions-three-months-ai-authored-code-up-69-percent-2026 > Inside of Airbnb, we're double this. So we're closer to 59% today. So more than half of code inside of Airbnb today is primarily authored by AI. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-inside-airbnb-were-double-so-were-closer-2026 > Diff Vader and the Death Star were built almost entirely with Claude Code, but not by typing “build me an AI reviewer” and shipping whatever came back. We used spec-based development — Superpowers for the core methodology, augmented with our own skills — which forces every piece of work through brainstorm → written spec → reviewed implementation plan → execution → code review, with discrete human checkpoints in between. - Who: Dave North, Leads the technical operations group, Rewind - Source: Review is the bottleneck now: How we let AI approve pull requests (safely), Rewind blog, June 1, 2026, at What I’d tell another engineering leader. https://rewind.com/blog/ai-approve-pull-requests-safely/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/rewind-built-almost-entirely-claude-code-spec-based-development-2026 > Our internal background coding agent went from <1% of all code changes to 8% in just a few months. There is zero human authoring. Engineers review and approve, but the code is written entirely by AI agents. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), March 16, 2026. https://x.com/praveenTweets/status/2033627282418655711 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/uber-internal-background-coding-agent-went-1-all-2026 > There is zero human authoring. Engineers review and approve, but the code is written entirely by AI agents. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), March 16, 2026. https://x.com/praveenTweets/status/2033627282418655711 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-there-zero-human-authoring-engineers-review-approve-2026 > building and shipping an internal beta of a software product with 0 lines of manually-written code - Who: OpenAI (the organization sells AI tools) - Source: Harness engineering: leveraging Codex in an agent-first world, OpenAI (own site), February 11, 2026. https://openai.com/index/harness-engineering/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-building-shipping-internal-beta-software-product-0-2026 > Given zero hand-written code and zero traditional review, we required a system that could: Grow from cascades of natural-language specifications Be validated automatically without semantic inspection of source - Who: StrongDM - Source: Techniques: The Validation Constraint (StrongDM Software Factory), StrongDM AI, February 2026, at The Validation Constraint. https://factory.strongdm.ai/techniques - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/strongdm-zero-hand-written-code-zero-traditional-review-2026 > what we call minion internally. So this is … background coding agent. So around 15% of all of the production … go through minion right now. … Once you have for the task that I have really good spec, or I want to do minimal steering, I can hand off that to a minion. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 50:46. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Medisetty is describing Uber's background coding agent, Minion. The word cut after 'production' is pull requests (PRs), which both transcriptions rendered as 'peers'; he says 'PRs' in the next sentence. - Passage page: https://the-ai-record.com/passages/uber-15-percent-production-prs-go-through-minion-background-agent-2026 > Errors arise then get fixed quickly. The error rate remains small and constant, perhaps rarely completely clean but steady and manageable, not exploding or deteriorating. This may indicate that the ideal efficient system accepts some error rate, but a final "green" branch is needed where an agent regularly takes snapshots and does a quick fixup pass before release. - Who: Wilson Lin, Researcher, Cursor (author of the Towards self-driving codebases post), Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026, at Throughput and tradeoffs: Commit correctness. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How autonomous are the agents, and who checks their work? - Passage page: https://the-ai-record.com/passages/cursor-ideal-system-accepts-error-rate-green-branch-2026 > Today, … used by several tens of thousands of engineers every single week to review all of their code. So we also know, generally speaking, the number of pull requests that each of these people are writing. The median … user writes 50 pull requests a month, so call it about … per work day. The 90th percentile writes 500 pull requests per month. That is a drastic difference between the median and the P90. The P99 is in the thousands of pull requests. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 10:00. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/greptile-median-user-writes-50-prs-month-p90-500-2026 > And I like to define what we've been calling frontier developers inside of Amazon by three behaviors that I've been seeing. One is hands off coding, frontier developers write maybe … of the code that they produce. The rest is agents. The second is that they interact with their agents infrequently. They'll aim to get their coding assistant to run for up to hours at a time without their intervention. And third is that they minimize idle time. These frontier developers tend to run multiple agents in parallel, … through a backlog of tasks. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 1:40. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How autonomous are the agents, and who checks their work?; How many agents can one person run at once?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Liguori is defining what Amazon calls frontier developers. - Passage page: https://the-ai-record.com/passages/amazon-frontier-developers-three-behaviors-hands-off-coding-2026 > And all of this led to twice the number of lines of code per engineer year over year. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory (talk, AI Engineer World's Fair 2026), AI Engineer (conference organizer) YouTube channel, August 21, 2026. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/uber-all-led-twice-number-lines-code-per-2026 > AI-authored backend code had a revert rate of 0.53%, compared to 5.39% for human-authored. On the frontend, it was 0.22% versus 2.00%. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at The data. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How is AI used in code review? - Passage page: https://the-ai-record.com/passages/intercom-ai-authored-backend-code-revert-rate-0-53-2026 > Pull requests arrive faster than human-driven review can stay meaningful. - Who: Bence A. Tóth, Author on the Zendesk Engineering blog (role not stated on the pages read), Zendesk (the organization sells AI tools) - Source: Zendesk Engineering blog on Medium (the company's own engineering blog; read through its RSS feed), March 16, 2026. https://medium.com/feed/zendesk-engineering - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Does AI make developers more productive?; What slows teams down once agents write the code? - Passage page: https://the-ai-record.com/passages/zendesk-pull-requests-arrive-faster-than-human-driven-2026 > we handled more than two hundred and fifty automated migrations, cumulatively nine million lines of code automatically for our engineers. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory (talk, AI Engineer World's Fair 2026), AI Engineer (conference organizer) YouTube channel, August 21, 2026. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Can AI do large code migrations? - Passage page: https://the-ai-record.com/passages/uber-handled-more-than-two-hundred-fifty-automated-2026 ## How autonomous are the agents, and who checks their work? How much agents do with no person in the loop, where people still approve, and which decisions organizations keep for people. Topic page: https://the-ai-record.com/topics/autonomy-and-oversight > In the first two weeks of April, Builderbot and other AI tools reviewed over 90% of production code change requests that were made. As of early April, Builderbot is executing over 200,000 operations per day and is making 15% of our production code changes nearly fully autonomously, with people only intervening to make the final decision to push to production. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To Our Shareholders. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/block-builderbot-reviewed-over-90-percent-code-change-requests-2026 > To carry out this work, the web team gave coding agents clearly defined tasks and a consistent process to follow. Each task started with a written plan agreed on by the engineer and agent. The agent made the changes, ran the relevant builds and tests locally, and recorded the results for an engineer to review before accepting the code. - Who: Dropbox - Source: How AI-assisted development helped us build the new Dropbox experience, Dropbox engineering blog (dropbox.tech), November 5, 2026, at Making the new experience usable by more people. https://dropbox.tech/machine-learning/how-aiassisted-development-helped-us-build-the-new-dropbox-experience - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/dropbox-web-team-gave-coding-agents-clearly-defined-2026 > A human intervenes at two points: approving the spec before implementation begins, and merging the PR after review. Everything else runs automatically. - Who: Warp (the organization sells AI tools) - Source: Set up your software factory, Warp docs, October 28, 2026, at How the loop works. https://docs.warp.dev/guides/agent-workflows/set-up-a-software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/warp-human-intervenes-two-points-approving-spec-2026 > Analysts review the agent-spread financials inside our existing credit tooling. Every line item traces back to a cell or page region in the source documents, so analysts can check the agent's work. - Who: Ramp - Source: Agentic Credit Analysis, Ramp Builders blog, October 22, 2026, at The agentic loop. https://builders.ramp.com/post/credit-analysis-using-semantics - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in finance, legal and operations? - Passage page: https://the-ai-record.com/passages/ramp-analysts-review-agent-spread-financials-2026 > Pulling a senior engineer out of deep work to catch a routine mistake costs far more than an automated review. People on the team are still responsible for: … Reviewing risky code such as billing, the auto-updater and CI workflows. … Owning the outcome. A person is accountable for every change that ships, whoever typed it. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Where humans stay in the loop. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Context (the archive's note, not their words): The two cuts drop three items from the post's list of what people stay responsible for: deciding whether to build a change, reviewing architecture that sets a pattern others will copy, and reading the evidence the checks produce. - Passage page: https://the-ai-record.com/passages/anthropic-people-still-own-risky-code-and-outcome-2026 > And the goal for all of this is the only time you're allowed to look is before you launch the experiment. So you review the data science findings, the hypotheses, the UX writing and the code all at once. So like one human touch. And then it launches and then you can review whether or not it's a winner or loss, but it has already started iterating on the next rounds. I would say like where we are in theory is pretty like is great, where we are in practice is almost there. … we're now at like five live concurrent experiments on things that we would not have been doing otherwise. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 25:48. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in product and design? - Context (the archive's note, not their words): Mechaber is describing how Auto XP, his team's automated growth-experiment setup, hands work to a person. - Passage page: https://the-ai-record.com/passages/figma-auto-xp-one-human-touch-five-live-experiments-2026 > We built a Software Factory: non-interactive development where specs + scenarios drive agents that write code, run harnesses, and converge without human review. - Who: StrongDM - Source: How StrongDM's AI team build serious software without even looking at the code, Simon Willison's Weblog, February 7, 2026. https://simonwillison.net/2026/feb/7/software-factory/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/strongdm-we-built-software-factory-non-interactive-development-2026 > We also use autonomous AI agents to support portions of our engineering work. These agents perform defined tasks such as contributing to code reviews, assisting with documentation, reviewing incidents, and handling certain migration and update activities across parts of our codebase. They operate as part of our broader development workflow and complement the work of our engineering teams. - Who: Mercado Libre - Source: MercadoLibre, Inc. Form 10-K, fiscal year 2025, SEC EDGAR, February 25, 2026, at Item 1. Business. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000006/meli-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/mercado-libre-autonomous-ai-agents-support-portions-engineering-2025 > So this probably understates the complexity. The reality is most things are multi-runner. Most things have deterministic code plus LLM plus statistical method, or human with LLM in conjunction. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 10:52. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Long has just walked through gathering, filtering and summarizing news as one small step of the investing function. - Passage page: https://the-ai-record.com/passages/coatue-most-steps-multi-runner-code-llm-statistics-human-2026 > And, of course, people, people are important runners in the system, right? So imagine a function meeting the management team of a company. … I probably don't want to send an if statement into that meeting. I think there's many cases where, because of judgment and the ability to navigate the real world, and all of these things, a person is the best runner for a given piece of the function. - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 08:35. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Long has just listed code, statistics and language models as the ways to carry out each step of the investing function. - Passage page: https://the-ai-record.com/passages/coatue-people-are-the-best-runner-for-some-steps-2026 > The hardest and most consequential work are the simulations and the evals in addition to the test pipeline. We need to test whether the rewritten software preserves the behavior customers depend on, performs under real conditions, and handles failures correctly. The agents can iterate, but passing those comprehensive evals has to be a prerequisite for deployment. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Jeetu Patel's LinkedIn post about modernizing Cisco's code, LinkedIn, November 2, 2026. https://www.linkedin.com/posts/jeetupatel_lot-of-discussion-on-a-fortune-article-link-activity-7523171493376090112-C67s - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; Can AI do large code migrations? - Passage page: https://the-ai-record.com/passages/cisco-hardest-most-consequential-work-simulations-evals-2026 > This is the most important human checkpoint in the loop. An accurate spec leads to a correct implementation; a wrong spec leads to a correct implementation of the wrong thing. - Who: Warp (the organization sells AI tools) - Source: Set up your software factory, Warp docs, October 28, 2026, at 2. Human review: approve the specs. https://docs.warp.dev/guides/agent-workflows/set-up-a-software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/warp-most-important-human-checkpoint-loop-accurate-spec-2026 > The models showed strong financial reasoning but had a few common failure modes. To address these gaps, we surround the agent with deterministic code that verifies calculations, avoids double-counting, and reconciles duplicate figures across statements. - Who: Ramp - Source: Agentic Credit Analysis, Ramp Builders blog, October 22, 2026, at Deterministic scaffolding. https://builders.ramp.com/post/credit-analysis-using-semantics - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in finance, legal and operations? - Passage page: https://the-ai-record.com/passages/ramp-models-showed-strong-financial-reasoning-2026 > This is agent chip working in incident mode and leveraging the MCP to get very relevant data. So, what happens here is whenever we identify and trigger a new incident, agent chip joins the channel and is able to start an investigation, understanding logs, understanding traces, being able to connect to that exact service. And of course, the code … to see what's going on and to help the human with a very high signal. We don't want the agent to be saying a lot of things during an incident … We wanted to maintain high signal and we give the humans also control of how much agent chip will be able to support there. - Who: Felippe Felisola Caso, Monzo (conference affiliation, 2026), Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 10:00. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI fix bugs and handle incidents? - Context (the archive's note, not their words): Felisola Caso is describing how Monzo's in-house agent, Agent Chip, takes part in incident response. - Passage page: https://the-ai-record.com/passages/monzo-agent-chip-working-incident-mode-joins-channel-investigation-2026 > This has become very powerful to us at Tessl. We have build a maintenance agent that periodically reads historic pull requests, findings and logs, then proposes changes like as pull requests that our team reviews. - Who: Simon Maple, Head of Developer Relations, Tessl (the organization sells AI tools) - Source: Anatomy of an Agentic Code Review, Tessl blog, October 9, 2026, at 6. A memory. https://tessl.io/blog/anatomy-of-an-agentic-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/tessl-maintenance-agent-periodically-reads-historic-pull-requests-2026 > Uber lawyers then review each suggestion, deciding whether to accept, reject, or modify it before final incorporation into the client response. - Who: Austin Greco, Staff Software Engineer, Michelangelo (Uber's AI platform), Uber - Source: Uber engineering blog, October 8, 2026. https://www.uber.com/us/en/blog/building-ubers-redlining-agent/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in finance, legal and operations?; Can AI do legal work like contract review? - Passage page: https://the-ai-record.com/passages/uber-uber-lawyers-then-review-each-suggestion-deciding-2026 > Thanks to these autonomous loops, 60%, 60% of UX issues that are identified by either a customer or a salesperson or a CX person or ourselves are fixed within 24 hours. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026, at 14:19. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/ramp-thanks-these-autonomous-loops-60-60-ux-2026 > But rather than killing the outer loop, I think that we believe and the industry has just started to really kind of coalesce on this idea that … really expanding the outer loop. Rather than removing humans from the code review process, we are moving their responsibilities up a layer. - Who: Will Bond, Staff Engineer, AI Foundations team, Development Platform organization, Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 13:43. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): Will Bond is closing the talk on what happens to human reviewers as agents write and review more of the code. - Passage page: https://the-ai-record.com/passages/uber-rather-than-killing-outer-loop-i-2026 > I am running an organization of around 50-60 agents, five of whom are interfacing with around 10 humans in the outside world: myself, my 5-person core game design team, my accountant, my chief of staff, and a few others. Only Fable is allowed to talk to humans, via Slack and email. - Who: Steve Yegge, Independent; builds and runs Wheelhouse, a solo multi-agent software factory (2026), Steve Yegge (independent) (the organization sells AI tools) - Source: Fences, not Sandboxes, yegge.ai (Steve Yegge), August 24, 2026. https://yegge.ai/essays/fences-not-sandboxes/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run?; How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/steve-yegge-running-organization-around-50-60-agents-2026 > if this is an autonomous diff coming from Minion, we want to give a human reviewer some confidence that this diff has gone through a lot of self-improvement already - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory (talk, AI Engineer World's Fair 2026), AI Engineer (conference organizer) YouTube channel, August 21, 2026. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/uber-if-autonomous-diff-coming-minion-want-give-2026 > So the key thing here is that this is actually a managed loop that you go to … We don't want thousands of loops being set up across the company without any bounds. You have a managed surface that you go to to set up the loop so it runs on Sunday when we know we have better CI capacity available. - Who: Adam Huda, Senior Engineering Leader, AI Developer Tools, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 16:25. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Huda is describing how Uber runs recurring maintenance agents against its services. - Passage page: https://the-ai-record.com/passages/uber-managed-loop-maintenance-skills-sunday-ci-capacity-2026 > Connect the factory with your GitHub issues, tag the ones you want it to pick up (choose trivial issues and bugs first), run it, and review the resulting PR and the agents' PR comments. - Who: Noah Blauensteiner, Author, Xebia articles, Xebia (the organization sells AI tools) - Source: Shining Some Light in Dark Agentic Software Factories, Xebia articles, August 12, 2026, at Shining some light on Agentic Software Factories. https://xebia.com/articles/shining-some-light-in-dark-agentic-software-factories/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/xebia-connect-factory-github-issues-tag-ones-2026 > PostHog is just a little further than that with agents writing about 70% of our PRs, but humans still skim at least 80% of those. - Who: Jina Yoon, Author, PostHog newsletter, PostHog (the organization sells AI tools) - Source: Can software factories actually work?, PostHog newsletter, August 11, 2026, at What is a software factory?. https://posthog.com/newsletter/software-factories - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/posthog-agents-writing-about-70-our-prs-2026 > But the real problem with all of this is I built this and I hit a little bit of AI psychosis. And I started shipping 10 design docs a day. And soon, the team was fighting my slop. … This last rule was … ship a design doc, you require people to actually go read it. And with this last standard, we suddenly had design docs … incredibly high quality. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 3:37. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design?; What AI efforts failed or were rolled back? - Context (the archive's note, not their words): Gupta is describing the design-document tool he built for the team and the rule that followed. - Passage page: https://the-ai-record.com/passages/boundary-ai-psychosis-ten-design-docs-day-read-rule-2026 > This tool basically visualizes our dependency graph internally with some external dependencies as well, … But what's more interesting is we can go build CLI tools that guarantee that certain invariants can't be broken. And what this does is when Claude builds a new package or adds a dependency that's leaky, we now have CI/CD changing or a simple git commit history that tells us exactly where things break. And by this, we're actually able to make our architecture change. We haven't changed our architecture in the last three or four months. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 4:05. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Context (the archive's note, not their words): Gupta is describing how the team keeps the BAML code base from drifting when engineers do not review each other's code. - Passage page: https://the-ai-record.com/passages/boundary-dependency-graph-cli-invariants-architecture-unchanged-months-2026 > We do no code reviews. We require every engineer to work on things in parallel, and we have no standardization how people do AI. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 0:26. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Does AI make developers more productive?; Do engineers still read the code agents write? - Context (the archive's note, not their words): Opening of a conference talk by the chief executive of Boundary, the company that builds the BAML programming language, on how its team works. - Passage page: https://the-ai-record.com/passages/boundary-no-code-reviews-parallel-no-ai-standardization-2026 > So today I want to talk about what I think works in the real world and what we've started doing at HumanLayer, which, to be clear, is still building loops, right? I think loops are super powerful, but we can design loops and still read the code. In fact, we can design loops that make it easier to read the code because the loops are making the code better. - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 4:03. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Do engineers still read the code agents write? - Context (the archive's note, not their words): Mistele is opening the part of his talk about how his company builds agent loops, after arguing that loops which produce unreadable pull requests are the wrong design. - Passage page: https://the-ai-record.com/passages/humanlayer-loops-still-read-the-code-2026 > And so my advice is … instead of trying to automate everything end-to-end, build these small incremental loops throughout your system, and you will wake up one day and you will be moving two to three times faster while still being able to read the code, while still owning the architecture. - Who: Dex Horthy, CEO, HumanLayer, HumanLayer (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 48:33. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; Do engineers still read the code agents write? - Context (the archive's note, not their words): Horthy is answering the moderator's question on how to reach fully autonomous software factories if two to three times is what coding agents give most engineers. - Passage page: https://the-ai-record.com/passages/humanlayer-small-incremental-loops-two-to-three-times-faster-owning-architecture-2026 > Just to kind of put an example to this, we've created these tools and these systems where researchers can run experiments directly from Slack. We … avoid this state of being bottlenecked on humans launching, and reviewing, and babysitting runs. And we actually have an entire team just working on automating every part of the research work that isn't freeing up the researchers' time to work on their most ambitious ideas. So every person on the ML team gets access to this fleet of agents they can basically train models directly from Slack. - Who: Lee Robinson, Machine Learning Engineer, Model Behavior, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Recursive Model Improvement, AI Engineer (YouTube), July 15, 2026, at 17:14. https://www.youtube.com/watch?v=q4Tr-DknG2M - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Robinson is describing how Cursor's machine learning team uses agents in its own model-training work. - Passage page: https://the-ai-record.com/passages/cursor-researchers-run-experiments-from-slack-2026 > One is you start measuring how many human touches you still do to have the agent do the right thing. That's supposed to go down. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Coding Agents Don't Scale Themselves. Neither Do Your Teams. The Rise of Agent Enablement, AI Engineer World's Fair 2026 (YouTube), July 2, 2026, at 09:38. https://www.youtube.com/watch?v=zCJtYuqwm7E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Debois names the first of two metrics he has come to believe in for judging how productive a team is with coding agents. - Passage page: https://the-ai-record.com/passages/tessl-human-touches-metric-agent-right-thing-2026 > With this architecture, agentic reasoning is used for intake and context gathering, but every autonomous risk decision is still made by an auditable model, governed under an approved policy. - Who: Blake Williams, Product Manager, Ramp - Source: Ramp engineering blog, June 30, 2026. https://builders.ramp.com/post/agentic-risk-operations - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in finance, legal and operations?; How do banks and other regulated organizations use AI within the rules? - Passage page: https://the-ai-record.com/passages/ramp-architecture-agentic-reasoning-used-intake-context-gathering-2026 > I got tired of watching old feature flags pile up… I built an agent for cleaning up stale feature flags. It runs in my local workflow and it doesn't merge any code. The models, the easy part. The judgment in front of it is the whole game. Every flag gets scored the way an experienced reviewer would analyze it. How many modules does it touch? Is it multi-variant? Does it share a component? Then it checks the experiment data. Is the rollout frozen? Is there a sample ratio mismatch? Is there a variant already rolled out to 100 percent? - Who: Andrew Orobator, Senior Software Engineer, Reddit - Source: Scale the Judgment, Not the Model, AI Engineer (YouTube), June 30, 2026, at 11:18. https://www.youtube.com/watch?v=6MudaeKdBSk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): The speaker is a Reddit Android engineer. In the talk he describes this agent as running in his local workflow and on his laptop, and says some of his other practices are from his personal side projects. - Passage page: https://the-ai-record.com/passages/reddit-built-agent-clean-stale-feature-flags-scored-reviewer-2026 > No organization starts with a fully autonomous software factory. Autonomy is a maturation process that is gradual and specific to every organization’s readiness and comfort level. - Who: Factory (the organization sells AI tools) - Source: Factory 2.0: From coding agents to software factories, Factory (own site), June 15, 2026. https://factory.ai/news/software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/factory-no-organization-starts-fully-autonomous-software-2026 > Well-defined, measurable tasks run with simple Droid agents or skills. Automations coordinate recurring workflows with a shared objective and memory. Remote and persistent execution leverages Droid Computers for long running or local agents. Multi-agent autonomous execution called Missions solve complex tasks over hours or days by decomposing work into parallel tracks to handle. - Who: Factory (the organization sells AI tools) - Source: Factory 2.0: From coding agents to software factories, Factory (own site), June 15, 2026. https://factory.ai/news/software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/factory-well-defined-measurable-tasks-run-simple-droid-2026 > Mode · Automation: PR review. Durable session, woken by an external system, often no human in the loop. - Who: Burke Libbey, Principal Engineer, Shopify - Source: Under the River, Shopify engineering blog, May 28, 2026. https://shopify.engineering/under-the-river - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/shopify-mode-automation-pr-review-durable-session-woken-2026 > One engineer might have three or four agents working in parallel: one implementing an API, one writing integration tests, one fixing a bug, one refactoring a legacy module. The engineer reviews each as it finishes, provides feedback, and merges. - Who: Coinbase - Source: Coding Had a Concurrency Problem: How Mux Helped Solve It, Coinbase engineering blog (byline: Alessandro Rodi, Adam Wolf), May 11, 2026, at The Shift. https://coinbase.com/blog/coding-had-a-concurrency-problem-how-mux-helped-solve-it - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/coinbase-one-engineer-might-have-three-four-agents-2026 > So for every ticket that's getting created in Linear, we spawn in the cloud agents. … if we have a feature flags for a specific thing that's rolled out. And if it's rolled out for two weeks with 100%, the system kind of like a single source like, hey, it's a stale feature flag at this time. You can remove it. So then we have this to create an automatic issue in Linear. And since that is hooked up with Cursor, it triggers a cloud agent to remove the feature flag. So it's kind of like completely automatic once the system knows that it's rolled out to everyone. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 1:09:19. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Zakariasson is answering an audience question about whether agents read tasks from an issue tracker, and describing how Cursor's own team does it. - Passage page: https://the-ai-record.com/passages/cursor-stale-feature-flag-cloud-agent-removes-itself-2026 > It doesn't really matter who generated what or what was behind it. If it’s an agent or a human, it’s very much outcome-based, and you also want to have someone who’s accountable for the outcome. - Who: Christian Ryan, Applied AI lead, Anthropic (the organization sells AI tools) - Source: Let’s Talk Agentic Development: Spotify x Anthropic Live, Spotify Engineering, April 6, 2026. https://engineering.atspotify.com/2026/4/anthropic-agentic-development - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/anthropic-doesn-t-really-matter-who-generated-what-agent-2026 > The human role shifts from guiding each line of code to defining the problem and setting review criteria. - Who: Michael Truell, Author of the Cursor blog post; CEO of Cursor per the lead record (the page gives no title), Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026, at Cloud agents and artifacts. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do engineering and product jobs change when agents write the code? - Passage page: https://the-ai-record.com/passages/cursor-human-role-shifts-guiding-each-line-2026 > More than one-third of the PRs we merge are now created by agents that run on their own computers in the cloud. - Who: Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/cursor-more-than-one-third-prs-merge-now-2026 > Thirty-five percent of the PRs we merge internally at Cursor are now created by agents operating autonomously in cloud VMs. - Who: Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/cursor-thirty-five-percent-prs-merge-internally-cursor-2026 > Some changes are beyond automation: architectural decisions, complex trade-offs, and organizational implications. The system recognizes these boundaries and escalates. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Lessons learned: Building systems engineers trust. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/customer-io-some-changes-beyond-automation-architectural-decisions-2025 > AskHR answers 94% of common employee inquiries in minutes or less, without human intervention, any time of the day or night. - Who: IBM (the organization sells AI tools) - Source: IBM (own site), August 26, 2025. https://www.ibm.com/think/insights/enterprise-transformation-extreme-productivity-ai - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/ibm-askhr-answers-94-common-employee-inquiries-minutes-2025 > When AI suggests code changes, it does not commit them directly. The author remains in control—reviewing, editing, and deciding whether to accept the suggestion by explicitly clicking ‘apply change’ option. - Who: Sneha Tuli, Principal Product Manager, Microsoft (the organization sells AI tools) - Source: Enhancing Code Quality at Scale with AI-Powered Code Reviews, Engineering@Microsoft blog, July 14, 2025, at Solving the real problems in PR reviews with AI. https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/microsoft-suggests-code-changes-does-not-commit-2025 > I also don't want to surrender understanding. … I want people to write this verification loops because it's going to give me conviction on the product that it delivered to my customer, and if anything goes wrong, I also want to … and I want to prevent broad categories of misalignment. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 36:26. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Context (the archive's note, not their words): Rauch is describing the checks he wants engineers to build around agent-written code at Vercel. - Passage page: https://the-ai-record.com/passages/vercel-i-want-people-to-write-verification-loops-conviction-product-2026 > One of the things that we're doing is we're shifting our focus from what you would call … or developer experience to verification, to evaluation. So … more about creating the checks and balances for the agent and the task. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 33:57. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Context (the archive's note, not their words): Rauch is answering how engineers building with agents at Vercel verify that the work meets his standard. - Passage page: https://the-ai-record.com/passages/vercel-shifting-our-focus-from-framework-developer-experience-to-verification-evaluation-2026 > If we have a very specific agent with a very specific prompt that we optimized — and we know that, over history, it was very successful in fixing a certain category of bugs — then we develop trust in that particular agent configuration - Who: Lars Grammel, Engineer, Vercel, Vercel (the organization sells AI tools) - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/vercel-very-specific-agent-prompt-optimized-develop-trust-2026 > actually talked to Claude about the code and say, all right, these are the questions that I would have. Can you go investigate it? So it is kind of … code review, but still human-driven, and for the really important ones. And for the ones that are cosmetic visual changes, it's much more like we'll fix forward if we need to fix forward. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 11:13. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review?; Do engineers still read the code agents write? - Context (the archive's note, not their words): Continuing his answer on review. The sentence before this passage, on not reviewing every line of a pull request, is not quoted because the two passes disagreed on its wording. - Passage page: https://the-ai-record.com/passages/anthropic-labs-talk-claude-about-code-instead-reading-every-line-2026 > I think we found that the state-of-the-art in December was you can probably, run this for about two weeks. By the end of those two weeks, you’d find that, hey, you want to, change the color of a button. Well, it turns out this button is implemented in, 10 different places, and they, have All these different variations, and oh, you forgot one of them, and actually it’s a slightly different color in one spot. - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:44:21. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review?; Do engineers still read the code agents write?; What AI efforts failed or were rolled back? - Context (the archive's note, not their words): Yan was describing an internal experiment building real products by auto-merging AI-written code with no code review, to see how many weeks it could run before the codebase became unmanageable. - Passage page: https://the-ai-record.com/passages/cognition-vibe-coding-no-review-experiment-two-weeks-2026 > we found it like a platform where we put this agent and a number of these other specialist agents into our CI infrastructure. And we accept prompts in order to do work, it goes through our CI checks and all our tests run and quality checks, … - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 05:30. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Ramakrishnan is describing how Minion and other specialist agents sit inside Uber's CI infrastructure. - Passage page: https://the-ai-record.com/passages/uber-minion-agents-in-ci-prompts-go-through-checks-tests-2026 > And I'm like, okay, I don't want you to be 100% percent using AI. I want you be like 90 or 95%. I want you to be able to go in and look at the code and say, oh yeah, there's a line that's wrong. - Who: Farhan Thawar, Head of Engineering, Shopify - Source: How AI is changing software engineering at Shopify with Farhan Thawar, The Pragmatic Engineer, July 2, 2025, at Interesting detail: Shopify’s interview process. https://newsletter.pragmaticengineer.com/p/how-ai-is-changing-software-engineering - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): From a live podcast conversation at LDX3 in London between Gergely Orosz and Farhan Thawar, about how Shopify lets candidates use AI in its interviews. - Passage page: https://the-ai-record.com/passages/shopify-dont-want-you-100-percent-using-ai-90-95-2025 > We are spending more and more compute on sort of like secondary monitoring. So you have, you know, you have all of the compute that is going into the primary system, you know, the agent that is like doing work. And then you have all of the compute that is going into monitoring, you know, the primary agent to make sure that it's like not taking too high risk of actions … If anything looks like, you know, maybe it's getting prompt injected and just like, you know, intervening with that. And the majority of our investment on the API stack is like, you know, actually going into the safety stack. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 34:37. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: How does AI help keep a product secure? - Context (the archive's note, not their words): Asked how he thinks about the risks of AI, Sottiaux describes how OpenAI spends compute on monitoring its agents. - Passage page: https://the-ai-record.com/passages/openai-compute-on-secondary-monitoring-safety-stack-2026 > What the working model that we've used for all of our internal agent development is basically a tripod. It's a GTM engineer, a data scientist, and then the best person, the best subject matter expert in that function. And they sit down shoulder to shoulder and document the best practice for that particular function and then encode that in workflows that ultimately form an agent. We then keep a human in the loop so that human will basically QA the agent. The agent won't be allowed to autonomously execute its action. And then over time as the subject matter expert has less feedback for the agent, we finally pull that person. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 03:24. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Do organizations reorganize teams around AI?; Who builds an organization's agents: every team, or a central group? - Context (the archive's note, not their words): Vercel's COO explains how her team builds internal agents ('GTM' is go-to-market). - Passage page: https://the-ai-record.com/passages/vercel-tripod-engineer-data-scientist-expert-human-in-loop-2026 > When we required 100% correctness before every single commit, it caused major serialization and slowdowns of effective throughput. Even a single small error, like an API change or typo, would cause the whole system to grind to a halt. - Who: Wilson Lin, Researcher, Cursor (author of the Towards self-driving codebases post), Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026, at Throughput and tradeoffs: Commit correctness. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/cursor-requiring-100-percent-correctness-every-commit-slowdown-2026 > And I think this is like a super, super important point. There is no glory in automating things for the sake of automating things. The goal is not to take as many things and make them … as possible for us. The goal is for any given piece of this function, what can I do to maximize absolute quality? - Who: Frank Long, Works at Coatue, Coatue - Source: AI Agents Will Upend Investing Even More Than Coding - Frank Long, Coatue, AI Engineer (YouTube), November 6, 2026, at 09:05. https://www.youtube.com/watch?v=AYFcmnQrZCg - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Long is explaining how his firm decides what to hand to a language model and what to leave to code or people. - Passage page: https://the-ai-record.com/passages/coatue-no-glory-in-automating-for-its-own-sake-2026 > Running an experiment and deploying its output are separate decisions. We can explore what is possible, assess the quality, and understand the economics before deciding what belongs in production. We won’t know how far this approach can take us until we try. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Jeetu Patel's LinkedIn post about modernizing Cisco's code, LinkedIn, November 2, 2026. https://www.linkedin.com/posts/jeetupatel_lot-of-discussion-on-a-fortune-article-link-activity-7523171493376090112-C67s - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations? - Passage page: https://the-ai-record.com/passages/cisco-running-experiment-deploying-its-output-separate-2026 > So what we built entire lifecycle around skills, so we have core skills and domain-specific skills, all of that go into a managed skills marketplace. We have 2,500 skills there right now, and it goes through a whole bunch of … automated reviews, which ensures a baseline skill quality for any skills that we have. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 07:35. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Medisetty is describing how Uber manages the agent skills its engineers build. - Passage page: https://the-ai-record.com/passages/uber-managed-skills-marketplace-2500-skills-automated-reviews-2026 > this is great for doing like toil sort of workloads, but to build more advanced like end-to-end features, we really … to be able to validate the feature first … validate sooner before we push to CI, that would be a big benefit. - Who: Adam Huda, Senior Engineering Leader, AI Developer Tools, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 13:48. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Huda is demonstrating Minion, Uber's cloud coding agent, which stops at a draft pull request. - Passage page: https://the-ai-record.com/passages/uber-minion-draft-pr-validate-feature-before-ci-2026 > And every morning, we walk into the office to a small incremental PR that's low risk. And when we first did this, it was actually really frustrating, and we turned the loop off … because we had to constantly update the skill, we had to constantly check out the branch, change the skill, change the code, commit and push, and our loop was actually really high friction, right? - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 13:48. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; What AI efforts failed or were rolled back? - Context (the archive's note, not their words): Mistele has just described scheduling the migration loop to run once a day as a CI workflow, and goes on to the problem HumanLayer hit with the first version. - Passage page: https://the-ai-record.com/passages/humanlayer-daily-loop-pr-turned-off-high-friction-2026 > So the next thing we're going to do is add flow control. Because the other problem that we had when we did this was that if we were at a customer site for a week or if we were traveling or spent six days working on slides instead of writing code, the PRs from all of our loops would just stack up. They'd duplicate work. They'd conflict. And we wouldn't get around to doing it. And the loop's work is important, but it's not that important. And so now we just had all this junk we had to deal with… - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 15:12. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What slows teams down once agents write the code? - Context (the archive's note, not their words): Mistele explains why HumanLayer added a rule that a loop skips its run while its previous pull request is still open. - Passage page: https://the-ai-record.com/passages/humanlayer-loop-prs-stack-up-customer-site-2026 > The job changed, and most of us didn't even notice, we're harness engineers now. Our work is the system that helps produce and verify the code. The constraints, the gates, the skills, the verification. - Who: Andrew Orobator, Senior Software Engineer, Reddit - Source: Scale the Judgment, Not the Model, AI Engineer (YouTube), June 30, 2026, at 02:06. https://www.youtube.com/watch?v=6MudaeKdBSk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): The speaker is a Reddit Android engineer speaking about engineers in general, not describing a Reddit policy. - Passage page: https://the-ai-record.com/passages/reddit-job-changed-harness-engineers-now-system-produce-verify-code-2026 > In the simplest of terms, I defined an … org as one where engineers leverage AI agents as their primary means of producing engineering outcomes. So that meant that our engineers needed to treat agents as core members of their engineering workflow. … using AI to help them write code here and there, but actually working with agents, decomposing problems, delegating work, reviewing and verifying what was done. And we wanted them directing the work as their default way of operating. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 2:29. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Jones is explaining how she defined an agent engineering organization after being asked to build one. - Passage page: https://the-ai-record.com/passages/block-defined-agent-engineering-org-agents-primary-means-2026 > Our engineers are orchestrating fully autonomous digital task forces, firing off agents and accomplishing incredible things. - Who: Google (Alphabet) (the organization sells AI tools) - Source: Google blog, April 22, 2026. https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/google-engineers-orchestrating-fully-autonomous-digital-task-forces-2026 > Our Network Firewall team audited rampartd using a multi-agent consensus process where every requirement was scored COMPLIANT, PARTIAL, or NON-COMPLIANT with specific violation details and remediation steps reducing what previously required weeks of manual work to a structured, repeatable process. - Who: Cloudflare (the organization sells AI tools) - Source: The AI engineering stack we built internally, on the platform we ship, Cloudflare blog, April 20, 2026, at Act 3, "Engineering Codex: engineering standards as agent skills". https://blog.cloudflare.com/internal-ai-engineering-stack/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/cloudflare-network-firewall-team-audited-rampartd-multi-agent-2026 > Today, when an engineer at Cloudflare opens a merge request, it gets an initial pass from a coordinated smörgåsbord of AI agents. Rather than relying on one model with a massive, generic prompt, we launch up to seven specialised reviewers covering security, performance, code quality, documentation, release management, and compliance with our internal Engineering Codex. These specialists are managed by a coordinator agent that deduplicates their findings, judges the actual severity of the issues, and posts a single structured review comment. - Who: Ryan Skidmore, Author of Cloudflare's "Orchestrating AI Code Review at scale" blog post, Cloudflare (the organization sells AI tools) - Source: Orchestrating AI Code Review at scale, Cloudflare blog, April 20, 2026, at Introduction. https://blog.cloudflare.com/ai-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): Describes the system as deployed on internal merge requests. - Passage page: https://the-ai-record.com/passages/cloudflare-up-seven-specialised-reviewers-coordinator-agent-2026 > The investigation that used to pull an engineer off their work for an hour now happens in the background. We review the PR, not the logs. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at End-to-End Bug Debugging. https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI fix bugs and handle incidents? - Passage page: https://the-ai-record.com/passages/cognition-investigation-used-pull-engineer-hour-background-2026 > The system peaked at ~1,000 commits per hour across 10M tool calls over a period of one week. - Who: Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cursor-system-peaked-1-000-commits-per-hour-2026 > Now, you asked specifically, do I write the code? I read … and what this says, I'm doing more and more and more, is asking the AI to explain back what it did to me, which is a little bit of the job I've compared it to being an engineering manager. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 12:17. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Do engineers still read the code agents write?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Rauch answers the host's question 'Are you still coding?' - Passage page: https://the-ai-record.com/passages/vercel-read-the-code-asks-ai-to-explain-back-engineering-manager-2026 > We're now seeing agentic code reviews that look at code changes through a series of different lenses. In the past, it would have been impractical for a cloud infrastructure engineer and a security engineer to review every single code change. With agents, that becomes possible. - Who: Venkat Venkataramani, VP of Engineering, Applied Infra, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 2. Death of the IDE & pull requests. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/openai-agentic-code-reviews-look-code-changes-series-2026 > If you submit a PR, no hard feelings, we’re just going to go and represent it for you as issues and discussions. And from there, trying to figure out the right way to bring people on. - Who: Fred K. Schott, Creator of Astro and Flue, Astro - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/astro-flue-submit-pr-no-hard-feelings-represent-2026 > We are definitely still bottlenecked on reviews, especially for things that are touching some architecture pieces. And it's actually more subtle than just being bottlenecked on review, … we can carve out time differently. It's … bottlenecked on human ability to even fully conceptualize what we're doing. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 10:13. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review?; What slows teams down once agents write the code? - Context (the archive's note, not their words): Answering a question from the host on whether code review is a bottleneck and whether code could be merged without a person looking at it. - Passage page: https://the-ai-record.com/passages/anthropic-labs-bottlenecked-on-review-human-understanding-2026 > We repurposed the word scenario to represent an end-to-end "user story", often stored outside the codebase (similar to a "holdout" set in model training), which could be intuitively understood and flexibly validated by an LLM. - Who: StrongDM - Source: How StrongDM's AI team build serious software without even looking at the code, Simon Willison's Weblog, February 7, 2026. https://simonwillison.net/2026/feb/7/software-factory/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/strongdm-repurposed-word-scenario-represent-end-end-2026 > But when they do have a copilot, I love seeing the generated code because I want to ask them, what do you think? Is this good code? Is this not good code? Are there problems? - Who: Farhan Thawar, Head of Engineering, Shopify - Source: How AI is changing software engineering at Shopify with Farhan Thawar, The Pragmatic Engineer, July 2, 2025, at Interesting detail: Shopify’s interview process. https://newsletter.pragmaticengineer.com/p/how-ai-is-changing-software-engineering - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Context (the archive's note, not their words): From a live podcast conversation at LDX3 in London between Gergely Orosz and Farhan Thawar, about how Shopify lets candidates use AI in its interviews. - Passage page: https://the-ai-record.com/passages/shopify-they-do-have-copilot-love-seeing-generated-code-2025 > …me five minutes before the live demo, which was actually … stressful because it was like, … production is down. And it's like, oh, well, you know, it's like, … want me to try and fix it? And I'm like, I don't think you're there yet, you know, a little dot, but, you know, thank you for trying. And then, you know, then, you know, I got in touch with like the engineering teams and we started to look, you know, what was going on. And, you know, they, it's like fixed by now. But having this thing that just understands like, okay, there's a pretty important thing happening. It's called Dev Day. There's like this production system, there's a live demo. It's probably using this production system. So these two things are connected and this is happening in five minutes. So I should ping him because like, he probably wants to know about it. I think that's quite remarkable. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 20:49. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: Can AI fix bugs and handle incidents? - Context (the archive's note, not their words): Asked for a recent AI use case that impressed him, Sottiaux describes his personal Dots agent warning him of a production outage shortly before the OpenAI DevDay live demo. - Passage page: https://the-ai-record.com/passages/openai-dot-warned-production-down-before-demo-2026 ## What slows teams down once agents write the code? Where the constraint moves when code gets cheap: review queues, CI capacity, decisions and other steps. Topic page: https://the-ai-record.com/topics/bottleneck-after-agents > Coding agents abruptly changed that reality. They can turn a ticket into a pull request in minutes, so teams end up getting greater numbers of PRs, much faster. Review didn't speed up with them and, as we well know, it became the bottleneck. - Who: Simon Maple, Head of Developer Relations, Tessl (the organization sells AI tools) - Source: Anatomy of an Agentic Code Review, Tessl blog, October 9, 2026, at Introduction. https://tessl.io/blog/anatomy-of-an-agentic-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/tessl-coding-agents-turn-ticket-into-pull-request-minutes-2026 > the amount of tests across our codebase grew 10x and we added a nominal amount of engineers. This all led to a 25x increase in CI jobs over a six month period - Who: Sachin Malhotra, Anthropic engineer (author), Anthropic (the organization sells AI tools) - Source: Agentic coding is straining CI. Here’s how we scaled test impact analysis at Anthropic, claude.com (Anthropic), September 14, 2026, at AI is evolving CI. https://claude.com/resources/articles/agentic-coding-is-straining-ci-heres-how-we-scaled-test-impact-analysis-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/anthropic-tests-grew-10x-ci-jobs-25x-six-months-2026 > Writing code is no longer the constraint, and once PR review gets accelerated, CI starts feeling the pressure. - Who: Sachin Malhotra, Anthropic engineer (author), Anthropic (the organization sells AI tools) - Source: Agentic coding is straining CI. Here’s how we scaled test impact analysis at Anthropic, claude.com (Anthropic), September 14, 2026, at AI is evolving CI. https://claude.com/resources/articles/agentic-coding-is-straining-ci-heres-how-we-scaled-test-impact-analysis-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/anthropic-writing-code-no-longer-constraint-ci-feels-pressure-2026 > Previously, writing code manually was the bottleneck. I find that with an Amazon, we've found the speed of decision making becomes a new bottleneck. … When it used to take nine to 12 months to build a new product, it didn't matter so much in the overall wash of things if it took two months to make the decision to build the product and then two months to approve the launch. But now, those are the bottlenecks. … Often, I find that frontier engineering teams spend more time making decisions than they do writing code. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 18:45. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Liguori is describing new organizational bottlenecks Amazon sees when coding is faster. - Passage page: https://the-ai-record.com/passages/amazon-decision-making-new-bottleneck-code-one-to-two-months-2026 > Back in 2024, we were seeing that engineers would get their first review within three hours. Now, in 2026, that has grown to nine hours, in addition to all of the volume changes. So in short, code review is now the bottleneck that we are running into. - Who: Will Bond, Staff Engineer, AI Foundations team, Development Platform organization, Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 01:05. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Does AI make developers more productive? - Context (the archive's note, not their words): Will Bond is describing a metric Uber tracks, the time to a first review of a pull request. - Passage page: https://the-ai-record.com/passages/uber-back-2024-we-were-seeing-engineers-2026 > One of the other things that we're really thinking about now is bottlenecks. We're now putting more strain on our infrastructure so we're trying to anticipate where our CI capacity needs to be and make the right foundational investments there. There's only so many experiments that we can feasibly run as well. So that's another bottleneck. And then lastly, too, … decision-making … It's not about … can we build? We know we can probably build it now. It's more of a question of should we build it? - Who: Adam Huda, Senior Engineering Leader, AI Developer Tools, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 17:29. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Huda closes the talk by naming the bottlenecks that remain once agents can write the code. - Passage page: https://the-ai-record.com/passages/uber-bottleneck-not-can-we-build-should-we-build-2026 > cortex hit exactly this. The moment we ran Claude Code at full throttle, writing speed jumped by an order of magnitude or more. Meanwhile the human time available to read and approve PRs only grew linearly. If the reviewer (=me) took a day off, the whole org stalled -- a classic single point of failure. - Who: Ryosuke Tsuji, CTO, airCloset - Source: Human-on-the-Loop: AI Reviewing AI PRs at cortex -- 769 PRs/month while raising the quality bar, DEV Community, May 26, 2026, at How the review bottleneck stops forming. https://dev.to/ryantsuji/human-on-the-loop-ai-reviewing-ai-prs-at-cortex-769-prsmonth-while-raising-the-quality-bar-4lh5 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/aircloset-moment-we-ran-claude-code-full-2026 > Code output per Anthropic engineer has grown 200% in the last year. Code review has become a bottleneck, and we hear the same from customers every week. - Who: Anthropic (the organization sells AI tools) - Source: Bringing Code Review to Claude Code, claude.com (Anthropic), March 9, 2026, at Managing the review bottleneck. https://claude.com/resources/articles/code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review? - Passage page: https://the-ai-record.com/passages/anthropic-code-output-per-engineer-grown-200-review-bottleneck-2026 > As we ship more code with agents, the bottleneck shifted from writing code to reviewing it. Devin Review turns large, complex GitHub PRs into intuitively organized diffs and precise explanations. We use it for every PR now. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at Automated Code Review. https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/cognition-bottleneck-shifted-writing-code-reviewing-devin-review-2026 > When we decided to rebuild our apps natively, we went back to the drawing board with a basic question: how should a mobile app be built when agents write most of the code? The biggest bottleneck we had was the feedback loop. An agent can write a screen in minutes, but proving that the screen works requires an emulator, a build, a tap script, and a lot of patience. - Who: Thiago Magalhaes, Author, Shopify Engineering blog, Shopify - Source: Reflex: an architecture for building mobile apps with agents, Shopify Engineering, November 5, 2026, at Introduction. https://shopify.engineering/reflex - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; How is AI used for testing? - Passage page: https://the-ai-record.com/passages/shopify-the-bottleneck-was-the-feedback-loop-2026 > My advice to engineering teams is, whether you build or buy, assume your architecture will be at a 25x load within two quarters. - Who: Sachin Malhotra, Anthropic engineer (author), Anthropic (the organization sells AI tools) - Source: Agentic coding is straining CI. Here’s how we scaled test impact analysis at Anthropic, claude.com (Anthropic), September 14, 2026, at What I would do differently. https://claude.com/resources/articles/agentic-coding-is-straining-ci-heres-how-we-scaled-test-impact-analysis-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/anthropic-assume-architecture-25x-load-within-two-quarters-2026 > Our team averages 10.8 PRs per engineer per week, 2.5x the 90th percentile according to DX benchmarks. But every one of those PRs still needed a human to look at it, and the review queue became the constraint. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at The problem: the review queue. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How is AI used in code review? - Passage page: https://the-ai-record.com/passages/ona-team-averages-10-8-prs-per-engineer-per-week-2026 > Right now, … the workflow for the developers has changed and is changing further. They're spending more and more time in planning and code review because there's so much more code being generated that they're being forced to do it. This probably isn't the favorite type of work that developers love doing, code review. - Who: Ty Smith, Principal Engineer, AI & Developer Productivity, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 18:38. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; Does AI make developers more productive?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Smith is describing how developers' time at Uber is shifting as agents write more code. - Passage page: https://the-ai-record.com/passages/uber-workflow-developers-changed-more-time-planning-code-review-2026 > At industrial scale, a flaky test or broken environment that a single developer can work around turns into a failure that interrupts every agent run. - Who: Michael Truell, Author of the Cursor blog post; CEO of Cursor per the lead record (the page gives no title), Cursor (Anysphere) (the organization sells AI tools) - Source: The third era of AI software development, Cursor (own site), February 26, 2026, at The shift is underway inside Cursor. https://cursor.com/blog/third-era - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/cursor-flaky-test-broken-environment-interrupts-every-2026 > year over year, the number of CI jobs also doubled or more than doubled. So we had to like also like beef up whole bunch of foundations in this space - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 58:38. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Speaking about readying Uber's systems for the influx of pull requests that agents produce. - Passage page: https://the-ai-record.com/passages/uber-ci-jobs-doubled-or-more-year-over-year-influx-of-prs-2026 > And … since the beginning of this year at Anthropic, we've seen an 8X increase in code output per engineer. And this is just unheard of in the industry. … And the thing that we did to get to the … 8X is we systematically improve a bottleneck at a time. … We throw [Claude] at it and [Claude] is the coding. The next bottleneck is the code review. … And so just like one step at a time, we unbottleneck every part of the process. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 14:09. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Context (the archive's note, not their words): Cherny is describing what he says Anthropic has seen in engineering output this year and how the company got there. - Passage page: https://the-ai-record.com/passages/anthropic-8x-increase-code-output-per-engineer-2026 > We can now take bigger swings that say, oh, reviews are a bottleneck, let's try to make high quality … reviews. Is toil work like cleaning up feature flags or cleaning up things or doing small changes of bottleneck? Let's give you that ability to get you out of that zone. And then measure directly. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 07:45. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in code review? - Context (the archive's note, not their words): Ramakrishnan is answering a host question on how Uber picks use cases; just before, he says Uber goes strictly by developer-satisfaction data when developers cannot choose. - Passage page: https://the-ai-record.com/passages/uber-reviews-bottleneck-agentic-reviews-toil-flags-measure-directly-2026 ## Can AI fix bugs and handle incidents? AI that triages and fixes bugs, responds to incidents, or takes on-call work. Topic page: https://the-ai-record.com/topics/bugs-and-incidents > In the past four weeks, OCA assisted with 575 of the 1,220 PRs merged to address incidents. - Who: Richard Marmorstein, Co-author of Ramp's OCA post (job title not stated), Ramp - Source: How we built OCA, our AI on-call assistant, Ramp engineering blog, September 9, 2026. https://builders.ramp.com/post/how-we-built-oca-our-ai-on-call-assistant - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/ramp-past-four-weeks-oca-assisted-575-1-2026 > We set up a !triage-bug playbook that captures how our engineers actually investigate: read the report, search for relevant code paths, check git history, then summarize findings with root cause and suggested fix. The automation fires when anyone adds the Bug label to a ticket—no assignment needed. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at Bug Triage Automation. https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cognition-bug-triage-playbook-fires-bug-label-linear-2026 > A combination of our AI investments and our broader focus on reliability are also driving tangible improvements in quality: incident rates after a production code change were down over 70% in the first quarter of 2026 compared to the first quarter of 2025, and down over 40% relative to the fourth quarter of 2025. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To Our Shareholders. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/block-incident-rates-after-production-change-down-70-percent-2026 > For bug fixes, River can reproduce the issue, record the broken behavior, implement the fix, and then repeat the same steps to show the before and after. - Who: Shopify - Source: We gave AI agents a mobile dev environment, Shopify Engineering, October 22, 2026, at It proves its work. https://shopify.engineering/remote-dev - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/shopify-bug-fixes-river-can-reproduce-issue-2026 > Some of these run overnight, automatically picking up bugs, implementing fixes, testing them, and creating PRs. So we can start the next morning with fixes already implemented and ready for review. - Who: Shopify - Source: We gave AI agents a mobile dev environment, Shopify Engineering, October 22, 2026, at Works in the background. https://shopify.engineering/remote-dev - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/shopify-some-run-overnight-automatically-picking-up-2026 > Testo, in the last 30 days, caught 425 bugs. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-testo-last-30-days-caught-425-bugs-2026 > To drive some long-term fixes, I started a long-running session in an internal version of Claude Tag dedicated to monitoring the service. Anytime the listener lag would get more than 50,000 jobs behind, Claude would ping me and resume our conversation on next steps. - Who: Sachin Malhotra, Anthropic engineer (author), Anthropic (the organization sells AI tools) - Source: Agentic coding is straining CI. Here’s how we scaled test impact analysis at Anthropic, claude.com (Anthropic), September 14, 2026, at The bumpy road to redesign: Patch 2: Sharding. https://claude.com/resources/articles/agentic-coding-is-straining-ci-heres-how-we-scaled-test-impact-analysis-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/anthropic-long-running-claude-tag-session-monitored-listener-lag-2026 > Right now, OCA is human-in-the-loop. - Who: Richard Marmorstein, Co-author of Ramp's OCA post (job title not stated), Ramp - Source: How we built OCA, our AI on-call assistant, Ramp engineering blog, September 9, 2026. https://builders.ramp.com/post/how-we-built-oca-our-ai-on-call-assistant - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-right-now-oca-human-loop-2026 > a growing share of sessions aren’t initiated by humans, but by automated managed agents handling code review, self-healing CI failures, completing E2E PRs with visual validation, triaging on-call alerts, debugging incoming bugs, and handling a variety of code maintenance tasks with human reviews/escalations. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Running a Software Factory Efficiently at Uber Scale, Uber engineering blog, August 27, 2026. https://www.uber.com/us/en/blog/efficient-software-factory/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in code review?; How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/uber-growing-share-sessions-arent-initiated-humans-automated-2026 > Alert investigation: automatic investigations of paging alerts in production - Who: Monzo - Source: Building Agent Chip: Monzo's in-house agentic tool, Monzo blog, August 13, 2026. https://monzo.com/blog/building-agent-chip - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/monzo-alert-investigation-automatic-investigations-paging-alerts-production-2026 > And we do this for like everything. So once you post something in Slack, we either have a Linear Slack agent look at it or we have a Cursor automation to … look at the message I was posted and triage it and … look for duplicates. Or … if it's determined to be easy, … start to implement the fix for it immediately. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 1:10:05. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Zakariasson is describing how Cursor's team handles messages posted in Slack, continuing an answer about connecting Linear to cloud agents. - Passage page: https://the-ai-record.com/passages/cursor-slack-post-triaged-duplicates-easy-fixes-started-2026 > So it's been really useful, even in our support channels, I would see like someone reports a bug, the on call ask minion, hey, like, … what is the issue and then minion will come back with, oh, this is the root cause, oh, this looks simple enough, let's just fix it right there and then we would like fix it, like there's no, then loop of like adding it to backlog, doing in the next, next thing, like you're able to fix it right there. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 53:00. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Context (the archive's note, not their words): Medisetty is describing Slack use of Minion, Uber's background coding agent. - Passage page: https://the-ai-record.com/passages/uber-on-call-asks-minion-root-cause-fix-it-right-there-2026 > agents spin up to then triage and run your on-call - Who: Ahmad Al-Dahle, Chief Technology Officer, Airbnb - Source: Latent Space (independent AI engineering publication), October 2, 2026. https://www.latent.space/p/airbnb - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/airbnb-agents-spin-up-then-triage-run-your-2026 > It’s totally shifted in the last six months … We can now solve these issues with these automations — handling triage, reproduction, getting the user to actually verify the fix that the bot is suggesting before we even look at it. - Who: Fred K. Schott, Creator of Astro and Flue, Astro - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/astro-totally-shifted-last-six-months-handling-triage-2026 > The SRE use case, internally we think about it as auto triage Because we just want every message that comes in, and that’s an alert, that’s a bug report, to have Devin just start triaging it before anything else. And we’ve leaned into this use case so much though that we’ve basically tried to make it so that you don’t ever have to leave Slack to interact with this. - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 01:03:24. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/cognition-sre-use-case-internally-think-auto-triage-2026 > At PostHog, for example, 63% of changed lines are in files that already exist, and fix is our largest commit type at 40%. - Who: Jina Yoon, Author, PostHog newsletter, PostHog (the organization sells AI tools) - Source: Can software factories actually work?, PostHog newsletter, August 11, 2026, at What software factories are missing. https://posthog.com/newsletter/software-factories - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/posthog-63-changed-lines-are-files-already-2026 > Over roughly six weeks, we’ve run thousands of scans, produced over 300 findings ranging from defense-in-depth improvements to resolved security incidents. - Who: Zack Deveau, Title not stated on Deveau's author page; the post is written from Shopify's Application Security team, Shopify - Source: Building an agentic harness that outlasts the model, Shopify engineering blog, July 29, 2026. https://shopify.engineering/building-an-agentic-harness-that-outlasts-the-model - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing?; How does AI help keep a product secure? - Passage page: https://the-ai-record.com/passages/shopify-over-roughly-six-weeks-weve-run-thousands-2026 > Early data shows that 70% of all our in house background agents, tasks which is Minions are toil. - Who: Abhishek Tibrewal, Senior Data Scientist, Platform Engineering, Uber - Source: Engineering Enablement podcast by DX (DX sells developer-productivity measurement software); recording of Uber's talk at DX Annual, June 22, 2026. https://getdx.com/podcast/ubers-journey-measuring-ai-impact-on-developer-productivity/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/uber-early-data-shows-70-all-house-background-2026 > We are now generating operation logs uh uh or daily summary of the platform - Who: Xiucheng Quek, Software engineer, Genentech - Source: Background Agents for Genomics and Cloud Ops, YouTube video of a talk or interview (speaker in notes), May 7, 2026. https://www.youtube.com/watch?v=GXDtw0EmoX0 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/genentech-now-generating-operation-logs-uh-uh-daily-2026 > And so Astra is just much, much more diligent about following directions than I am. And so I will run it through and I will say … keep doing all of these specifically and just like it will run for all of these scenarios and it will use the browser. And it's a dream. And then when it's done, it writes out back to the fig jam board of all of the things that it has done. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 19:49. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Context (the archive's note, not their words): Mechaber is describing how he uses a coding agent with a browser to bug bash a redesign of Figma's request modals. - Passage page: https://the-ai-record.com/passages/figma-codex-bug-bash-diligent-browser-writes-back-figjam-2026 > We are building an agent that "handholds" a change all the way to production — whether it's a code change or a change behind a feature flag. It observes the relevant monitoring graphs, but can also build its own dashboard to monitor important signals. More of our code changes are going to production with this kind of agent monitoring. - Who: Venkat Venkataramani, VP of Engineering, Applied Infra, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 2. Death of the IDE & pull requests. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/openai-building-agent-handholds-change-all-way-2026 > When we asked questions related to why Codex runs in a certain way, Codex decided to debug its own systems. It connected to logs, SSH'd into research dev boxes, and analyzed ML instabilities. - Who: Thibault Sottiaux, Head of Codex, OpenAI (the organization sells AI tools) - Source: How Codex is built, The Pragmatic Engineer, February 17, 2026, at 3. How Codex builds itself, Using Codex to debug Codex. https://newsletter.pragmaticengineer.com/p/how-codex-is-built - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/openai-asked-questions-why-codex-runs-certain-2026 ## How do organizations choose and switch AI models? How organizations pick among frontier, open and fine-tuned models, avoid lock-in to one provider, and test before switching. Topic page: https://the-ai-record.com/topics/choosing-models > You have labeled data. Examples teach distinctions and edge cases that instructions struggle to spell out. Hundreds of thousands of in-domain examples did what a 2,700 word prompt couldn’t. - Who: Airbnb - Source: Frontier quality at 1% of the cost: a production benchmark of LLM classifiers, Airbnb Tech blog, November 3, 2026, at When should you fine-tune?. https://airbnb.tech/ai-ml/frontier-quality-at-1-of-the-cost-a-production-benchmark-of-llm-classifiers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/airbnb-labeled-data-examples-teach-distinctions-prompt-2026 > Our platform is provider agnostic by definition. So if we come back to the diagram, the little blue … green agent box there can be very easily swapped for any model. We can dynamically choose which model best fits the purposes that we have. So we are not specifically tied to any provider by definition - Who: Felippe Felisola Caso, Monzo (conference affiliation, 2026), Monzo - Source: A Bank Built Its Own Coding Agent. It Took Off (talk, AI Engineer), AI Engineer (conference organizer) YouTube channel, October 20, 2026, at 13:46. https://www.youtube.com/watch?v=adfw3xJBx1o - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Felisola Caso is explaining why Monzo expects to keep its in-house agent platform as models change. - Passage page: https://the-ai-record.com/passages/monzo-platform-provider-agnostic-by-definition-agent-box-swapped-any-model-2026 > And since all the requests go through one place, we could also add like audit logging, session traces, and a whole bunch of infrastructure around this. Every model request at Uber goes through this. So there's no model request that doesn't go through this. So that gives us like one single entry point for across all the models. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 24:22. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations keep AI spending in check? - Context (the archive's note, not their words): 'This' is Uber's internal model gateway, which he has just been describing (one OpenAI-compatible or Anthropic-compatible API in front of models from all vendors, with an AI Guard safety layer). - Passage page: https://the-ai-record.com/passages/uber-every-model-request-goes-through-one-gateway-2026 ## How is AI used in code review? AI that reviews code changes, finds problems before merge, or picks the human reviewer. Topic page: https://the-ai-record.com/topics/code-review > Approximately 1 million automated code reviews completed by Citi’s Generative AI tools year-to-date, saving approximately 100,000 hours per week across Citi’s developer population - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended September 30, 2025, SEC EDGAR, November 6, 2025, at Management's Discussion and Analysis, Transformation progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000154/c-20250930.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/citigroup-approximately-1-million-automated-code-reviews-2025 > Our most successful example of this has been review buddy, our code review agent. It started with engineers on another team hooking up a GitHub webhook to Inspect's API with a review-oriented prompt. They could just focus on making the code review part great while we provided them all the infrastructure to actually run this agent. Now, review buddy runs on every pull request at Ramp, and we've replaced all of our third-party code review tools. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 06:52. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes?; Who builds an organization's agents: every team, or a central group? - Context (the archive's note, not their words): Bruggeman describes Review Buddy, a code review agent built on Inspect's API and platform. - Passage page: https://the-ai-record.com/passages/ramp-our-most-successful-example-review-buddy-code-review-2026 > every time there's a pull request, we have a swarm of agents. And they go and look at every file that's changed, every file that's related to the files that change to figure out if there's bugs. And it also spins up your code in a sandbox, … dependencies, spins up local host, clicks around to try to break things, mocks inputs, and everything else to figure out if the code is broken. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 00:27. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/greptile-swarm-agents-every-pull-request-sandbox-2026 > 93% of our PRs are now automatically handled by Review Buddy so that the highest quality engineers are spending time on the 7% of PRs that truly matter, - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026, at 09:34. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/ramp-93-prs-now-automatically-handled-review-buddy-2026 > When the reviewer's machine receives an event, the script spawns claude -p and walks through 9 dimensions (Graph / Architecture / Security / Test / Doc / Impact / Observability / AI-Antipattern / Recurrence) sequentially, then reads the verdict marker the AI emitted at the end and posts APPROVE or REQUEST_CHANGES via gh pr review. - Who: Ryosuke Tsuji, CTO, airCloset - Source: Human-on-the-Loop: AI Reviewing AI PRs at cortex -- 769 PRs/month while raising the quality bar, DEV Community, May 26, 2026, at How the auto-review system is wired. https://dev.to/ryantsuji/human-on-the-loop-ai-reviewing-ai-prs-at-cortex-769-prsmonth-while-raising-the-quality-bar-4lh5 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/aircloset-reviewer-machine-receives-event-script-spawns-2026 > Our PR review Agent doesn’t treat code review as a single task. It decomposes it into separate sub-jobs, each handled by an independent sub-Agent. One assesses the quality of the problem description. Another checks whether the diff actually aligns with the stated intent. Another reviews for safety concerns. Another checks for logical correctness. Another reviews against best practices and known anti-patterns. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at How our PR review Agent works. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/intercom-pr-review-agent-decomposes-review-sub-agents-2026 > offers the promise of accelerating feedback cycles (i.e., decreasing the PR cycle time by 30.8%), alleviating reviewer workload (i.e., reducing the number of human-written comments by 35.6%), and improving overall software quality (i.e., finding errors with actionable suggestions). - Who: Atlassian (the organization sells AI tools) - Source: RovoDev Code Reviewer: A Large-Scale Online Evaluation of LLM-based Code Review Automation at Atlassian, arXiv (Atlassian researchers; accepted at ICSE 2026, SEIP track), January 5, 2026, at Abstract. https://arxiv.org/abs/2601.01129 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/atlassian-rovodev-code-reviewer-decreasing-pr-cycle-time-2026 > Rovo Dev is an LLM-powered agent that can review pull requests and post actionable suggestions to accelerate merges and deployments. It targets issues developers most often act on, i.e., readability, bugs, and maintainability, and formats each suggestion so authors can apply the in-line changes by just one click. - Who: Atlassian (the organization sells AI tools) - Source: Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?, Inside Atlassian (Atlassian engineering blog), November 10, 2025, at What is Rovo Dev Code Reviews?. https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/atlassian-rovo-dev-llm-powered-agent-review-pull-requests-2025 > Per early experiments and data science studies, 5000 repositories onboarded to AI code reviewer observed 10 – 20% median PR completion time improvements. - Who: Sneha Tuli, Principal Product Manager, Microsoft (the organization sells AI tools) - Source: Enhancing Code Quality at Scale with AI-Powered Code Reviews, Engineering@Microsoft blog, July 14, 2025, at Impact on quality and velocity. https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/microsoft-5000-repositories-onboarded-ai-code-reviewer-2025 > Completed approximately 220,000 automated code reviews in the Generative AI developer tool, considerably increasing coding capacity - Who: Citigroup - Source: Citigroup Inc. Form 10-Q, quarter ended March 31, 2025, SEC EDGAR, May 8, 2025, at Management’s Discussion and Analysis, Transformation, Progress. https://www.sec.gov/Archives/edgar/data/831001/000083100125000086/c-20250331.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Listed under the 10-Q bullet "Significantly expanding adoption of Generative AI tools, increasing efficiency and productivity across Citi". - Passage page: https://the-ai-record.com/passages/citigroup-220-000-automated-code-reviews-coding-capacity-2025 > From June to September, merged pull requests to Claude Code's codebase more than doubled while the share that had to be reverted fell by more than 40 percent. We didn't remove review. But we did move most of the reading to machines, optimized for human time and attention, and invested heavily in making it fast and easy to undo mistakes. - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Introduction. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/anthropic-claude-code-merged-prs-doubled-reverts-fell-2026 > We think "did a human read the code?" is an imprecise proxy for "what could go wrong with this change, and which check—automated or human—is the right one to catch it?" - Who: Anthropic (the organization sells AI tools) - Source: How the Claude Code team does code review, claude.dev Blog (Anthropic), October 15, 2026, at Rethinking the purpose of code review. https://claude.dev/blog/solving-code-review-at-anthropic/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/anthropic-human-read-code-imprecise-proxy-2026 > We built an automated AI code review GitHub Action in a single day, and we reverted it before the day was over. The idea was simple: someone comments a trigger phrase on a pull request, and a workflow runs our deepest review and posts the result. It took seven small pull requests to get that far, and one revert pull request to undo them. - Who: TradersPost - Source: The AI Review Bot We Reverted, TradersPost blog, October 13, 2026, at Introduction. https://blog.traderspost.io/article/the-ai-review-bot-we-built-and-reverted - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: What AI efforts failed or were rolled back? - Passage page: https://the-ai-record.com/passages/traderspost-built-ai-code-review-action-single-day-2026 > Checks that return pass or fail belong in CI, where they are cheap and repeatable. Model judgment is reserved for what a gate cannot decide. - Who: TradersPost - Source: The AI Review Bot We Reverted, TradersPost blog, October 13, 2026, at Deterministic gates first. https://blog.traderspost.io/article/the-ai-review-bot-we-built-and-reverted - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/traderspost-checks-return-pass-fail-belong-ci-2026 > A verifier runs in seconds, costs almost nothing and gives the same answer every time. Anything a verifier can catch, review shouldn't have to, which leaves the lenses free to spend their attention on problems that genuinely need reasoning about. - Who: Simon Maple, Head of Developer Relations, Tessl (the organization sells AI tools) - Source: Anatomy of an Agentic Code Review, Tessl blog, October 9, 2026, at Beyond the code review: Tessl verifiers and security findings. https://tessl.io/blog/anatomy-of-an-agentic-code-review - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/tessl-verifier-runs-seconds-costs-almost-nothing-same-answer-2026 > And so I started tracking the number of iterations, the number of review rounds between when the pull request was opened and when it was merged. Sure enough, very little difference. … Once again, very little if any statistical difference between human-generated and AI-generated pull requests in terms of how many iterations before they're ready to merge. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 07:57. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/greptile-very-little-difference-review-rounds-human-ai-prs-2026 > It finds the right human reviewer to loop in. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/ramp-finds-right-human-reviewer-loop-2026 > One thing that we learned is that actually writing the skill was very easy, like teams just very quickly wrote a skill by asking Claude to write one, go over my previous PR reviews and write me a skill, but the hard part was how to run these skills at scale with consistent quality and low cost, and that required a lot of iterations, not only from the [uReview] team side, but also … for each team who was trying to write these rules. - Who: Ameya Ketkar, Speaker, AI Engineer World's Fair 2026 (job title not stated), Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 09:44. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Ameya Ketkar is describing how Uber's teams wrote their own code-review skills for uReview. - Passage page: https://the-ai-record.com/passages/uber-one-thing-we-learned-actually-writing-2026 > Over the last 30 days: 100% AI code reviewer coverage across all repos on our standard CI pipeline. - Who: Cloudflare (the organization sells AI tools) - Source: The AI engineering stack we built internally, on the platform we ship, Cloudflare blog, April 20, 2026. https://blog.cloudflare.com/internal-ai-engineering-stack/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cloudflare-over-last-30-days-100-ai-code-2026 > code reviewed over 50,000 pull requests authored by humans - Who: HubSpot (the organization sells AI tools) - Source: Cloud Coding Agents at HubSpot, HubSpot product blog, January 23, 2026. https://product.hubspot.com/blog/cloud-coding-agents-at-hubspot - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/hubspot-code-reviewed-over-50-000-pull-requests-2026 > For PRs that got AI approval, merge time dropped from 3.5 hours to 33 minutes. That's 6.4x faster. - Who: Joey Bates, Developer Enablement, Customer.io - Source: How we taught AI to approve pull requests (not just review them), Customer.io, November 10, 2025, at Results: Speed and cognitive relief. https://customer.io/learn/how-we-work/how-we-taught-ai-to-approve-pull-requests - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/customer-io-merge-time-dropped-3-5-hours-33-minutes-2025 > We have seen scenarios where PRs waited days and even weeks before getting merged, or where important feedback was missed. These pain points motivated us to try AI assistance in the review process. The goal was to let AI handle the repetitive or easily overlooked aspects of reviews, freeing human reviewers to focus on higher-level concerns. - Who: Sneha Tuli, Principal Product Manager, Microsoft (the organization sells AI tools) - Source: Enhancing Code Quality at Scale with AI-Powered Code Reviews, Engineering@Microsoft blog, July 14, 2025, at Solving the real problems in PR reviews with AI. https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/microsoft-goal-let-ai-handle-repetitive-2025 > Teams across the company are leveraging these customizations to perform specialized reviews, like identifying regressions based on historical crash patterns or ensuring flight and change gates are in place. - Who: Sneha Tuli, Principal Product Manager, Microsoft (the organization sells AI tools) - Source: Enhancing Code Quality at Scale with AI-Powered Code Reviews, Engineering@Microsoft blog, July 14, 2025, at Customization: tailoring review for your team. https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/microsoft-teams-customizations-specialized-reviews-historical-crash-2025 > So one of the reasons we built … artifacts that we shipped a couple of weeks ago was partially for that, which is you would send somebody a PR, and then they'd be like, I don't know, man, this is like 2000 lines of code. It looks like code to me. And what we started doing instead is sharing much more like, here's a … artifact. Here's the explanation. Here's the intention of the change. Here's the trade-offs that were made. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 10:30. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Context (the archive's note, not their words): Continuing his answer on review bottlenecks. The product name before 'artifacts' is rendered unclearly in the recording's transcripts, so the passage is cut around it. - Passage page: https://the-ai-record.com/passages/anthropic-pull-requests-sent-with-artifact-explaining-intent-tradeoffs-2026 > As of June 2025, 80% of our engineers and data scientists connected to their work software an AI copilot that can create and review code - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-june-2025-80-engineers-data-scientists-connected-2025 > So there didn't seem to be very big difference between the rate at which pull requests were reverted from people versus agents in my study. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 06:05. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/greptile-little-difference-revert-rate-people-versus-agents-2026 > So then I started tracking the counts of P0s, P1s, and P2s in all of these changes. And interestingly, once again, there was not that much of a difference. In fact, three out of the four agents that we tested performed better than humans in terms of the rate at which … were producing P0s. They were producing fewer P0s than humans were. Similarly, the case with P1s and P2s. Broadly speaking, human-generated PRs were about equal in quality to … PRs based on this data. - Who: Daksh Gupta, Co-founder, Greptile (the organization sells AI tools) - Source: AI-Generated Code Is Already Competing With Human Code, AI Engineer (YouTube), September 27, 2026, at 07:07. https://www.youtube.com/watch?v=474j-n1Ltxc - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/greptile-three-out-four-agents-fewer-p0s-than-humans-2026 > In August, people at Coval opened 2,578 pull requests and pushed roughly 3,400 further commits to them after they were first marked ready. - Who: Dana Dzik, Engineer, Coval (author of the Coval blog post on its code reviewer), Coval (the organization sells AI tools) - Source: Notes on Building a Code Reviewer, Coval blog, September 10, 2026, at Introduction. https://www.coval.ai/blog/how-we-built-our-code-reviewer/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/coval-august-people-opened-2-578-pull-2026 > Uber spent the time to invest in building an in-house solution due to some of the constraints that we have. One of those is we currently use [Phabricator] and have for a long time and are in the process of migrating to GitHub. Most of the solutions do not provide support for [Phabricator]. - Who: Will Bond, Staff Engineer, AI Foundations team, Development Platform organization, Uber - Source: Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026), AI Engineer (YouTube), August 28, 2026, at 01:22. https://www.youtube.com/watch?v=EL123UNokkI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/uber-uber-spent-time-invest-building-house-2026 > In particular, this paper aims at answering the practical question: how can we design a review-guided, context-aware, quality-checked code review comment generation without fine-tuning? - Who: Atlassian (the organization sells AI tools) - Source: RovoDev Code Reviewer: A Large-Scale Online Evaluation of LLM-based Code Review Automation at Atlassian, arXiv (Atlassian researchers; accepted at ICSE 2026, SEIP track), January 5, 2026, at Abstract. https://arxiv.org/abs/2601.01129 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/atlassian-rovodev-code-reviewer-enterprise-grade-llm-based-2026 > For open-source projects, it’s worth considering having your own agents and your own setup, and not necessarily trusting the community, because it can actually cut down your time to review - Who: Lars Grammel, Engineer, Vercel, Vercel (the organization sells AI tools) - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do coding agents open and merge changes? - Passage page: https://the-ai-record.com/passages/vercel-open-source-projects-worth-considering-own-agents-2026 ## How do coding agents open and merge changes? Agents that write pull requests on their own: how many, how they are reviewed, and how many merge. Topic page: https://the-ai-record.com/topics/coding-agents > Inspect has a million sessions, 75% of our PRs is actually built by Inspect, and 1,000 of those PRs over the last month was submitted by a non-engineer. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026, at 08:41. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can people outside engineering ship code with agents? - Passage page: https://the-ai-record.com/passages/ramp-inspect-million-sessions-75-prs-actually-built-2026 > More than 70% of pull requests are attributed to local or cloud agents. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Running a Software Factory Efficiently at Uber Scale, Uber engineering blog, August 27, 2026. https://www.uber.com/us/en/blog/efficient-software-factory/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-more-than-70-pull-requests-attributed-local-2026 > Between 25-35% of the PRs we merge on a weekly basis are now authored by `ai-sdk-factory` agents. - Who: Vercel (the organization sells AI tools) - Source: Building a software factory for AI SDK, Vercel blog, August 12, 2026. https://vercel.com/blog/building-a-software-factory-for-ai-sdk - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-between-25-35-prs-merge-weekly-basis-2026 > And today at Uber, one in nine of our PRs is fully production ready with zero human authoring from our background agents. - Who: Ty Smith, Principal Engineer, AI & Developer Productivity, Uber - Source: Engineering Enablement podcast by DX (DX sells developer-productivity measurement software); recording of Uber's talk at DX Annual, June 22, 2026. https://getdx.com/podcast/ubers-journey-measuring-ai-impact-on-developer-productivity/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/uber-today-uber-one-nine-prs-fully-production-2026 > Builderbot executes over 200,000 operations per day and merges approximately 1,500 pull requests per week, about 15% of all production code changes across Block. - Who: Block - Source: Block rolls out BuilderBot, a new suite of AI-native tools that changes the way we ship, Block (own site), June 17, 2026. https://block.xyz/inside/block-rolls-out-builderbot-a-new-suite-of-ai-native-tools-that-changes-the-way-we-ship - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/block-builderbot-executes-over-200-000-operations-per-2026 > It picks up tickets directly from Linear and Jira, creates the branch, writes the code, opens the pull request, watches CI, and iterates based on feedback. - Who: Block - Source: Block rolls out BuilderBot, a new suite of AI-native tools that changes the way we ship, Block (own site), June 17, 2026. https://block.xyz/inside/block-rolls-out-builderbot-a-new-suite-of-ai-native-tools-that-changes-the-way-we-ship - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/block-picks-up-tickets-directly-linear-jira-creates-2026 > About one in eight pull requests merged into our codebase last week was authored by River, reviewed by us. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Learning on the Shop floor, X post by Tobi Lutke, CEO, May 9, 2026. https://x.com/tobi/status/2053121182044451016 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-about-one-eight-pull-requests-merged-into-2026 > In the last 30 days, 5,938 Shopify employees worked with River across 4,450 different Slack channels. It opened 1,870 pull requests in the last week alone in our main monorepo. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify careers page reproducing an internal post from the CEO to all staff, May 2026, May 2026. https://www.shopify.com/careers/candidate-guide/shop-floor - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-last-30-days-5-938-shopify-employees-2026 > Last week, we merged 659 Devin PRs into our own codebase, up from 154 in our best week in 2025. - Who: Cognition (the organization sells AI tools) - Source: How Cognition Uses Devin to Build Devin, Cognition, February 27, 2026, at Intro. https://cognition.com/blog/how-cognition-uses-devin-to-build-devin - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/cognition-merged-659-devin-prs-last-week-2026 > Over 1,300 Stripe pull requests (up from 1,000 as of Part 1) merged each week are completely minion-produced - Who: Alistair Gray, Software engineer, Leverage team, Stripe - Source: Minions: Stripe’s one-shot, end-to-end coding agents—Part 2, Stripe developer blog, February 19, 2026. https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents-part-2 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-over-1-300-stripe-pull-requests-up-2026 > They pull review feedback, respond inline, push updates, and often squash and merge their own pull requests. - Who: OpenAI (the organization sells AI tools) - Source: Harness engineering: leveraging Codex in an agent-first world, OpenAI (own site), February 11, 2026. https://openai.com/index/harness-engineering/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-they-pull-review-feedback-respond-inline-push-2026 > Over a thousand pull requests merged each week at Stripe are completely minion-produced - Who: Alistair Gray, Software engineer, Leverage team, Stripe - Source: Minions: Stripe’s one-shot, end-to-end coding agents, Stripe developer blog, February 9, 2026. https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-over-thousand-pull-requests-merged-each-week-2026 > we have merged over 7,000 fully AI-generated pull requests - Who: HubSpot (the organization sells AI tools) - Source: Cloud Coding Agents at HubSpot, HubSpot product blog, January 23, 2026. https://product.hubspot.com/blog/cloud-coding-agents-at-hubspot - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/hubspot-merged-over-7-000-fully-ai-generated-2026 > ~30% of all pull requests merged to our frontend and backend repos are written by Inspect. - Who: Zach Bruggeman, Principal engineer; founding engineer of Inspect, Ramp - Source: Why We Built Our Own Background Agent, Ramp engineering blog, January 12, 2026. https://builders.ramp.com/post/why-we-built-our-background-agent - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-30-all-pull-requests-merged-frontend-backend-2026 > our agents have generated more than 1,500 pull requests that teams across Spotify have merged into our production codebase. - Who: Spotify - Source: 1,500+ PRs Later: Spotify’s Journey with Our Background Coding Agent (Honk, Part 1), Spotify engineering blog, November 6, 2025. https://engineering.atspotify.com/2025/11/spotifys-background-coding-agent-part-1 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/spotify-agents-generated-more-than-1-500-pull-2025 > it’s probably up to about 50% of pull requests in Shopify, which again, pull requests, every time you change your production system, you write a pull request, are created now not by engineers doing engineering work in the traditional sense - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: The Knowledge Project (podcast by Shane Parrish of Farnam Street, independent), transcript on fs.blog, September 10, 2026. https://fs.blog/knowledge-project-podcast/tobi-lutke-3/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-its-probably-up-about-50-pull-requests-2026 > one in three pull requests on GitHub now involves an agent - Who: Satya Nadella, Chairman and Chief Executive Officer, Microsoft (the organization sells AI tools) - Source: Microsoft investor relations, July 29, 2026. https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4.aspx - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/github-one-three-pull-requests-github-now-involves-2026 > As the work progressed, we moved from using AI to help with individual coding steps to delegating defined tasks—including implementation and testing—for engineers to review. - Who: Dropbox - Source: How AI-assisted development helped us build the new Dropbox experience, Dropbox engineering blog (dropbox.tech), November 5, 2026. https://dropbox.tech/machine-learning/how-aiassisted-development-helped-us-build-the-new-dropbox-experience - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/dropbox-moved-using-ai-help-individual-coding-steps-2026 > Inspect is Ramp's in-house team-based coding agent. It lives in Slack, GitHub, our own web interface in every Ramp Links Chrome, linear, and more. It works fully remotely, spawning a sandbox powered by modal with an entire Ramp development environment for each session. … spin up in under five seconds and are set up with all the dependencies and connections you need to let you prototype, build, verify, and ship your work. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 00:18. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Opening of Bruggeman's talk at Modal's Runtime 2026 conference, describing Inspect, Ramp's internal coding agent. - Passage page: https://the-ai-record.com/passages/ramp-inspect-ramps-house-team-based-coding-agent-2026 > GitHub labels serve as the state machine that connects the agents. Each agent watches for a specific label, does its work, and applies the next label: - Who: Warp (the organization sells AI tools) - Source: Set up your software factory, Warp docs, October 28, 2026, at How the loop works. https://docs.warp.dev/guides/agent-workflows/set-up-a-software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/warp-github-labels-serve-state-machine-connects-agents-2026 > warpdotdev/oz-for-oss is Warp's complete, deployable software factory for GitHub-hosted repositories. It includes all four agent roles, a Vercel webhook layer that handles GitHub events without GitHub Actions boilerplate, and self-improvement loops that propose skill updates based on maintainer feedback. - Who: Warp (the organization sells AI tools) - Source: Set up your software factory, Warp docs, October 28, 2026, at Reference implementation. https://docs.warp.dev/guides/agent-workflows/set-up-a-software-factory - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/warp-oz-for-oss-complete-deployable-software-factory-2026 > In May 2026, Anthropic averaged around 2,500 code changes merged per day; as of September, we merged over 30,000 per day. - Who: Anthropic (the organization sells AI tools) - Source: Anthropic (own site), October 13, 2026. https://www.anthropic.com/engineering/redesigning-source-control - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/anthropic-may-2026-anthropic-averaged-around-2-500-2026 > I aligned with three other teams for another week on that and then I just sent it to an agent to implement overnight and it came back, this is probably from two … not one, but it's basically six weeks of coding work … only one week. So that's where I get … 5X speed up if I include the review cycle at the end… - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 11:20. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/figma-plan-week-aligned-teams-agent-overnight-six-weeks-5x-2026 > And so it's not uncommon to spend a week writing a very detailed plan, making all the decisions, flushing it out, iterating, sending it out to teammates to review. And then only when it's read and you've … out all the decision, … to the agent, the agent will send it back to you when it's implemented … has been really successful also in accelerating and also really restoring some of the joy into the development process. - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 08:07. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/figma-week-writing-detailed-plan-agent-implements-restores-joy-2026 > Over the last year, all of the investments we made in agentic AI have led to more than seventy percent of our PRs now either by local or cloud agents. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory (talk, AI Engineer World's Fair 2026), AI Engineer (conference organizer) YouTube channel, August 21, 2026. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-over-last-year-all-investments-made-agentic-2026 > authoring ~10% of all merged PRs at Monzo - Who: Monzo - Source: Building Agent Chip: Monzo's in-house agentic tool, Monzo blog, August 13, 2026. https://monzo.com/blog/building-agent-chip - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/monzo-authoring-10-all-merged-prs-monzo-2026 > A human stays in control throughout the entire process, including merging every change. - Who: Vercel (the organization sells AI tools) - Source: Building a software factory for AI SDK, Vercel blog, August 12, 2026. https://vercel.com/blog/building-a-software-factory-for-ai-sdk - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-human-stays-control-throughout-entire-process-including-2026 > This is how it works today internally at Notion. Almost all of our polish and large feedback like this is actually coordinated through our software factories, both in terms of writing to the right teams and also having coding agents take the first step. - Who: Sarah Sachs, Leads engineering teams for AI, Notion, Notion (the organization sells AI tools) - Source: Notion's Token Town, AI Engineer (YouTube), July 23, 2026, at 22:12. https://www.youtube.com/watch?v=-I5W5QVAT8E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Sachs is showing an example of internal product feedback at Notion that she says a colleague agreed she could post. - Passage page: https://the-ai-record.com/passages/notion-polish-feedback-coordinated-software-factories-2026 > More than 70% of pull requests are attributed to local or cloud agents. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), July 7, 2026. https://x.com/praveenTweets/status/2074605343439810922 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-more-than-70-pull-requests-attributed-local-2026-2 > Engineers were also able to assign … tickets as well as GitHub issues to an agent and have the agent implement the work … So now the agents have become a part of the sprint. This blew everyone's minds. The first time we did this, the team ran out of work and had to pull in more tickets like twice, right? So of course, engineering managers, product teams, they love this flow. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 11:40. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Jones is describing what happened when Block engineers began assigning tickets to coding agents. - Passage page: https://the-ai-record.com/passages/block-agents-part-of-sprint-team-ran-out-of-work-2026 > accounting for roughly 1 in 12 pull requests at Dropbox today - Who: Dropbox - Source: Beyond code generation: rethinking engineering productivity in the age of AI agents, Dropbox engineering blog, May 28, 2026. https://dropbox.tech/culture/beyond-code-generation-rethinking-engineering-productivity-in-the-age-of-ai-agents - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/dropbox-accounting-roughly-1-12-pull-requests-dropbox-2026 > 3,536 River-coauthored pull requests merged in that window. - Who: Burke Libbey, Principal Engineer, Shopify - Source: Under the River, Shopify engineering blog, May 28, 2026. https://shopify.engineering/under-the-river - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-3-536-river-coauthored-pull-requests-merged-2026 > Merged 160+ PRs using the automated upgrader - Who: Mahib Hosain, Engineer, Developer Productivity team, Stripe - Source: Stripe developer blog, May 27, 2026. https://stripe.dev/blog/modern-java-at-stripe-language-upgrades-as-a-service - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-merged-160-prs-using-automated-upgrader-2026 > A very typical user interaction these days is some people discussing some problem they want to solve on Slack and then just @mentioning Honk — like, go solve this. - Who: Niklas Gustavsson, Chief Architect, Spotify - Source: Let’s Talk Agentic Development: Spotify x Anthropic Live, Spotify Engineering, April 6, 2026. https://engineering.atspotify.com/2026/4/anthropic-agentic-development - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/spotify-typical-user-interaction-discussing-problem-slack-mentioning-honk-2026 > 1,800 code changes per week are now written entirely by Uber's internal background coding agent - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), March 16, 2026. https://x.com/praveenTweets/status/2033627282418655711 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-1-800-code-changes-per-week-now-2026 > what we found is by really focusing on this project, building a custom agent on top of our internal … we were able to get a much higher quality type of unit test output. And so at this point, we're seeing about 5,000 tests generated and merged per month around the company from this. And almost a 3x rate of quality versus something that would be generated from your typical generic agent. - Who: Ty Smith, Principal Engineer, AI & Developer Productivity, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 24:45. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Context (the archive's note, not their words): Smith is describing Auto Cover, Uber's custom unit-test generation agent, and why Uber built it instead of using a general coding agent. - Passage page: https://the-ai-record.com/passages/uber-autocover-5-000-tests-generated-merged-per-month-3x-quality-2026 > So if you think about back in the good old days of 2022 and 2023, when GitHub co-pilot first came out, it was a pretty novel way of augmenting developments. … we saw about maybe a 10 to 15% bump in overall diff velocity. … but this by itself didn't push us in the direction that we've seen over the past, let's say, year, where the paradigm has shifted to peer programming, where you could hand off workloads that are running asynchronously. - Who: Anshu Chadha, Director of Engineering, Dev Platform, Uber, Uber - Source: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 03:46. https://www.youtube.com/watch?v=i1tZN41VKcE - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Chadha is describing what Uber measured when GitHub Copilot arrived and how the work has shifted since. - Passage page: https://the-ai-record.com/passages/uber-back-2022-2023-copilot-10-15-diff-velocity-peer-programming-2026 > roughly 1,500 pull requests have been opened and merged with a small team of just three engineers driving Codex. - Who: OpenAI (the organization sells AI tools) - Source: Harness engineering: leveraging Codex in an agent-first world, OpenAI (own site), February 11, 2026. https://openai.com/index/harness-engineering/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-roughly-1-500-pull-requests-been-opened-2026 > This translates to an average throughput of 3.5 PRs per engineer per day, and surprisingly the throughput has increased as the team has grown to now seven engineers. - Who: OpenAI (the organization sells AI tools) - Source: Harness engineering: leveraging Codex in an agent-first world, OpenAI (own site), February 11, 2026. https://openai.com/index/harness-engineering/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/openai-translates-average-throughput-3-5-prs-per-2026 > while they're human-reviewed, they contain no human-written code. - Who: Alistair Gray, Software engineer, Leverage team, Stripe - Source: Minions: Stripe’s one-shot, end-to-end coding agents, Stripe developer blog, February 9, 2026. https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/stripe-while-theyre-human-reviewed-they-contain-no-2026 > Codex reviews every internal pull request. - Who: OpenAI (the organization sells AI tools) - Source: OpenAI forum (OpenAI-run), December 4, 2025. https://forum.openai.com/home/blogs/how-openai-engineers-use-codex-to-tackle-big-projects-with-rigor-2025-12-04 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-codex-reviews-every-internal-pull-request-2025 > It gets assigned issues by human engineers, opens pull requests, and does the work assigned. - Who: GitHub (the organization sells AI tools) - Source: GitHub blog, November 12, 2025. https://github.blog/ai-and-ml/github-copilot/how-copilot-helps-build-the-github-platform/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/github-gets-assigned-issues-human-engineers-opens-pull-2025 > All the surrounding Fleet Management infrastructure — targeting repositories, opening pull requests, getting reviews, and merging into production — remains exactly the same. - Who: Spotify - Source: 1,500+ PRs Later: Spotify’s Journey with Our Background Coding Agent (Honk, Part 1), Spotify engineering blog, November 6, 2025. https://engineering.atspotify.com/2025/11/spotifys-background-coding-agent-part-1 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/spotify-all-surrounding-fleet-management-infrastructure-targeting-repositories-2025 > You just tell the AI to go nuts. It makes a pull request. You just say, “Yep, that’s good.” You don’t really read it, and now it has to be reviewed by your colleagues. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: The Knowledge Project (podcast by Shane Parrish of Farnam Street, independent), transcript on fs.blog, September 10, 2026. https://fs.blog/knowledge-project-podcast/tobi-lutke-3/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/shopify-you-just-tell-ai-go-nuts-makes-2026 > It just makes less sense to have people contributing code if the issue is decently well-specified and the code can be written by agents. - Who: Steve Ruiz, Founder, tldraw, tldraw - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/tldraw-makes-less-sense-people-contributing-code-2026 > about 30% of our PRs in that week came from Minions - Who: Will Gaybrick, President, Technology and Business (also styled President of Product and Business), Stripe - Source: Tokens Are the New Dollars | Stripe's Will Gaybrick & David George (podcast episode: "Stripe’s AI Strategy: Build More, Not Less"), The a16z Show (podcast and YouTube channel of Andreessen Horowitz, an investor in Stripe), August 17, 2026. https://www.youtube.com/watch?v=P5iICDVn5gc - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-about-30-prs-week-came-minions-2026 > last week, I think 7,000 PRs came from minions - Who: Will Gaybrick, President, Technology and Business (also styled President of Product and Business), Stripe - Source: Tokens Are the New Dollars | Stripe's Will Gaybrick & David George (podcast episode: "Stripe’s AI Strategy: Build More, Not Less"), The a16z Show (podcast and YouTube channel of Andreessen Horowitz, an investor in Stripe), August 17, 2026. https://www.youtube.com/watch?v=P5iICDVn5gc - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-last-week-i-think-7-000-prs-2026 > Can we give them a way to win back time by a stack ranking that work in a way where they can background some of the toil, right? So that when they get to the validation, 50% to 70% of the work required has been completed by a background agent. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 04:27. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Ramakrishnan is describing Minion, Uber's background-agent platform, and the engineers who get toil tickets, such as a vulnerability fix with a long validation loop, in the middle of product work. - Passage page: https://the-ai-record.com/passages/uber-minion-backgrounds-toil-50-70-percent-done-before-validation-2026 > I had large pieces of code that I was using, maybe some helpers and some other pieces, and I just turned them into a skill. We had that into our product. It was an onboarding from … AI agents. People have … Node.js, all the various things, then they have different tools for packaging, and it is impossible to actually code that. Like it will require a lot of coding. But if I just say a skill says, please first figure out what their package manager is, then figure out what their ecosystem is, and then do these steps together with the user, you know, it … solved a lot more problems that we could ever code. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Context Is the New Code, AI Engineer (YouTube), May 3, 2026, at 1:37. https://www.youtube.com/watch?v=bSG9wUYaHWU - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used in product and design?; How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Debois describes turning code helpers into a skill for the onboarding of AI agents into Tessl's product. - Passage page: https://the-ai-record.com/passages/tessl-turned-code-helpers-into-skill-onboarding-agents-2026 > We found it was better to not instruct for things the model knows how to do, only things it doesn't know (e.g. multi-agent collaboration) or that are specific to the relevant domain (e.g. how to run tests, your deploy pipeline). Treat the model like a brilliant new hire who knows engineering but not your specific codebase and processes. - Who: Wilson Lin, Researcher, Cursor (author of the Towards self-driving codebases post), Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026, at Optimizing prompts. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/cursor-treat-model-brilliant-new-hire-prompts-2026 > Our developer workforce is using IBM Bob, our AI-based software development system that automates the full software lifecycle, driving developer productivity and predictable enterprise costs. - Who: IBM (the organization sells AI tools) - Source: IBM Form 10-Q for the quarterly period ended September 30, 2026, SEC (Form 10-Q), October 22, 2026, at Management Discussion, Looking Forward. https://www.sec.gov/Archives/edgar/data/51143/000005114326000095/ibm-20260930.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/ibm-developer-workforce-using-ibm-bob-2026 > In early 2024, we began developing goose, which was the first agentic harness to enable foundation models to execute work across an enterprise. That work is now changing how Block runs. Velocity continues to increase. As of mid-April, production code changes per engineer are up over 2.5x compared to January. - Who: Block - Source: Block, Inc. Q1 2026 Shareholder Letter (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To Our Shareholders. https://www.sec.gov/Archives/edgar/data/1512673/000119312526212032/d132441dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/block-goose-first-agentic-harness-production-changes-per-engineer-2026 > With agents writing most of the code, a second native implementation costs far less than it used to. An abstraction layer still carries its full cost: it adds a maintenance surface, hides platform behavior you later need for performance work, and turns every platform upgrade into a two-step migration. Writing the same ViewModel twice is cheaper than maintaining a bridge between the two. - Who: Thiago Magalhaes, Author, Shopify Engineering blog, Shopify - Source: Reflex: an architecture for building mobile apps with agents, Shopify Engineering, November 5, 2026, at Share the test, not the code. https://shopify.engineering/reflex - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/shopify-second-native-implementation-costs-far-less-2026 > The idea is to combine rigorous simulations and evals, with swarms of AI agents to rewrite a piece of existing code in a memory-safe language. We want to see how quickly we can migrate it, make the code more compact, and create a better foundation for shipping new features. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Jeetu Patel's LinkedIn post about modernizing Cisco's code, LinkedIn, November 2, 2026. https://www.linkedin.com/posts/jeetupatel_lot-of-discussion-on-a-fortune-article-link-activity-7523171493376090112-C67s - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations? - Passage page: https://the-ai-record.com/passages/cisco-combine-rigorous-simulations-evals-swarms-ai-2026 > We have a dumb but effective approach. Every 30 minutes, we clone every repository, install all of its dependencies, and do any initial setup, and then take a file system snapshot. Then when a new inspect session is created, we start the sandbox from that snapshot, and the sandbox starts up in just a couple seconds from that exact file system state. This means when you send a prompt, you're not waiting on Git or npm or really anything, it just starts the sandbox and begins processing your prompt. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 03:23. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Bruggeman explains how Inspect sandboxes (run on Modal) start quickly from file system snapshots. - Passage page: https://the-ai-record.com/passages/ramp-dumb-effective-approach-every-30-minutes-clone-2026 > We have on-call assistant that begins investigating the moment an incident is declared. We have Testo, a QA-focused agent. Our data team built Ramp research, creating an entirely custom tool set for the agent that's built on top of Inspect's access to our data warehouses. Our product operations team built Voice of the Customer, which synthesizes feedback from all the sources we have access to and compiles reports for teams. These, along with 200 other custom agents, all run on the Inspect's platform. - Who: Zach Bruggeman, Principal engineer, Ramp, Ramp - Source: @Inspect fix this | Zach Bruggeman Principal Software Engineer at Ramp, Runtime 2026, Modal (YouTube), November 2, 2026, at 07:16. https://www.youtube.com/watch?v=VTu6wAKgiwY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; Who builds an organization's agents: every team, or a central group? - Context (the archive's note, not their words): Bruggeman lists agents that teams at Ramp built on the Inspect platform. - Passage page: https://the-ai-record.com/passages/ramp-on-call-assistant-begins-investigating-moment-incident-2026 > Coding agents can use that environment too, running sessions against the live system to validate their changes. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/ramp-coding-agents-can-use-environment-too-2026 > and 1,000 of those PRs over the last month was submitted by a non-engineer. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-1-000-those-prs-over-last-month-2026 > Total merged changes more than doubled year over year in August, from roughly 8,100 to 17,000. - Who: Spotify - Source: AI Changed How Spotify Builds. What We Learned (and Fixed) About Quality at Higher Velocity, Spotify engineering blog, September 16, 2026. https://engineering.atspotify.com/2026/9/ai-changed-how-spotify-builds-what-we-learned-and-fixed-about-quality-at-higher-velocity - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/spotify-total-merged-changes-more-than-doubled-year-2026 > And sometime last year, we knew, or I say we, but the bedrock team knew that they were going to need to build a new inference data plane. But they had estimated it at 30 people over 18 months. This is a big, big service, and it was … take time to build the new one, migrate customers over, migrate models over, and they decided to take a step back. They took six people and they built it in 76 days with … - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 2:35. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; Do organizations reorganize teams around AI? - Context (the archive's note, not their words): Liguori is describing the Bedrock team (Amazon's model hosting service) and its build of a new inference data plane. - Passage page: https://the-ai-record.com/passages/amazon-bedrock-30-people-18-months-six-people-76-days-2026 > So another experiment that I want to talk about is an experimental sprint that was done in the Prime Video Organization. They took a 10-day sprint and they … experiment where they put, again, six engineers in a room and they let them go wild with Kiro. They brought down the project delivery time estimate from what was going to be 90 weeks down to 24, based on all of the progress they had made in this 10-day sprint. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 4:05. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Context (the archive's note, not their words): Liguori is describing an experimental sprint in Amazon's Prime Video organization. - Passage page: https://the-ai-record.com/passages/amazon-prime-video-ten-day-sprint-90-weeks-down-to-24-2026 > the number of issues with a pull request attached by someone in an engineering, product, or design function has grown sevenfold - Who: Linear - Source: Linear (own site), August 26, 2026. https://linear.app/now/sharing-growth-with-the-people-building-linear - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/linear-number-issues-pull-request-attached-someone-engineering-2026 > At HumanLayer, we like to build out what we call golden patterns by hand before setting the agent loose. These are just like idiomatic hand-written examples for the agent to follow because they're just pattern replicators and otherwise you're getting what's in the docs or what the agent knows from the internet. - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 12:47. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Mistele describes how HumanLayer prepares the skill given to the agent that makes each change in its migration loop. - Passage page: https://the-ai-record.com/passages/humanlayer-golden-patterns-by-hand-before-agent-2026 > To illustrate that, I'm going to walk through a control loop that we use internally at HumanLayer. For our loop, we are incrementally migrating our RPC API to Effect. We adopted it for some of our race prone code. We like it, so we're adopting it across the rest of our code base. - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 9:09. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations?; Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Mistele introduces the worked example he uses for the rest of the talk: a daily agent loop at HumanLayer that migrates RPC procedures to the Effect library. - Passage page: https://the-ai-record.com/passages/humanlayer-migrating-rpc-api-to-effect-loop-2026 > if you're just getting started with AI, try to keep it around 100,000 tokens. For larger million context window, we … this up to … 200,000 tokens, but I've regularly tried to keep it under 60 for the hardest problems. I've regularly gone over 300K for things where I'm just kind of riffing with the agent, and I'm just … too lazy to compact it and move on and do a new one. - Who: Dex Horthy, CEO, HumanLayer, HumanLayer (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 28:36. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Horthy is answering the moderator's question on whether context rot and context engineering are solved now that context windows are larger. - Passage page: https://the-ai-record.com/passages/humanlayer-context-window-guideline-100000-tokens-under-60-hardest-problems-2026 > What I've seen is that when we started introducing … and loops and eventually more autonomous work within the whole organization, a new technical path opened. All of a sudden, we were helping the agent with tooling, building tooling for the agent and that kind of reignited some of the developers … kind of felt that it wasn't for them. Not all of a sudden, they were like, yes, we can do this, we have that knowledge, we're like somehow helping this even with a kind of programmatic way. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Coding Agents Don't Scale Themselves. Neither Do Your Teams. The Rise of Agent Enablement, AI Engineer World's Fair 2026 (YouTube), July 2, 2026, at 03:53. https://www.youtube.com/watch?v=zCJtYuqwm7E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations train people to use AI?; How do engineering and product jobs change when agents write the code? - Context (the archive's note, not their words): Debois, who says earlier in the talk that his observations include Tessl, is describing developers' reaction to prompting and spec-writing and then to building harnesses; the speaker does not say which team 'we' is. - Passage page: https://the-ai-record.com/passages/tessl-building-tooling-for-agent-reignited-developers-2026 > What I started seeing in some of the more advanced kind of teams is that … rituals of, Hey, we're doing a planning and we're doing a retro in a team that they weren't about like, Hey, we had issues with the code, but we're seeing we had issues with the system. So on the retro part is like, Hey, the agent went over and over hit this problem. Can we fix the system? That's something you'll learn in the retro. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Coding Agents Don't Scale Themselves. Neither Do Your Teams. The Rise of Agent Enablement, AI Engineer World's Fair 2026 (YouTube), July 2, 2026, at 07:07. https://www.youtube.com/watch?v=zCJtYuqwm7E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Context (the archive's note, not their words): Debois describes what he has seen in advanced teams adopting coding agents; he does not name the teams. - Passage page: https://the-ai-record.com/passages/tessl-retros-agent-hit-problem-fix-system-2026 > We have like 50 users and I don't want to scare people, but the PR throughput for these folks is like 3X higher again. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-like-50-users-i-dont-want-scare-2026 > People that were using agentic AI for four hours or more per day have seen PR output more than double from before AI. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Why do some teams gain far more from AI than others? - Passage page: https://the-ai-record.com/passages/airbnb-people-were-using-agentic-ai-four-hours-2026 > Our PR throughput inside of Airbnb is 65% higher than it was before we introduced agentic AI. - Who: Christopher Sanson, Product Lead, AI Developer Experience, Airbnb - Source: DX podcast page (DX sells developer-productivity measurement software) carrying a talk given at its DX Annual conference, June 22, 2026. https://getdx.com/podcast/beyond-the-cli-agentic-ai-for-async-workloads-and-non-developers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-pr-throughput-inside-airbnb-65-higher-than-2026 > On the Square side, we took a list of features sellers had been waiting on for months and our engineers shipped them in days. Builderbot handled the scaffolding and the repetitive work, and our engineers made the decisions that shaped the product. - Who: Brad Axen, Head of AI Capabilities, Block - Source: Block rolls out BuilderBot, a new suite of AI-native tools that changes the way we ship, Block (own site), June 17, 2026, at What it means for how we build. https://block.xyz/inside/block-rolls-out-builderbot-a-new-suite-of-ai-native-tools-that-changes-the-way-we-ship - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/block-square-features-sellers-waiting-months-shipped-days-2026 > we're seeing a 76% increase in pull request frequency - Who: Spotify - Source: Coding Is No Longer the Constraint: Scaling Developer Experience to Teams and Agents at Spotify, Spotify engineering blog, June 3, 2026. https://engineering.atspotify.com/2026/6/code-with-claude-coding-is-no-longer-the-constraint - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/spotify-were-seeing-76-increase-pull-request-frequency-2026 > In a recent 30 day period: 59,918 River sessions happened in 5,170 Slack channels, touching the work of 7,000+ people inside the company - Who: Burke Libbey, Principal Engineer, Shopify - Source: Under the River, Shopify engineering blog, May 28, 2026. https://shopify.engineering/under-the-river - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-recent-30-day-period-59-918-river-2026 > Mux is an internal multi-agent coding tool that started when one engineer built it to solve their own problem. The idea was minimal: each agent gets its own git worktree, its own branch, its own terminal. No conflicts. No stashing. - Who: Coinbase - Source: Coding Had a Concurrency Problem: How Mux Helped Solve It, Coinbase engineering blog (byline: Alessandro Rodi, Adam Wolf), May 11, 2026, at The Evidence: Mux. https://coinbase.com/blog/coding-had-a-concurrency-problem-how-mux-helped-solve-it - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/coinbase-mux-internal-multi-agent-coding-tool-started-2026 > the code tells the model what's in front of it, but it does not tell the model who owns the service, what the dependencies are, … what databases are being talked to, what conventions are that the teams considers non-negotiable … they reduce the amount of guessing the models have to do. And one thing we learned the hard way is that even with the decent context file, it is better than having none. And when the model stops guessing, the quality goes up everywhere else. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 16:24. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Explaining why Cloudflare invested in Backstage, its internal developer portal, and AGENTS.md files as context for coding agents. - Passage page: https://the-ai-record.com/passages/cloudflare-code-tells-model-what-front-guessing-2026 > But the way that I see it is they're just the engine. If you give the engine the wrong fuel, which is context, they're not going to perform. So and you can't do anything on the LLMs, at least not me, right? I'm just using the coding agent, I'm using whatever they give me. But I can optimize my context. And that's I think the message, doing this more in an engineered way than just copy and … things and hoping for the best. - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Context Is the New Code, AI Engineer (YouTube), May 3, 2026, at 23:45. https://www.youtube.com/watch?v=bSG9wUYaHWU - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations give agents the knowledge they need? - Context (the archive's note, not their words): Debois closes the talk by saying where he puts his effort when using a coding agent he did not build. - Passage page: https://the-ai-record.com/passages/tessl-llms-just-the-engine-context-the-fuel-optimize-2026 > Escalate to a human only when judgment is required - Who: OpenAI (the organization sells AI tools) - Source: Harness engineering: leveraging Codex in an agent-first world, OpenAI (own site), February 11, 2026. https://openai.com/index/harness-engineering/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-escalate-human-only-when-judgment-required-2026 > A final tip: probably the most important thing to get great results out of Claude Code -- give Claude a way to verify its work. If Claude has that feedback loop, it will 2-3x the quality of the final result. … Verification looks different for each domain. It might be as simple as running a bash command, or running a test suite, or testing the app in a browser or phone simulator. Make sure to invest in making this rock-solid. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: X post by Boris Cherny (the last post of a numbered thread of tips), X post by Boris Cherny, January 2, 2026. https://x.com/bcherny/status/2007179861115511237 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/anthropic-give-claude-way-to-verify-its-work-2026 > engineers using it ship about 70% more merged pull requests - Who: OpenAI (the organization sells AI tools) - Source: OpenAI forum (OpenAI-run), December 4, 2025. https://forum.openai.com/home/blogs/how-openai-engineers-use-codex-to-tackle-big-projects-with-rigor-2025-12-04 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/openai-engineers-using-ship-about-70-more-merged-2025 > roughly corroborates the 67% increase in merged pull requests - Who: Anthropic (the organization sells AI tools) - Source: How AI is transforming work at Anthropic, Anthropic (own site), December 2, 2025. https://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make developers more productive?; How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/anthropic-roughly-corroborates-67-increase-merged-pull-requests-2025 > the main success driver we saw was choosing the right related files (finding nearby files, good example files from the same project, filtering the dependencies for files that were relevant to the component, etc.), rather than getting the prompt engineering perfect. - Who: Charles Covey-Brandt, Author of the Airbnb engineering blog post on LLM test migration (role not stated on the page), Airbnb - Source: Accelerating Large-Scale Test Migration with LLMs, Airbnb Engineering & Data Science (own site), March 13, 2025, at 3. Increasing the Context. https://airbnb.tech/?p=366 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations?; How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/airbnb-main-success-driver-choosing-right-related-2025 > So we did this example with Mercedes. So when we started partnering with them, they asked us, how can you make our F1 team better? And I being sort of, you know, performance maniac, I said, well, we can make your website much faster because when people go to … buy your merch or want to get the latest results from the race or whatever, I want to make you faster, I want to make you agentic. And so that process would have taken quite a bit of time. But the first iteration of the migration of their old stack to the new one actually took five minutes. We used an agent to rewrite the code from the old platform to the new one. Five minutes. … I was like, this used to take months. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 80:35. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Also filed under: Can AI do large code migrations? - Context (the archive's note, not their words): Rauch is giving an example of a customer migration done by an agent. - Passage page: https://the-ai-record.com/passages/vercel-mercedes-first-iteration-migration-took-five-minutes-agent-rewrote-code-2026 > every engineer must be able to explain, even if an AI generated the PR, must be able to explain what they built. - Who: Ahmad Al-Dahle, Chief Technology Officer, Airbnb - Source: Latent Space (independent AI engineering publication), October 2, 2026. https://www.latent.space/p/airbnb - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do engineers still read the code agents write? - Passage page: https://the-ai-record.com/passages/airbnb-every-engineer-must-able-explain-even-if-2026 > it’s better to limit community contribution to the places it still matters: reporting, discussion, perspective, and care. - Who: Steve Ruiz, Founder, tldraw, tldraw - Source: PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors, Latent Space, September 1, 2026. https://latent.space/p/pr-not-welcome - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/tldraw-better-limit-community-contribution-places-matters-2026 > And the way we get there is with dynamic workflows. And this is essentially using [Claude] to orchestrate just very large teams of agents to do the most complex work. Like, large … migrations. … Like, we just migrated … from Zig to Rust. And this is like a JavaScript runtime. It's a lot of code. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 21:44. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: Can AI do large code migrations? - Context (the archive's note, not their words): Cherny is describing what he calls dynamic workflows, in which the model coordinates many agents. - Passage page: https://the-ai-record.com/passages/anthropic-dynamic-workflows-large-teams-agents-migrations-2026 > I was like, OK, I need to port this whole thing from Python to TypeScript. Like, as a-- you know, if I put on my, like, 2010s engineering hat, or even my early 2020s, like, that's a dumb idea. … But I was like, I think this is doable now. And I basically created this dynamic workflow set up. And over the weekend, … port the whole thing, like, verify it, double check it, … and then came back Monday to complete a workflow that was a ported version of that thing. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 4:38. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: Can AI do large code migrations? - Context (the archive's note, not their words): Describing a Labs project he moved from Python to TypeScript, in answer to a question on being more ambitious with prompts. - Passage page: https://the-ai-record.com/passages/anthropic-labs-port-python-to-typescript-over-weekend-2026 > So we have like some Slack channels internally, maybe a Slack channel dedicated to, a specific product like DeepWiki maybe. And you can imagine that, or you want a Devin that never stops, it’s just always awake, but it has this like memory dock that it can just maintain for itself about, okay, what are like the number one priorities of what we have to fix and prioritize? Who is responsible for some upcoming work? - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:33:01. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them?; How do organizations give agents the knowledge they need? - Passage page: https://the-ai-record.com/passages/cognition-devin-permanent-pm-memory-file-slack-channel-2026 > You could give your agent a Slack MCP and okay, it can post messages back to you on Slack. But we actually use Devin like a coworker in Slack, and that’s how it’s been built from the ground up. But to do that, you actually need to, support webhooks that come back, right? And then Devin has to respond in a natural way and then hopefully don’t spam your threads too much and annoy the people in your company. - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:26:46. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/cognition-use-devin-like-coworker-slack-webhooks-2026 > Codex is really built for multitasking. There's this understanding that most tasks will just get done to completion. People on our team have figured out what Codex is and isn't capable of. There is a tricky thing in all of this, though: we have to relearn these capabilities with every model - Who: Thibault Sottiaux, Head of Codex, OpenAI (the organization sells AI tools) - Source: How Codex is built, The Pragmatic Engineer, February 17, 2026, at 3. How Codex builds itself. https://newsletter.pragmaticengineer.com/p/how-codex-is-built - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI make developers more productive? - Passage page: https://the-ai-record.com/passages/openai-codex-really-built-multitasking-understanding-most-2026 ## How do organizations give agents the knowledge they need? Context files, skills, rules, internal knowledge graphs and tool connections that tell agents how the organization works. Topic page: https://the-ai-record.com/topics/context-for-agents > Engineers have built over 3,600 agent skills across the software development life cycle, and executed more than 30K agent skill executions per day. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Running a Software Factory Efficiently at Uber Scale, Uber engineering blog, August 27, 2026. https://www.uber.com/us/en/blog/efficient-software-factory/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations train people to use AI? - Passage page: https://the-ai-record.com/passages/uber-engineers-built-over-3-600-agent-skills-2026 > So we have one context graph, we took all of the information of how Uber runs into one context graph, this has 150 unique node and edge types, we have 40 million entries there right now, it captures all the way from how our mobile apps are built, to our backend, to our … all the design docs, Jira, incident bugs, everything is connected. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Agentic SDLC at Uber: Building Blocks for Uber's Software Factory, AI Engineer (YouTube), August 21, 2026, at 09:15. https://www.youtube.com/watch?v=17-YSUHo6Lk - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Medisetty is describing the single store of company context that Uber's agents read. - Passage page: https://the-ai-record.com/passages/uber-context-graph-150-node-edge-types-40-million-entries-2026 > We built an architecture.md file. Instead of using … MD, just pick something that every model can just understand. This file has to be incredibly small, and it can only have things that will not change for months or for years. In our case, it's the layers of the compiler. … So now we have standards so anyone can use whatever they want. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 2:10. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Gupta is describing how the team sets shared standards when engineers choose their own AI tools. - Passage page: https://the-ai-record.com/passages/boundary-architecture-md-small-stable-layers-compiler-2026 > So we built a company world … based on the entirety of our 25,000 repo …, right? And this was a machine readable view of every single service and how they all connect. And this allowed orchestrators and the agents that they delegated to to pull context in as needed and understand the landscape as they implement. And this enabled multiple agents to explore different parts of the system in parallel. - Who: Angie Jones, Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation, Block - Source: Building an Autonomous Engineering Org, AI Engineer (YouTube), June 28, 2026, at 15:23. https://www.youtube.com/watch?v=whue9_YquGA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Jones is describing the shared view of Block's code that its agent orchestrator draws on. - Passage page: https://the-ai-record.com/passages/block-machine-readable-view-every-service-orchestrators-2026 > Laminar adopts the concept of deferred tools: most of the tools are hidden behind a tool with a way for product teams to choose which tools they don’t want to defer. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-laminar-adopts-concept-deferred-tools-most-2026 > We’ve found that skills help to establish a more consistent baseline of performance for how well agents use tools. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-skills-help-establish-more-consistent-baseline-2026 > So far, 323 employees (PMs, CX agents, salespeople, marketers, analysts) have made 6,500+ edits to what Glass knows. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Ramp company blog (signed by CEO Eric Glyman), October 13, 2026. https://ramp.com/blog/introducing-glass - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/ramp-so-far-323-employees-pms-cx-agents-2026 > We tried to model some of that here where the harness has, of course, these core facts. It also has these dated logs of kind of your past conversation history centrally. And then also these kind of scratch pad notes … little things that you've been working on right now that are designed to be ephemeral, and they will go away, but they might be relevant just for the next 30 minutes or an hour. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 33:51. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Robinson has just compared the agent's memory to how a brain keeps relevant facts in mind, and is describing the three memory layers in the Grok Bot harness. - Passage page: https://the-ai-record.com/passages/spacex-grok-bot-memory-core-facts-logs-scratchpad-2026 > This single entry point encompasses more than 1,000 MCP servers across internal and third-party SaaS MCP - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber - Source: Running a Software Factory Efficiently at Uber Scale, Uber engineering blog, August 27, 2026. https://www.uber.com/us/en/blog/efficient-software-factory/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-single-entry-point-encompasses-more-than-1-2026 > I was playing with Notion and we used Notion a lot at Keycard, but I really just wanted all my Notion things available to me as Markdown files. And because … just made it easier for the agent to work with it instead of going through MCP - Who: Ian Livingstone, CEO and co-founder, Keycard, Keycard (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 40:42. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Livingstone is answering the moderator's question about how to control access to a memory store shared by many agents. - Passage page: https://the-ai-record.com/passages/keycard-notion-as-markdown-files-easier-for-agent-than-mcp-2026 > Something I've done for a long time, and it's very silly, is I run a model … without any skills or any markdown. … actually, I get rid of all my skills and all my markdown and everything when the new model is released. - Who: Geoffrey Huntley, Creator of the Ralph loop; Latent Patterns, Latent Patterns (the organization sells AI tools) - Source: The Great Loops Debate, AI Engineer (YouTube), July 17, 2026, at 32:35. https://www.youtube.com/watch?v=c35YoMdnI78 - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Huntley is answering the moderator's question on how he keeps a model from cheating when verifying its own work. - Passage page: https://the-ai-record.com/passages/latent-patterns-run-model-without-skills-or-markdown-new-release-2026 > When it comes to context management and context engineering, I think having a good set of actually fairly simplistic setup that is reproducible across engineers, with a good set of Claude MD setups, a good set of skills that really capture the essence of the role you're trying to do or the domain you're trying to operate in. I think that's really it and don't overthink it. - Who: Christian Ryan, Applied AI lead, Anthropic (the organization sells AI tools) - Source: Let’s Talk Agentic Development: Spotify x Anthropic Live, Spotify Engineering, April 6, 2026. https://engineering.atspotify.com/2026/4/anthropic-agentic-development - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations train people to use AI? - Passage page: https://the-ai-record.com/passages/anthropic-context-engineering-good-set-fairly-simplistic-setup-reproducible-2026 > We have our own internal agent that we call AirChat, which [includes] basically all the necessary MCP organizational context, - Who: Ahmad Al-Dahle, Chief Technology Officer, Airbnb - Source: Latent Space (independent AI engineering publication), October 2, 2026. https://www.latent.space/p/airbnb - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/airbnb-own-internal-agent-call-airchat-which-includes-2026 > I think like having to set up and fiddle with your loops and, you know, figuring that out is something that, you know, maybe people got excited about, but I don't think this is the way that it's going to work. I think the way that it's going to work is how we're positioning and what we're shipping with dots, where you have an incredibly smart agent that works 24/7, understands your goals, understands your preferences, learns from feedback. And it's like, … not perfect. I will learn very much from making this available to our pro users. But over time, you just want … system that learns from what you want to achieve. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 03:19. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Context (the archive's note, not their words): Asked about the shift from hand-built agent loops and graphs, Sottiaux describes the approach OpenAI took with its Dots personal agent. - Passage page: https://the-ai-record.com/passages/openai-dots-agent-learns-no-loops-to-configure-2026 > But, you know, seriously, like … research, long persistent, long horizon tasks, this is something that we've worked on, you know, for more than two years. … And it's like, you know, it's just memory works very, very well in … and so we've put a lot of research in that. And, you know, all of this coming together on top of like the Codex harness and also the ability to just like run 24/7 with, you know, good productivity. That's what we launched with dots. And so it's the work of like many, many months. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 15:14. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Context (the archive's note, not their words): Asked why Dots took so long to ship, Sottiaux describes the research OpenAI built it on. - Passage page: https://the-ai-record.com/passages/openai-dots-built-on-research-memory-codex-harness-2026 ## Does AI make developers more productive? Claims about output per engineer, speed of delivery and how they were measured. Topic page: https://the-ai-record.com/topics/developer-productivity > we are seeing doubling in the code output for engineers - Who: Balaji Krishnamurthy, Chief Financial Officer, Uber - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/UBER-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/uber-seeing-doubling-code-output-engineers-2026 > Productivity gains across our technology teams are among the most tangible results of our AI investment. Headcount grew 8% YoY in Q1’26 – a carryover effect of 2025 hiring – but productivity KPIs are growing 7-10x faster. Many of our most senior engineers that were previously spending most of their time reviewing code are now also building code because of the productivity gains enabled by AI tools. Rollbacks – code that is returned to its developer due to errors – are materially lower YoY. - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To our Shareholders, Artificial Intelligence. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How do engineering and product jobs change when agents write the code? - Passage page: https://the-ai-record.com/passages/mercado-libre-productivity-gains-across-our-technology-teams-2026 > And completely anecdotally based on my own experience, I've really only felt maybe 10 to 20% more productive with all of these phases that have come before. But now inside of Amazon, we've been running pilots with different teams across the company, and we've been seeing a median of 4.5x productivity improvement and sometimes more than 10x. - Who: Clare Liguori, Senior Principal Engineer, AWS, Amazon (the organization sells AI tools) - Source: From AI-Assisted to AI-Native: Building a Frontier Development Team, AI Engineer (YouTube), August 28, 2026, at 1:11. https://www.youtube.com/watch?v=pqlWNihgdjI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Liguori is describing how pilots with Amazon teams compare with her own earlier experience of AI coding assistants. - Passage page: https://the-ai-record.com/passages/amazon-median-4-5x-productivity-pilots-felt-10-to-20-2026 > The impact on developer velocity internally is clear: we’ve never seen a quarter-to-quarter increase in merge requests to this degree. As AI tooling adoption has grown the 4-week rolling average has climbed from ~5,600/week to over 8,700. - Who: Cloudflare (the organization sells AI tools) - Source: The AI engineering stack we built internally, on the platform we ship, Cloudflare blog, April 20, 2026. https://blog.cloudflare.com/internal-ai-engineering-stack/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/cloudflare-impact-developer-velocity-internally-clear-never-2026 > The team was roughly the same size throughout (39 unique contributors before, 43 after), but each developer went from merging 34 PRs per month to 96: a 2.9x increase in individual throughput with no headcount change. - Who: Ona (the organization sells AI tools) - Source: How auto-approving low-risk PRs with AI cut our lead time by 74%, Ona blog, April 9, 2026, at The results. https://ona.com/stories/auto-approving-low-risk-prs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/ona-each-developer-went-merging-34-prs-month-96-2026 > Our early 2025 study found the use of AI causes tasks to take 19% longer, with a confidence interval between +2% and +39%. For the subset of the original developers who participated in the later study, we now estimate a speedup of -18% with a confidence interval between -38% and +9%. Among newly-recruited developers the estimated speedup is -4%, with a confidence interval between -15% and +9%. - Who: METR - Source: We are Changing our Developer Productivity Experiment Design, METR, February 24, 2026, at Introduction. https://evals.alignment.org/blog/2026-02-24-uplift-update/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): METR is a nonprofit that studies AI agents; this post reports on its own randomized study of AI use by experienced open-source developers, not on METR's internal AI use. - Passage page: https://the-ai-record.com/passages/metr-early-2025-tasks-take-19-percent-longer-speedup-18-2026 > So here we took 46 teams that were using AI, and we matched them with 46 similar teams that were not using AI. And we measured their net productivity gains from AI quarterly. And the shaded area is the middle 50% of the data, and the dark blue line is the median, which is of July of this year, stands at about 10% for this cohort. I'd like to direct your attention to the fact that the discrepancy between the top performers and the bottom ones is increasing. There's a widening gap. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 02:03. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Denisov-Blanch is describing Stanford's comparison of teams using AI with matched teams that were not, as his study's first result. - Passage page: https://the-ai-record.com/passages/stanford-46-teams-ai-matched-46-without-median-about-10-2025 > For example, groceries took eight months, nine months, and airport pickups took about six weeks to develop. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/airbnb-example-groceries-took-eight-months-nine-months-2026 > reduced the time from concept to delivery by as much as 60% - Who: Airbnb - Source: SEC filing (8-K exhibit: the company's own shareholder letter), August 6, 2026. https://www.sec.gov/Archives/edgar/data/0001559720/000119312526337928/d70413dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-reduced-time-concept-delivery-much-60-2026 > we've increased the number of features and improvements we shipped this year by nearly 80%. - Who: Brian Chesky, Co-founder and Chief Executive Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/airbnb-weve-increased-number-features-improvements-shipped-year-2026 > Product Development expenses also scaled, falling from 9.3% of revenue in Q1’25 to 7.9% in Q1’26. This reflects the deceleration in headcount growth, and the increasing use of AI among our engineers. Gains in productivity mean that we intend to keep engineer headcount broadly stable in 2026 - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at Costs and expenses, Product Development. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/mercado-libre-gains-productivity-mean-we-intend-keep-2026 > We're seeing even stronger gains with power users of AI coding tools, whose output has increased 80% year over year. - Who: Meta - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), January 28, 2026. https://transcripts.platformaeronaut.com/transcripts/META-4Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do organizations measure AI's impact?; Why do some teams gain far more from AI than others? - Passage page: https://the-ai-record.com/passages/meta-were-seeing-even-stronger-gains-power-users-2026 > As we prepared for launch, engineers used AI-assisted development to deliver 22 accessibility features across Home, All Files, file previews, and Photos. - Who: Dropbox - Source: How AI-assisted development helped us build the new Dropbox experience, Dropbox engineering blog (dropbox.tech), November 5, 2026, at Making the new experience usable by more people. https://dropbox.tech/machine-learning/how-aiassisted-development-helped-us-build-the-new-dropbox-experience - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/dropbox-engineers-used-ai-assisted-development-deliver-22-2026 > 19,000 software developers have experienced efficiency gains of over 20% through real-time coding assistance - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), October 14, 2026. https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/bank-of-america-19-000-software-developers-experienced-efficiency-gains-2026 > This project took three weeks for a single engineer. A year ago it would have been closer to a quarter. - Who: Sachin Malhotra, Anthropic engineer (author), Anthropic (the organization sells AI tools) - Source: Agentic coding is straining CI. Here’s how we scaled test impact analysis at Anthropic, claude.com (Anthropic), September 14, 2026, at The redesign. https://claude.com/resources/articles/agentic-coding-is-straining-ci-heres-how-we-scaled-test-impact-analysis-at-anthropic - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/anthropic-redesign-took-three-weeks-one-engineer-year-ago-quarter-2026 > In the first 11 days of running the dependency workflow, the backlog of open issues fell by about 70%. - Who: Erin Son, Senior Engineer, Shopify - Source: How River takes security work from a fix to merge, Shopify engineering blog, September 2, 2026. https://shopify.engineering/river-vulnerability-remediation - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How does AI help keep a product secure? - Passage page: https://the-ai-record.com/passages/shopify-first-11-days-running-dependency-workflow-backlog-2026 > When we started building BAML, I didn't think it would be possible to build some of the software we did. And just yesterday, one of … engineers built a partial C compiler, purely in BAML. … Our processes have to evolve if we're going to ship at agent speed. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 19:47. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Gupta is describing what the team has been able to build in BAML, the programming language it makes. - Passage page: https://the-ai-record.com/passages/boundary-engineer-built-partial-c-compiler-processes-evolve-agent-speed-2026 > Organizations running this model report 3 to 5x productivity gains. Legacy backlogs that once ran to thousands of work days shrink to hundreds. Time to market compresses from quarters to weeks. - Who: BCG Platinion (the organization sells AI tools) - Source: The Agentic Software Factory: How AI Agents Are Delivering Software at 5x Speed (BCG Platinion blog; BCG Platinion sells agentic software factory pilots), BCG Platinion blog, July 3, 2026. https://www.bcgplatinion.com/blog/the-agentic-software-factory-blog - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations? - Passage page: https://the-ai-record.com/passages/bcg-platinion-organizations-running-model-report-3-5x-2026 > These numbers come with caveats. Mux users likely skew toward engineers who were already high-output and AI-forward, so we don't claim Mux is the sole cause. But the consistency across orgs and repos suggests this isn't a marginal effect. - Who: Coinbase - Source: Coding Had a Concurrency Problem: How Mux Helped Solve It, Coinbase engineering blog (byline: Alessandro Rodi, Adam Wolf), May 11, 2026, at The Evidence: Mux. https://coinbase.com/blog/coding-had-a-concurrency-problem-how-mux-helped-solve-it - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/coinbase-these-numbers-come-caveats-mux-users-2026 > we complete about 189 tickets for this sprint, which is about four times as much or two to three times more than what we usually accomplish - Who: Xiucheng Quek, Software engineer, Genentech - Source: Background Agents for Genomics and Cloud Ops, YouTube video of a talk or interview (Xiucheng Quek, software engineer (Genentech)), May 7, 2026. https://www.youtube.com/watch?v=GXDtw0EmoX0 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/genentech-complete-about-189-tickets-sprint-which-about-2026 > Last year, our CTO Darragh Curran set an explicit goal: double the productivity of our entire R&D organization within 12 months. Because the faster we can build and ship, the faster our customers get the capabilities they need. Nine months later, we did it. The results have been significant across the board, but the number that matters most for this post: downtime from breaking code changes dropped 35%, even as our deployments doubled. - Who: Intercom (renamed Fin in May 2026) (the organization sells AI tools) - Source: AI is approving our pull requests: Here's how we made it safe, Intercom Blog, April 21, 2026, at The 2x effort. https://www.intercom.com/blog/ai-is-approving-our-pull-requests-heres-how-we-made-it-safe - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/intercom-double-productivity-r-d-downtime-dropped-35-2026 > Normally, this sort of activity might take a team of 40 people about a year to carefully build. Instead, the Bedrock team spun up a separable group of six very skilled engineers who were excited about starting over and building on our agentic coding service (Kiro) - Who: Amazon (the organization sells AI tools) - Source: Amazon (own news site), April 2026. https://www.aboutamazon.com/news/company-news/amazon-ceo-andy-jassy-2025-letter-to-shareholders - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Passage page: https://the-ai-record.com/passages/amazon-normally-sort-activity-might-take-team-40-2026 > Based on conversations with study participants, we believe it is likely that developers are more sped up from AI tools now — in early 2026 — compared to our estimates from early 2025. However, because of the selection effects in our experiment, our data is only very weak evidence for the size of this increase. - Who: METR - Source: We are Changing our Developer Productivity Experiment Design, METR, February 24, 2026, at Introduction. https://evals.alignment.org/blog/2026-02-24-uplift-update/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): METR is a nonprofit that studies AI agents; this post reports on its own randomized study of AI use by experienced open-source developers, not on METR's internal AI use. - Passage page: https://the-ai-record.com/passages/metr-developers-more-sped-up-early-2026-weak-evidence-2026 > Unfortunately, given participant feedback and surveys, we believe that the data from our new experiment gives us an unreliable signal of the current productivity effect of AI tools. The primary reason is that we have observed a significant increase in developers choosing not to participate in the study because they do not wish to work without AI, which likely biases downwards our estimate of AI-assisted speedup. - Who: METR - Source: We are Changing our Developer Productivity Experiment Design, METR, February 24, 2026, at Introduction. https://evals.alignment.org/blog/2026-02-24-uplift-update/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): METR is a nonprofit that studies AI agents; this post reports on its own randomized study of AI use by experienced open-source developers, not on METR's internal AI use. - Passage page: https://the-ai-record.com/passages/metr-unreliable-signal-developers-choosing-not-participate-without-ai-2026 > And here, AI usage had two effects. Firstly, is that rework went up by 2.5 times, which is really bad, and effective output, which is kind of like a proxy for productivity or so, didn't really change. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 14:33. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Denisov-Blanch is describing results for one team at an unnamed large enterprise in his case study. - Passage page: https://the-ai-record.com/passages/stanford-ai-usage-two-effects-rework-up-2-5-times-effective-output-2025 > So we saw that PRs went up by 14%, but this is inconclusive, because more PRs doesn't mean better. We saw that code quality decreased by 9%, which is problematic. We saw that effective output didn't increase meaningfully. And then, we saw that rework increased by a lot. And so then, the question here is, what is the ROI of this AI adoption … It might be negative. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 14:48. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Denisov-Blanch is recapping the case study of one team at an unnamed large enterprise. - Passage page: https://the-ai-record.com/passages/stanford-prs-up-14-percent-code-quality-down-9-roi-might-negative-2025 > So like one, for example, one of the things we did is tagging all of our PRs to different categories, like on complexity level, and is it a feature PR bug fix maintenance testing documentation on different categories. And we could see year over year, how is that mix changing? Luckily, like at least like right now, you can see our percentage of feature PRs is going up, our percentage of maintenance PRs is going down. - Who: Uday Kiran Medisetty, Distinguished Engineer, Uber, Uber - Source: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 1:24:33. https://www.youtube.com/watch?v=MDUYSMJmi8A - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Answering the host's question how Uber thinks about engineering productivity now that agents write more code. He gives no figures for the shift. - Passage page: https://the-ai-record.com/passages/uber-feature-pr-share-up-maintenance-pr-share-down-2026 > it grew like 7X over, the last, I think it was, two months, three months, something like that. And then you see our engineering headcount growth. It’s, gone up by, 10% or something. - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:03:01. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Walden Yan was describing an internal graph of Cognition's merged pull requests, which he said had grown 7X; the passage is the continuation of that answer. - Passage page: https://the-ai-record.com/passages/cognition-merged-prs-grew-7x-engineering-headcount-2026 > And we came up with an environment cleanliness index. It's quite experimental. It's a composite score that looks at tests, … documentation, and at modularity and at code quality. … And so what you can see is that there is a point 40 R squared, meaning a pretty decent correlation around environment cleanliness and gains from AI, … productivity gains from using AI. And so the takeaway here is to invest in code-based hygiene, to unlock these AI productivity gains. - Who: Yegor Denisov-Blanch, Researcher, Stanford, Stanford - Source: How to Quantify AI ROI in Software Engineering (Stanford Study, 120k Devs), AI Engineer (YouTube), December 11, 2025, at 04:09. https://www.youtube.com/watch?v=JvosMkuNxF8 - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Denisov-Blanch is explaining a measure Stanford built to test whether a team's working environment affects its gains from AI. - Passage page: https://the-ai-record.com/passages/stanford-environment-cleanliness-index-composite-tests-modularity-2025 > at a minimum, it has allowed us to moderate the pace with which we are adding headcount there. - Who: Balaji Krishnamurthy, Chief Financial Officer, Uber - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 5, 2026. https://transcripts.platformaeronaut.com/transcripts/UBER-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/uber-minimum-allowed-us-moderate-pace-which-adding-2026 > Our bankers automate the research and presentation materials. Our developers code more efficiently, and all our teammates improve productivity, consistency, and client service - Who: Brian Moynihan, Chair and Chief Executive Officer, Bank of America - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15267023092982784 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-bankers-automate-research-presentation-materials-developers-code-2026 > Some of the technology is letting our people do more and be more productive. - Who: Goldman Sachs - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15226560906273792 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/goldman-sachs-some-technology-letting-people-do-more-more-2026 > It started with productivity and a more narrow component of the firm called engineering, but based on leveraging the firm's relationships with a number of different large language model providers - Who: Goldman Sachs - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15226560906273792 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/goldman-sachs-started-productivity-more-narrow-component-firm-called-2026 > The upside is that the suite has no I/O and no flakiness; the price is that fixtures get big. A single journey can carry a dozen GraphQL responses, which is tedious for a human to maintain but fine for an agent. - Who: Thiago Magalhaes, Author, Shopify Engineering blog, Shopify - Source: Reflex: an architecture for building mobile apps with agents, Shopify Engineering, November 5, 2026, at Our test trophy. https://shopify.engineering/reflex - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/shopify-fixtures-big-tedious-human-fine-agent-2026 > So for example, when … playwright … came out instead of having humans navigate the code now, the agent can explore the code. That was a big win unlock for productivity in a lot of our team. - Who: Eyal Blum, Software Engineer, Figma (the organization sells AI tools) - Source: How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage), AI Engineer (YouTube), August 28, 2026, at 05:20. https://www.youtube.com/watch?v=5Bn0xro2ol8 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/figma-playwright-agent-explore-code-big-win-productivity-2026 > And of course, once you're feeling confident in the loop, we're going to want to speed it up, right? I have 150 RPC procedures to migrate. If I do one at a time, it's going to take six months, which is way longer than I want to wait. - Who: Kyle Mistele, Co-founder, HumanLayer, HumanLayer (the organization sells AI tools) - Source: Loop Engineering from first principles, AI Engineer (YouTube), July 25, 2026, at 16:23. https://www.youtube.com/watch?v=xIt_mTQp6mY - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do large code migrations? - Context (the archive's note, not their words): Mistele is explaining why HumanLayer would raise the number of procedures its loop handles per run. - Passage page: https://the-ai-record.com/passages/humanlayer-150-rpc-procedures-six-months-one-at-a-time-2026 > A year ago, agents triggered less than 3% of the deployments on Vercel. Now, they trigger around 29%, and we expect half of all deployments to come from agents soon. - Who: Vercel (the organization sells AI tools) - Source: Vercel blog, June 17, 2026. https://vercel.com/blog/introducing-eve - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-year-ago-agents-triggered-less-than-3-2026 > AI changed the build-vs-buy calculus for internal tooling. Before AI, building something like Mux internally would have meant vendor evaluations, procurement, and months of negotiation. Now, one engineer with access to the right models and infrastructure can prototype in days what used to require a vendor relationship. - Who: Coinbase - Source: Coding Had a Concurrency Problem: How Mux Helped Solve It, Coinbase engineering blog (byline: Alessandro Rodi, Adam Wolf), May 11, 2026, at What About Open Source?. https://coinbase.com/blog/coding-had-a-concurrency-problem-how-mux-helped-solve-it - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Passage page: https://the-ai-record.com/passages/coinbase-ai-changed-build-vs-buy-calculus-internal-2026 > requests and tokens tell us like usage, like merge requests and pipelines tell us about platform activity, but they do not tell us what value … providing. … the more meaningful long-term question really is like what is the future velocity and outcomes that we can measure with less friction, getting changes out safely. - Who: Rajesh Bhatia, Senior Director of Engineering, Cloudflare, Cloudflare (the organization sells AI tools) - Source: From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026), Ona (YouTube), May 6, 2026, at 20:49. https://www.youtube.com/watch?v=MbLdrAZFQRs - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Context (the archive's note, not their words): Asked how Cloudflare thinks about measuring return on AI spend. - Passage page: https://the-ai-record.com/passages/cloudflare-requests-tokens-tell-usage-not-value-2026 > more than 12,000 engineers bankwide now have access to AI coding assistants, supporting efficiency and faster feature deployment. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-more-than-12-000-engineers-bankwide-now-2026 > AI and Automation have helped IBM unlock extreme productivity gains across the company since January 2023, and we are on track to reach USD 4.5B in savings by the end of 2025. - Who: IBM (the organization sells AI tools) - Source: IBM (own site), August 26, 2025. https://www.ibm.com/think/insights/enterprise-transformation-extreme-productivity-ai - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ibm-ai-automation-helped-ibm-unlock-extreme-productivity-2025 > uReview today analyzes over 90% of the weekly ~65,000 diffs (equivalent of pull requests) landed at Uber. - Who: Sonal Mahajan, Staff Engineer, Programming Systems team, Uber - Source: uReview: Scalable, Trustworthy GenAI for Code Review at Uber, Uber engineering blog, August 12, 2025. https://www.uber.com/us/en/blog/ureview/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-ureview-today-analyzes-over-90-weekly-65-2025 > we had one problem where, like, it wouldn't cite properly, and so it would be difficult to follow, and it was like, next model release solved, it went from, 70% to, like, 99. It's just like, we did nothing and, like, great. - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 10:30. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact? - Passage page: https://the-ai-record.com/passages/legora-next-model-release-citations-70-to-99-2026 > And so my hope is that on … we'll see maybe one day, the single-person trillion dollar company. And that's not to dismiss the power of teams, but I do think it's amazing to see how one person plus agents can now do the work. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 06:25. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Context (the archive's note, not their words): Rauch is describing what he hopes companies built on Vercel's platform will look like. - Passage page: https://the-ai-record.com/passages/vercel-my-hope-single-person-trillion-dollar-company-one-person-plus-agents-2026 > We're talking about roughly a 10x increase in load on some systems. At most companies, that kind of growth might happen over two or three years. At OpenAI, we see it in about six months. - Who: Venkat Venkataramani, VP of Engineering, Applied Infra, OpenAI (the organization sells AI tools) - Source: Inside OpenAI's agentic software factory, The Pragmatic Engineer, September 15, 2026, at 2. Death of the IDE & pull requests. https://newsletter.pragmaticengineer.com/p/openai-software-factory - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/openai-roughly-10x-increase-load-some-systems-2026 > He's orchestrating 16 agents and he's going a whole lot faster. - Who: Will Gaybrick, President, Technology and Business (also styled President of Product and Business), Stripe - Source: Tokens Are the New Dollars | Stripe's Will Gaybrick & David George (podcast episode: "Stripe’s AI Strategy: Build More, Not Less"), The a16z Show (podcast and YouTube channel of Andreessen Horowitz, an investor in Stripe), August 17, 2026. https://www.youtube.com/watch?v=P5iICDVn5gc - Grade C: An interview or press quote. Checked against the source. - Also filed under: How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/stripe-hes-orchestrating-16-agents-hes-going-whole-2026 > seller productivity has increased by 20% - Who: Will Gaybrick, President, Technology and Business (also styled President of Product and Business), Stripe - Source: Tokens Are the New Dollars | Stripe's Will Gaybrick & David George (podcast episode: "Stripe’s AI Strategy: Build More, Not Less"), The a16z Show (podcast and YouTube channel of Andreessen Horowitz, an investor in Stripe), August 17, 2026. https://www.youtube.com/watch?v=P5iICDVn5gc - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/stripe-seller-productivity-increased-20-2026 > It's now more than half. So more than 50% of deployments to Vercel are being made by coding agents, which is insane. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel (the organization sells AI tools) - Source: CEO Interviews (interview transcript aggregator), June 18, 2026. https://ceointerviews.ai/interview/1044131/ - Grade C: An interview or press quote. Not yet checked against the source. - Passage page: https://the-ai-record.com/passages/vercel-its-now-more-than-half-so-more-2026 ## How do engineering and product jobs change when agents write the code? What engineers, product managers and designers spend their time on when agents do the building. Topic page: https://the-ai-record.com/topics/engineer-role > Yeah, yeah, yeah, merging some code every once in a while. Technically, like a lot of code is getting written for me to do all sorts of kinds of like analysis, like, you know, understand like trends, understand the business, understand the next feature, you know, how well our previous launches are doing. A lot of code is written by Codex for that. You know, I'm obviously not writing it by hand anymore. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 01:56. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Asked whether he still codes or ships pull requests, Sottiaux describes how much of his analysis work is now done by code that Codex writes. - Passage page: https://the-ai-record.com/passages/openai-codex-writes-analysis-code-for-me-2026 > 85% of questions that are being asked to PMs now are fully answered with AI. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026, at 13:01. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/ramp-85-questions-being-asked-pms-now-fully-2026 > I archive roughly 10 channels a day, it's bananas. And so one of the things that we were really focused on was how do you help PMs know where to focus and what to do and like go run down all the things that are important. And so I run this skill, what's up like six times a day, it pulls all of my relevant slacks and projects together. It pulls all of the things that I like, who I owe stuff to, what decisions are outstanding. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 08:13. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in product and design?; Do organizations build their own AI tools or buy them? - Context (the archive's note, not their words): Mechaber is demonstrating one of the skills in Figma's PM OS, a shared repository of AI skills for product managers. - Passage page: https://the-ai-record.com/passages/figma-pm-os-whats-up-skill-slack-six-times-a-day-2026 > I tell my teams that I work with now that hiring is the least interesting way to solve every problem. We should almost never talk about hiring. - Who: Will Larson, CTO, Imprint, Imprint - Source: The revised rules of engineering management with Will Larson, The Pragmatic Engineer (newsletter), October 21, 2026, at Takeaway 6. https://newsletter.pragmaticengineer.com/p/the-revised-rules-of-engineering - Grade C: An interview or press quote. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/imprint-hiring-least-interesting-way-solve-every-problem-2026 > One of the worst parts of being an engineering manager in the 2010s was when you joined the ranks of engineering management, one of the first lessons you'd get was: 'you need to never make a technical decision again.' That's literally what I was told. - Who: Will Larson, CTO, Imprint, Imprint - Source: The revised rules of engineering management with Will Larson, The Pragmatic Engineer (newsletter), October 21, 2026, at Takeaway 10. https://newsletter.pragmaticengineer.com/p/the-revised-rules-of-engineering - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/imprint-never-make-technical-decision-again-engineering-manager-2026 ## What AI efforts failed or were rolled back? Tools turned off, experiments abandoned, policies reversed and results that came out negative. Topic page: https://the-ai-record.com/topics/failures-and-reversals > Locking also caused too much contention. 20 agents would slow to the throughput of 1-3 with most time spent waiting on locks. - Who: Wilson Lin, Researcher, Cursor (author of the Towards self-driving codebases post), Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026, at Self-coordination. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How many agents can one person run at once? - Passage page: https://the-ai-record.com/passages/cursor-twenty-agents-slow-throughput-one-three-locks-2026 ## How is AI used in finance, legal and operations? Back-office uses: accounting, risk, compliance, legal review and operations. Topic page: https://the-ai-record.com/topics/finance-legal-operations > The result is an agentic pipeline that automatically decisions more than 30% of our credit risk account reviews that were historically all manual. - Who: Ramp - Source: Agentic Credit Analysis, Ramp Builders blog, October 22, 2026, at Introduction. https://builders.ramp.com/post/credit-analysis-using-semantics - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-agentic-pipeline-automatically-decisions-more-30-2026 > I think that we're processing over 100,000 expenses a day that are being reviewed agentically. - Who: Eric Glyman, Co-founder and co-CEO, Ramp - Source: Cheeky Pint (podcast hosted by John Collison, Stripe co-founder); transcript on podscripts.co, February 17, 2026. https://podscripts.co/podcasts/cheeky-pint/ramp-founder-eric-glyman-on-the-many-ways-ai-is-changing-corporate-spending - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-i-think-were-processing-over-100-000-2026 > While we continue to utilize well-established ML techniques in our underwriting processes, we do not use generative AI for credit underwriting. - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How do banks and other regulated organizations use AI within the rules? - Passage page: https://the-ai-record.com/passages/klarna-while-continue-utilize-well-established-ml-techniques-2026 > have handheld devices, and they're using computer vision to map our inventory, to know what we have, to know exactly where it is - Who: Walmart - Source: earnings call transcript via MarketBeat (third-party transcript site), February 19, 2026. https://www.marketbeat.com/earnings/reports/2026-2-19-walmart-inc-stock/ - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/walmart-handheld-devices-theyre-using-computer-vision-map-2026 > Our legal teams use AI to expedite document review. - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: Can AI do legal work like contract review? - Passage page: https://the-ai-record.com/passages/klarna-legal-teams-use-ai-expedite-document-review-2025 > Manual review scales with headcount. Our credit agents scale with compute, so every dollar we save goes back into our customers. - Who: Ramp - Source: Agentic Credit Analysis, Ramp Builders blog, October 22, 2026, at Decision automation as a customer advantage. https://builders.ramp.com/post/credit-analysis-using-semantics - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI change headcount and hiring? - Passage page: https://the-ai-record.com/passages/ramp-manual-review-scales-headcount-credit-2026 > Since launching with our Legal team, we’ve seen an over 20% reduction in average contract review time and 91% accuracy in AI-generated decisions. - Who: Austin Greco, Staff Software Engineer, Michelangelo (Uber's AI platform), Uber - Source: Uber engineering blog, October 8, 2026. https://www.uber.com/us/en/blog/building-ubers-redlining-agent/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do legal work like contract review? - Passage page: https://the-ai-record.com/passages/uber-since-launching-legal-team-weve-seen-over-2026 > In finance and operations, Kai is helping Stripes analyze messy data, generate recurring digests, and turn fragmented context into usable artifacts. - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-finance-operations-kai-helping-stripes-analyze-messy-2026 > Capital allocation across 150 cities: 15 hours → 30 minutes. • Financial pacing reports: 2 days → 10 minutes. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), July 7, 2026. https://x.com/praveenTweets/status/2074605343439810922 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-capital-allocation-across-150-cities-15-hours-2026 > Klarna's lawyers are using ChatGPT Enterprise to create the first draft of common types of contract, massively reducing the time it takes to draw up a contract. - Who: Klarna - Source: Klarna press releases, May 14, 2024. https://www.klarna.com/international/press/90-of-klarna-staff-are-using-ai-daily-game-changer-for-productivity/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Can AI do legal work like contract review? - Passage page: https://the-ai-record.com/passages/klarna-klarnas-lawyers-using-chatgpt-enterprise-create-first-2024 > five agents now live supporting our colleagues in core processes - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), July 31, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/31072026/nwg-h1-2026-results.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-five-agents-now-live-supporting-colleagues-core-2026 > It offers a secure way to provide identity documents to NatWest digitally, reducing completion time from days to minutes. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-offers-secure-way-provide-identity-documents-natwest-2026 ## Does AI change headcount and hiring? What organizations say about hiring, team size and work AI does in place of staff. Topic page: https://the-ai-record.com/topics/headcount-and-hiring > Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Reflexive AI usage is now a baseline expectation at Shopify, X post by Tobi Lütke, CEO (CEO memo shared publicly), April 7, 2025. https://x.com/tobi/status/1909251946235437514 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do leaders push AI adoption? - Passage page: https://the-ai-record.com/passages/shopify-before-asking-more-headcount-resources-teams-must-2025 > we don't need to grow our headcount. at levels that we did in the past because we're getting so much more output and speed from our existing workforce - Who: Ellie Mertz, Chief Financial Officer, Airbnb - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), August 6, 2026. https://transcripts.platformaeronaut.com/transcripts/ABNB-2Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-dont-need-grow-headcount-levels-did-past-2026 > We have had discrete areas where we did reduce jobs by 30% or 40%, and most of those people offered jobs elsewhere. - Who: JPMorgan Chase - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15271033854174208 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/jpmorgan-chase-had-discrete-areas-where-did-reduce-jobs-2026 > our investment in AI tools and infrastructure is increasing. That will be offset by slower head count growth. - Who: Dara Khosrowshahi, Chief Executive Officer, Uber - Source: Uber Q1 2026 earnings call transcript (FactSet CallStreet corrected transcript, hosted on Uber's investor site), May 6, 2026. https://s23.q4cdn.com/407969754/files/doc_financials/2026/q1/Uber-Q1-26-Earnings-Call-Transcript.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-investment-ai-tools-infrastructure-increasing-will-offset-2026 > We shipped over 300. new products and features last year alone. We kept our flat headcount, which we're very proud of - Who: Harley Finkelstein, President, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-shipped-over-300-new-products-features-last-2026 > We've obviously been disciplined for three years now. We're, on any given year, we're in fact slightly down from the year before. I don't see that changing. - Who: Jeff Hoffmeister, Chief Financial Officer, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), May 5, 2026. https://transcripts.platformaeronaut.com/transcripts/SHOP-1Q26-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-weve-obviously-been-disciplined-three-years-now-2026 > A significantly smaller team, using the tools we're building, can do more and do it better. - Who: Block - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/1512673/000119312526076557/d108590dex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/block-significantly-smaller-team-using-tools-were-building-2026 > Revenue per employee reached $1.24 million, up 3.6 times since 2022 as headcount declined 49% - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 19, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000002/klarnaq425earningspressr.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-revenue-per-employee-reached-1-24-million-2026 > For over two years, total headcount has consistently been flattened down, both sequentially and year over year - Who: Jeff Hoffmeister, Chief Financial Officer, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), November 4, 2025. https://transcripts.platformaeronaut.com/transcripts/SHOP-3Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-over-two-years-total-headcount-consistently-been-2025 > Entry-level software engineer work that largely consisted of "here's the ticket, go implement it" can now probably be replaced by some form of AI or factory. - Who: Noah Blauensteiner, Author, Xebia articles, Xebia (the organization sells AI tools) - Source: Shining Some Light in Dark Agentic Software Factories, Xebia articles, August 12, 2026, at If agentic software factories are coding, where does that leave us humans?. https://xebia.com/articles/shining-some-light-in-dark-agentic-software-factories/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/xebia-entry-level-software-engineer-work-largely-consisted-2026 > I think what I hear from most companies is they say, first step is we give them an exercise and we want them to really go nuts on the … to solve this. … Now, after they pass this, you do a walkthrough and you actually say, please explain me what happened. Why is this a good idea? - Who: Patrick Debois, Product DevRel Lead, Tessl (the organization sells AI tools) - Source: Coding Agents Don't Scale Themselves. Neither Do Your Teams. The Rise of Agent Enablement, AI Engineer World's Fair 2026 (YouTube), July 2, 2026, at 16:09. https://www.youtube.com/watch?v=zCJtYuqwm7E - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): Debois relays what he hears from companies about interviewing engineers in the age of coding agents; it is his account of other companies' practice, not Tessl's. - Passage page: https://the-ai-record.com/passages/tessl-hiring-exercise-go-nuts-on-ai-walkthrough-2026 > While it will likely be the outcome in certain jobs, the goal is not fewer headcount; the goal is compounding performance - Who: JPMorgan Chase - Source: JPMorgan Chase (own site), 2026. https://www.jpmorganchase.com/ir/annual-report/2025/ar-ceo-letter-jennifer-piepszak - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/jpmorgan-chase-while-will-likely-outcome-certain-jobs-goal-2026 > I've made the decision to reduce our workforce by 8% at Zendesk. - Who: Tom Eggemeier, Chief Executive Officer, Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (CEO's note to employees, posted by the company), May 31, 2023. https://www.zendesk.com/newsroom/articles/zendesk-workforce-reduction/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-ive-made-decision-reduce-workforce-8-zendesk-2023 > We’re reducing the size of our team by around 14% - Who: Patrick Collison, CEO, Stripe - Source: Stripe newsroom (company post of the CEO's email), November 3, 2022. https://stripe.com/newsroom/news/ceo-patrick-collisons-email-to-stripe-employees - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-were-reducing-size-team-around-14-2022 > So, yes, it's a mix. We have engineers, and then we have lawyers, and there's one team. So there's, like, three engineers and, like, three lawyers right now, and then we have a pool of what we call legal engineers, which is people that have worked in big law for like three or four years. Typically, maybe they have a CS degree and like they're very tech forward - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 9:00. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/legora-three-engineers-three-lawyers-legal-engineers-pool-2026 > we're not saying, "Wow, we need fewer sellers." We're saying, "So, we need a lot more sellers." - Who: Will Gaybrick, President, Technology and Business (also styled President of Product and Business), Stripe - Source: Tokens Are the New Dollars | Stripe's Will Gaybrick & David George (podcast episode: "Stripe’s AI Strategy: Build More, Not Less"), The a16z Show (podcast and YouTube channel of Andreessen Horowitz, an investor in Stripe), August 17, 2026. https://www.youtube.com/watch?v=P5iICDVn5gc - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations measure AI's impact?; How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/stripe-were-not-saying-wow-need-fewer-sellers-2026 > It’s got 800 employees working here, fully distributed, far more agents than there are people. - Who: Wade Foster, Co-founder and Chief Executive Officer, Zapier - Source: Madrona (venture capital firm) podcast, April 2, 2026. https://www.madrona.com/zapier-has-more-ai-agents-than-employees-heres-how-that-happened/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/zapier-its-got-800-employees-working-here-fully-2026 > So we launched a lead qualification agent back in August of last year. We started that with about 20% of a single engineer working on that agent. And with a human in the loop over the course of six weeks, we're able to completely move … our team of 10 on that function down to one. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 01:33. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used in sales and marketing? - Context (the archive's note, not their words): Vercel's COO describes the first agent her go-to-market engineering team shipped, for sales development. - Passage page: https://the-ai-record.com/passages/vercel-lead-qualification-agent-20-percent-engineer-team-10-down-to-one-2026 > We now still have one person running that in the U.S. And then about 20% of a person doing all of Europe and all of APAC. And when this came out, people decried the end of the knowledge worker. But for us, really … what we did was to shift all of those people into more valuable roles. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 02:57. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: How is AI used in sales and marketing? - Context (the archive's note, not their words): Continues the account of the lead qualification agent, right after a sponsor break in the recording. 'That' is the lead qualification work. - Passage page: https://the-ai-record.com/passages/vercel-one-person-us-20-percent-europe-apac-shift-roles-2026 > 850 full-time agents - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-850-full-time-agents-2026 > From 2020 – 2022, our hiring outpaced our business realities. - Who: Tom Eggemeier, Chief Executive Officer, Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (CEO's note to employees, posted by the company), May 31, 2023. https://www.zendesk.com/newsroom/articles/zendesk-workforce-reduction/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-2020-2022-hiring-outpaced-business-realities-2023 > The only way for us to keep accelerating is to make sure we’re surrounded by people with a broad range of experience… Roughly one-third of my team is experienced, scaled operators who understand how the Cisco engine works. The second third are systems thinkers from the market who bring the outside-in perspective. And the final third is made of entrepreneurial founders, CEOs, and engineers who are AI-native who operate in founder mode. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Cisco chief found a 27-year-old ex-founder ‘buried away in the organization doing something unstrategic’—now he’s upgrading 98% of its code, Fortune (Orianna Rosa Royle), October 30, 2026. https://fortune.com/2026/10/30/cisco-president-jeetu-patel-gen-z-27-year-old-unlimited-ai-tokens-rewrite-legacy-code-entry-level-jobs/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Passage page: https://the-ai-record.com/passages/cisco-roughly-one-third-my-team-experienced-scaled-2026 > It’s like if someone is like an AI skeptic and … just kind of trying to lay low. They are never … be like a top performer in a company where everyone around them is just using like a … data center of … agents to accomplish their tasks. - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: Shopify CEO on why using AI isn’t optional, Sources with Alex Heath (YouTube), October 28, 2026, at 51:42. https://www.youtube.com/watch?v=r9SHzCW9zxo - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Lütke is talking about open source projects that have banned AI contributions and about a company being clear about who it is for and who it is not for. - Passage page: https://the-ai-record.com/passages/shopify-ai-skeptic-never-top-performer-data-center-agents-2026 > The reason for that is because I’ve seen the acceleration of AI, and I know we can ship all these things on the existing organization, - Who: Sebastian Siemiatkowski, Chief Executive Officer and co-founder, Klarna - Source: Fortune (business magazine), February 17, 2026. https://fortune.com/2026/02/17/klarnas-ceo-dario-amodei-ai-white-collar-workforce-shrink-2030 - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-reason-because-ive-seen-acceleration-ai-i-2026 > will be cutting almost 5pc of its staff worldwide. - Who: Elaine Burke, Former editor of Silicon Republic (until 2023), Zendesk (the organization sells AI tools) - Source: Silicon Republic (Irish technology news site), November 9, 2022. https://www.siliconrepublic.com/business/zendesk-jobs-cuts-ireland-dublin - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-will-cutting-almost-5pc-its-staff-worldwide-2022 > Yeah, I would say like, … skill that is trending … is typing fast. That is not that useful … Skills that are trending up is, you know, just great taste. Thinking about the user connecting to the audience that you're building for. A lot of ex-founders are incredibly successful right now, or like, you know, just like repeat founders. We're also finding this, like when we're hiring at OpenAI, like, you know, we have a lot of founders. I think we have more than 120 … founders currently at OpenAI. So it's like this mega startup. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 23:45. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): Asked which skills are trending up and down in what OpenAI looks for when hiring, Sottiaux answers. - Passage page: https://the-ai-record.com/passages/openai-hiring-typing-fast-down-taste-up-2026 ## Do organizations build their own AI tools or buy them? Internal agents and platforms versus bought tools, and why. Topic page: https://the-ai-record.com/topics/internal-tools > We wanted the app to be agentic-first, and the same state-and-actions surface the CLI uses is what Sidekick, Shopify's AI assistant, uses to take actions inside the app. There's no separate integration layer for it. Anything an agent can do from the terminal, an agent inside the app can do too. - Who: Thiago Magalhaes, Author, Shopify Engineering blog, Shopify - Source: Reflex: an architecture for building mobile apps with agents, Shopify Engineering, November 5, 2026, at A contract built on first principles. https://shopify.engineering/reflex - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/shopify-agentic-first-sidekick-same-surface-2026 > Instead, we test a production classifier that screens content for violations of Airbnb’s policy on pricing and payment transparency. The task requires recognizing subtle policy distinctions with high precision, since false positives prevent legitimate content from being published. The model is called thousands of times per minute, and because it gates publication in real time, a score must return within 200ms. - Who: Airbnb - Source: Frontier quality at 1% of the cost: a production benchmark of LLM classifiers, Airbnb Tech blog, November 3, 2026, at Case study: benchmarking a production classifier. https://airbnb.tech/ai-ml/frontier-quality-at-1-of-the-cost-a-production-benchmark-of-llm-classifiers/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/airbnb-production-classifier-gates-publication-200ms-2026 > … there's only one tool that communicates between the client and the server. And it's just send user. And the great thing about this is that it's actually very architecturally simple where the client is just texting the server in some ways. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 18:21. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Robinson is describing how the Grok Bot harness differs from other tools on the market. The quote starts mid-sentence, after a lead-in the two transcription passes heard differently. - Passage page: https://the-ai-record.com/passages/spacex-grok-bot-one-tool-client-server-2026 > So we use an open source tool called temporal, which is interesting to look into if you're curious, so that we don't have to rebuild a lot of this very complicated stuff from scratch. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 17:08. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Robinson has just explained that an agent task that fails midway should resume at the failed step, not replay earlier steps, and says Grok Bot uses Temporal for these durable workflows. - Passage page: https://the-ai-record.com/passages/spacex-grok-bot-temporal-durable-workflows-2026 > we have a very simple rule in our team. Code can be slop, writing cannot. … What this design doc tool does, it's a replacement for both Notion and GitHub, effectively for design docs. It allows versioning, commenting, all the stuff you want. … We built another tool on top of that. And this tool was a slack integration for that tool. Every time a design doc got updated, this channel got notifications. And what ended up happening is this channel became the most popular channel in our company really fast. - Who: Vaibhav Gupta, Co-founder and Chief Executive, Boundary (BoundaryML) (the organization sells AI tools) - Source: Fighting Slop with Slop, AI Engineer (YouTube), July 31, 2026, at 2:43. https://www.youtube.com/watch?v=AMiyLItEtLA - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in product and design? - Context (the archive's note, not their words): Gupta is describing how the team keeps design documents good while letting code be generated. - Passage page: https://the-ai-record.com/passages/boundary-design-docs-writing-cannot-be-slop-slack-channel-2026 > again, cloud agents also come from this … internal need of us just wanting to … run things asynchronously. And because of that … we have also … put a lot … of effort into making our own code base work really well in cloud agents. - Who: Eric Zakariasson, Engineer, Developer Experience and Product, Cursor, Cursor (Anysphere) (the organization sells AI tools) - Source: Building your own software factory, AI Engineer (YouTube), April 28, 2026, at 1:21:41. https://www.youtube.com/watch?v=rnDm57Py54A - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Context (the archive's note, not their words): Zakariasson is answering an audience question about cloud agents that go off on tangents, and says where Cursor's cloud agents came from. - Passage page: https://the-ai-record.com/passages/cursor-cloud-agents-came-from-internal-need-2026 > We recently rolled out the Employee Assistant, a personalized AI-powered agent to provide employees a single resource to get help and take action across the firm. - Who: JPMorgan Chase - Source: JPMorgan Chase (own site), 2026. https://www.jpmorganchase.com/ir/annual-report/2025/ar-ceo-letter-jennifer-piepszak - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/jpmorgan-chase-recently-rolled-out-employee-assistant-personalized-ai-2026 > And so I call it Auto XP. And right now one of our engineers is running it and has like five different experiments going that I have not taken a look at … and it found a huge win on its very first experiment. … there are three cursor automations that trigger based on stuff in Slack. Auto XP has its own channel where it can communicate with itself. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 23:39. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used in product and design? - Context (the archive's note, not their words): Mechaber is describing Auto XP, an automated agent setup at Figma for running growth experiments. - Passage page: https://the-ai-record.com/passages/figma-auto-xp-growth-pm-agent-huge-win-first-experiment-2026 > Yeah, so it's a really, really basic repo. And so it's got an agents.md, which just tells it how to behave. It's got a bunch of related skills that tell it when to fire and what to do. It's got a couple of command hooks. So things like on boot go find out if there's a new version of PMOS, because we wanted a good way to like keep our 60 to 70 PM kind of roughly in sync. - Who: Ezra Mechaber, Growth product manager, Figma, Figma (the organization sells AI tools) - Source: The AI chief of staff that this Figma PM uses all day (How I AI, hosted by Claire Vo), How I AI (YouTube), November 2, 2026, at 10:56. https://www.youtube.com/watch?v=ztt8irafLMQ - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): Mechaber is answering the host's question about what the PM OS repository contains. - Passage page: https://the-ai-record.com/passages/figma-pm-os-basic-repo-agents-md-skills-hooks-60-70-pms-2026 > There is also tools that refuse to be agentic that you should delete like the plague. So if something is not exposing its data and capability in a way that agents can consume, they're breaking this new loop that I just talked about that is so amazing. - Who: Guillermo Rauch, Founder and Chief Executive Officer, Vercel, Vercel (the organization sells AI tools) - Source: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch, Delphi (YouTube), October 29, 2026, at 43:20. https://www.youtube.com/watch?v=W9YfKZaK7KI - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do leaders push AI adoption? - Context (the archive's note, not their words): Rauch is describing how he decides which software to keep, in the same answer about what Vercel has replaced with coding agents. - Passage page: https://the-ai-record.com/passages/vercel-tools-that-refuse-to-be-agentic-you-should-delete-like-the-plague-2026 > Then there's Vertex. This is our customer service agent, … Vertex costs us $300 a month in infrastructure, plus 12,000 in tokens. It's $150k a year. It operates at significant scale, handling thousands of technical cases a week. And three engineers work on it, three. We started with an off-the-shelf solution, but weren't seeing enough results. We built something … in two months and had it in production. - Who: Jeanne DeWitt Grosser, Chief Operating Officer, Vercel, Vercel (the organization sells AI tools) - Source: Going from 10 to 1 SDRs. How We Deployed AI Across our Sales Org with Vercel's COO, SaaStr AI (YouTube); talk at SaaStr AI Annual 2026, November 4, 2026, at 16:21. https://www.youtube.com/watch?v=wXHLbr4W-8s - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Does AI customer service cut the cost of support? - Context (the archive's note, not their words): Vercel's COO describes Vertex, the internal customer service agent behind its help page, as an example of building rather than buying. - Passage page: https://the-ai-record.com/passages/vercel-vertex-300-month-three-engineers-off-the-shelf-2026 > We leverage our daily member interactions, using AI to regularly monitor and analyze member feedback, and provide data-driven insights to make real-time changes to our products and build new products to serve our members’ needs. Our member experience team tracks trends across contact rates, NPS scores, and app store reviews, and augments these with detailed member feedback calls, to identify insights into member pain points. We have also trained GPTs on member feedback to quickly iterate on project plans and product innovations. - Who: Chime - Source: Chime Financial, Inc. prospectus (Form 424B4), SEC EDGAR, June 12, 2025, at Business, Member Support. https://www.sec.gov/Archives/edgar/data/1795586/000162828025030855/chimefinancialinc-finalpro.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in product and design? - Passage page: https://the-ai-record.com/passages/chime-ai-monitor-member-feedback-trained-gpts-2025 > even when you have hundreds of sessions running in parallel - Who: Zach Bruggeman, Principal engineer; founding engineer of Inspect, Ramp - Source: Why We Built Our Own Background Agent, Ramp engineering blog, January 12, 2026. https://builders.ramp.com/post/why-we-built-our-background-agent - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-even-when-you-hundreds-sessions-running-parallel-2026 > in the latest version of the agent, I, it’s not in custom agents yet, but it will be like, like next week, a week after or so, um, there’s now like over a hundred tools. Just for all, all the crazy notion stuff. - Who: Simon Last, Co-founder and head of AI, Notion (the organization sells AI tools) - Source: Notion's Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future (Latent Space), Latent Space, April 15, 2026, at 00:54:57. https://www.latent.space/p/notion - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/notion-latest-agent-now-over-hundred-tools-2026 ## How do leaders push AI adoption? Mandates, memos, expectations and incentives from leadership, and how organizations bring skeptical staff along. Topic page: https://the-ai-record.com/topics/leadership-and-mandates > But really, for a lot of people, it's actually just really exciting to get to try all these tools to get to experiment. And so I actually see a lot of my role as creating space for people to try stuff out, to feel safe doing it. … And my other job is to just give people context, so to give them business context, product context, so that they can make better decisions. - Who: Boris Cherny, Creator and head of Claude Code, Anthropic, Anthropic (the organization sells AI tools) - Source: Agentic AI and the Future of Software Development: S3 E4 (AMD Advanced Insights, hosted by Mark Papermaster), AMD (YouTube), July 29, 2026, at 18:41. https://www.youtube.com/watch?v=eQ6tb7j3Z2U - Grade C: An interview or press quote. Checked against the source. - Also filed under: How do organizations train people to use AI? - Context (the archive's note, not their words): Cherny is answering the host's remark that learning cycles are vastly accelerated, about how leaders should take people with them. - Passage page: https://the-ai-record.com/passages/anthropic-leader-role-creating-space-to-experiment-2026 > This is not just a tech topic anymore—it's changing the economics of industries. It belongs on the agenda of every CEO. - Who: Axel Griewel, Managing Director, BCG Platinion, BCG Platinion (the organization sells AI tools) - Source: The Agentic Software Factory: How AI Agents Are Delivering Software at 5x Speed (BCG Platinion blog; BCG Platinion sells agentic software factory pilots), BCG Platinion blog, July 3, 2026. https://www.bcgplatinion.com/blog/the-agentic-software-factory-blog - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bcg-platinion-not-just-tech-topic-anymore-changing-2026 ## How do organizations measure AI's impact? The metrics organizations use to judge whether AI is working, and whether counts of usage, tokens and pull requests show value or only activity. Topic page: https://the-ai-record.com/topics/measuring-impact > When Account Executives use Kai, they produce 2x the sales activity, create 17% more opportunities, generate 26% more revenue opportunities, and close 39% more deals when compared to the same sellers in weeks they don't use it. - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/stripe-when-account-executives-use-kai-they-produce-2026 > Our proprietary AI and ML models have helped reduce our fraud loss rates, which declined by 29% between the year ended December 31, 2022 and the year ended March 31, 2025. - Who: Chime - Source: Chime Financial, Inc. prospectus (Form 424B4), SEC EDGAR, June 12, 2025, at Business, Risk and Fraud. https://www.sec.gov/Archives/edgar/data/1795586/000162828025030855/chimefinancialinc-finalpro.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How does AI help keep a product secure? - Passage page: https://the-ai-record.com/passages/chime-ai-ml-models-reduce-fraud-loss-rates-29-2025 > EricaAssist reduces average call times by nearly one minute per interaction - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Does AI make human support agents faster? - Passage page: https://the-ai-record.com/passages/bank-of-america-ericaassist-reduces-average-call-times-nearly-one-2026 > Employee engagement is measured through Amplifier, Klarna's AI-powered engagement survey. - Who: Klarna - Source: SEC filing (the company's own regulatory report), furnished on Form 6-K, May 27, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000162828026038366/annualreport2025.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-employee-engagement-measured-through-amplifier-klarnas-ai-2026 > Revenue per employee reached nearly $1.4 million, four times the 2022 level - Who: Klarna - Source: SEC filing (the company's own regulatory report), furnished on Form 6-K, May 14, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000162828026034877/exhibitno991pressrelease.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-revenue-per-employee-reached-nearly-1-4-2026 > Our average annual revenue per employee at period end has increased from approximately $344,000 in 2022 to approximately $1,240,000 in the year ended December 31, 2025. - Who: Klarna - Source: SEC filing (the company's own regulatory report), February 26, 2026. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/klar-20251231.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-average-annual-revenue-per-employee-period-end-2026 > Our average annual revenue per employee at period end has increased from approximately $344,000 in 2022 to approximately $972,000 in the last twelve months ended June 30, 2025. - Who: Klarna - Source: SEC filing (the company's own regulatory report), September 8, 2025. https://www.sec.gov/Archives/edgar/data/2003292/000200329225000036/klarnagroupplcf-1a4.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-average-annual-revenue-per-employee-period-end-2025 > Most people misunderstand this as tokenmaxxing. It is actually evalsmaxxing. We’re in the early stages here and I love that at Cisco we’re on the frontier and learning how to best shape the future of software engineering. - Who: Jeetu Patel, President and Chief Product Officer, Cisco (the organization sells AI tools) - Source: Jeetu Patel's LinkedIn post about modernizing Cisco's code, LinkedIn, November 2, 2026. https://www.linkedin.com/posts/jeetupatel_lot-of-discussion-on-a-fortune-article-link-activity-7523171493376090112-C67s - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/cisco-most-people-misunderstand-tokenmaxxing-evalsmaxxing-2026 ## Can AI do large code migrations? Agents moving large codebases between languages, frameworks or versions. Topic page: https://the-ai-record.com/topics/migrations > We'd originally estimated this would take 1.5 years of engineering time to do by hand, but — using a combination of frontier models and robust automation — we finished the entire migration in just 6 weeks. - Who: Charles Covey-Brandt, Author of the Airbnb engineering blog post on LLM test migration (role not stated on the page), Airbnb - Source: Accelerating Large-Scale Test Migration with LLMs, Airbnb Engineering & Data Science (own site), March 13, 2025. https://airbnb.tech/?p=366 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/airbnb-originally-estimated-1-5-years-engineering-time-2025 > Retiring an older file list had been deferred for two quarters because updating the automated tests required substantial engineering time. AI-assisted development gave the team a practical way to tackle that cleanup alongside its launch work. Engineers divided the cleanup into 53 tasks. Coding agents then updated or removed the affected tests using the same testing and review process, and the team finished all of those tasks in one week. - Who: Dropbox - Source: How AI-assisted development helped us build the new Dropbox experience, Dropbox engineering blog (dropbox.tech), November 5, 2026, at Making the new experience usable by more people. https://dropbox.tech/machine-learning/how-aiassisted-development-helped-us-build-the-new-dropbox-experience - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/dropbox-retiring-older-file-list-deferred-two-quarters-2026 > a recent Java migration across backend services completed in three days - Who: Spotify - Source: AI Changed How Spotify Builds. What We Learned (and Fixed) About Quality at Higher Velocity, Spotify engineering blog, September 16, 2026. https://engineering.atspotify.com/2026/9/ai-changed-how-spotify-builds-what-we-learned-and-fixed-about-quality-at-higher-velocity - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/spotify-recent-java-migration-across-backend-services-completed-2026 > when we ran our first bulk run, we successfully migrated 75% of our target files in just four hours. - Who: Charles Covey-Brandt, Author of the Airbnb engineering blog post on LLM test migration (role not stated on the page), Airbnb - Source: Accelerating Large-Scale Test Migration with LLMs, Airbnb Engineering & Data Science (own site), March 13, 2025, at 3. Increasing the Context. https://airbnb.tech/?p=366 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/airbnb-first-bulk-run-migrated-75-target-files-2025 > we found the most effective route to improve outcomes was simply brute force: retry steps multiple times until they passed or we reached a limit. - Who: Charles Covey-Brandt, Author of the Airbnb engineering blog post on LLM test migration (role not stated on the page), Airbnb - Source: Accelerating Large-Scale Test Migration with LLMs, Airbnb Engineering & Data Science (own site), March 13, 2025, at 2. Retry Loops & Dynamic Prompting. https://airbnb.tech/?p=366 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/airbnb-found-most-effective-route-improve-outcomes-2025 > we reached 97% of the 3.5K original Enzyme files - Who: Charles Covey-Brandt, Author of the Airbnb engineering blog post on LLM test migration (role not stated on the page), Airbnb - Source: Accelerating Large-Scale Test Migration with LLMs, Airbnb engineering blog, March 2025. https://airbnb.tech/?p=366 - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/airbnb-reached-97-3-5k-original-enzyme-files-2025 > Engineers reviewed requirements and plans before proceeding with implementation. - Who: Max Da Silva, VP of Engineering, Growth, Shopify - Source: Shopify engineering blog, September 10, 2026. https://shopify.engineering/shop-app-migration - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-engineers-reviewed-requirements-plans-before-proceeding-implementation-2026 > We’ve also internally done a big shift to make a lot of our core, components of code testable as purely local dev without needing to actually, integrate with, any live services for this reason. And honestly, the older the company, the more you have to change to shift in this direction. But you can use AI to help you perform this migration nowadays. - Who: Walden Yan, Co-founder and Chief Product Officer, Cognition (the organization sells AI tools) - Source: The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray, Latent Space, May 28, 2026, at 00:56:10. https://latent.space/p/cognition - Grade C: An interview or press quote. Checked against the source. - Also filed under: How is AI used for testing? - Passage page: https://the-ai-record.com/passages/cognition-big-shift-core-components-testable-purely-local-2026 > And I think the angle with Shepherd ultimately is we need to keep evergreen code. We want to keep our code healthy, … we want to give people the primitive to be able to do that. - Who: Nikhil Ramakrishnan, AI Foundations & Developer Experience, Uber - Source: Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs, Ona (YouTube), Background Agents Summit, May 7, 2026, at 10:08. https://www.youtube.com/watch?v=o7G5EHDqQ1g - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Ramakrishnan is answering a question about Shepherd, which the host describes as Uber's platform for large-scale code changes such as migrations and upgrades. The recording is captioned 'Shepard' on the session page; both passes heard 'Shepherd'. - Passage page: https://the-ai-record.com/passages/uber-shepherd-evergreen-code-primitive-large-scale-change-2026 ## Can people outside engineering ship code with agents? Product managers, designers, executives and other staff who change production code through agents, and the checks around it. Topic page: https://the-ai-record.com/topics/non-engineers-ship-code > With River and Remote Dev, a web developer can fix a bug, a designer can iterate on a UI, and a PM can test a feature, all through Slack. - Who: Shopify - Source: We gave AI agents a mobile dev environment, Shopify Engineering, October 22, 2026, at Anyone can now contribute to mobile. https://shopify.engineering/remote-dev - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-river-remote-dev-web-developer-fix-2026 ## How many agents can one person run at once? How many agents people run side by side, how they keep track of them, and what limits the number. Topic page: https://the-ai-record.com/topics/parallel-agents > We built agentic workflows to tackle mobile development tasks in parallel, but quickly hit the limits of our development machines. Just three or four workflows were enough to max out a laptop, with compilers, simulators, and emulators competing for resources. - Who: Shopify - Source: We gave AI agents a mobile dev environment, Shopify Engineering, October 22, 2026, at Introduction. https://shopify.engineering/remote-dev - Grade B: The organization's own post, paper or talk. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/shopify-three-four-workflows-enough-max-laptop-2026 > Workers pick up tasks and are solely responsible for driving them to completion. They're unaware of the larger system. They don't communicate with any other planners or workers. They work on their own copy of the repo, and when done, they write up a single handoff that the system submits to the planner that requested the task. - Who: Wilson Lin, Researcher, Cursor (author of the Towards self-driving codebases post), Cursor (Anysphere) (the organization sells AI tools) - Source: Towards self-driving codebases, Cursor (own site), February 5, 2026, at The final system design. https://cursor.com/blog/self-driving-codebases - Grade B: The organization's own post, paper or talk. Not yet checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Passage page: https://the-ai-record.com/passages/cursor-workers-own-copy-repo-single-handoff-planner-2026 ## How is AI used in product and design? AI in product management, research, prototyping and design. Topic page: https://the-ai-record.com/topics/product-and-design > We rolled out our first AI-powered search experience in our marketplace in Q1’26, shifting the architecture away from keywords and rebuilding it around LLMs. In Brazil and Mexico, the improvement in product relevance led to uplifts in conversion and click-through-rate for sponsored listings, both of which represent incremental revenue. These are early results, which we believe have the potential to transform how our customers search and discover products on our platform. - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To our Shareholders, Artificial Intelligence. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Also filed under: How is AI used in sales and marketing? - Passage page: https://the-ai-record.com/passages/mercado-libre-rolled-out-our-first-ai-powered-search-2026 > Scout is an internal voice of the customer system that indexes hundreds of millions of merchant feedback items, making them searchable within our tools. - Who: Harley Finkelstein, President, Shopify - Source: earnings call transcript via Platform Aeronaut (third-party transcript site), November 4, 2025. https://transcripts.platformaeronaut.com/transcripts/SHOP-3Q25-transcript - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-scout-internal-voice-customer-system-indexes-hundreds-2025 > One of the principles that we had at Cursor and now at SpaceX is to delete the product. And what we mean by this is you kind of have to internalize and assume that the models today are the worst they'll ever be, which means that in six months you might need to completely rebuild the UI. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 37:45. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Robinson is describing the product principle his teams followed at Cursor and now at SpaceX while building harnesses for always-on agents. - Passage page: https://the-ai-record.com/passages/spacex-delete-the-product-models-worst-theyll-ever-be-2026 > We built an actual customer's insight agent that agent pulls from all the data sources at a company. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-built-actual-customers-insight-agent-agent-pulls-2026 > But especially in a labs-type group where our whole cadence is two-week reviews where every project goes up for, we call it persevere or pivot. So basically, every project is up for review, and either it's time to keep going, … or it's time to pivot it, or even shut down. And we shut down projects basically every single one of those cycles. And the more you do it, the less it's just like, oh, no, my project is shut down. I failed. It's like, no, that is definitely the intention of the lab team is to prototype quickly, try to ship internally, maybe get it to early access, and if it doesn't work, wind it down. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 12:11. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Context (the archive's note, not their words): Answering how Anthropic Labs is structured. - Passage page: https://the-ai-record.com/passages/anthropic-labs-shuts-down-projects-every-two-week-cycle-2026 > if you align the org chart too much to the individual projects, you're going to end up re-orging every two weeks, which would be a total nightmare. … And there is the concept of a bet lead or a directly responsible individual. But the interesting thing is that they don't manage usually any of the other people, which … breaks the … previous way in which a lot of these things were done. But I think it leads us to be really flexible when you say, OK, actually, this project is not going to work out. Let's disband and keep going, and it's not a big deal. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 12:46. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Context (the archive's note, not their words): Continuing his answer on how Anthropic Labs is structured; 'bets' are the project teams inside Labs. - Passage page: https://the-ai-record.com/passages/anthropic-labs-bet-leads-manage-nobody-2026 > We have one of our Slack challenges, like Project Unship, which is like, what is in the product right now? … Like, we unshipped styles, I think, recently, … used by a small percentage of people … very, sort of, prescriptive in the way that it worked, and skills were a much better application, something like that. - Who: Mike Krieger, Head of Anthropic Labs, Anthropic (the organization sells AI tools) - Source: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 15:43. https://www.youtube.com/watch?v=qqrk7CtkuIw - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Answering a question on what Anthropic would delete from its products. - Passage page: https://the-ai-record.com/passages/anthropic-labs-unship-styles-project-unship-2026 > I don't know if that's surprising, but it is, you know, as I said, a lot of ex-founders doing a lot of bottoms up, you know, very exciting ideas. You know, for example, decisions API, you know, came together very quickly. We realized like we have a really good model with Luna. We can just do some, you know, constraint sampling and then, you know, have it, you know, shipped as like a different shape to the Responses API. And then, you know, when we do that, you know, it's like, it's faster and it's quite delightful for some cases. And it's just, if you were to like, … the hood, you realize like, okay, there's this … channel that came together. You know, initially, it was like four people kind of hacking on it on the weekend. - Who: Thibault Sottiaux, Head of ChatGPT and Codex, OpenAI, OpenAI (the organization sells AI tools) - Source: OpenAI's Head of ChatGPT: We're entering a new era of AI (again), Lenny's Podcast (YouTube), October 4, 2026, at 27:34. https://www.youtube.com/watch?v=MM-C3JqCXBk - Grade C: An interview or press quote. Not yet checked against the source. - Also filed under: Do organizations reorganize teams around AI? - Context (the archive's note, not their words): Asked what would surprise people about working inside OpenAI, Sottiaux describes how one API feature came together. - Passage page: https://the-ai-record.com/passages/openai-decisions-api-bottoms-up-weekend-hack-2026 ## How is AI used in sales and marketing? AI that prospects, answers sales questions, writes campaigns or supports account teams. Topic page: https://the-ai-record.com/topics/sales-and-marketing > Klarna has built an AI-powered copywriting tool, Copy Assistant, which allows the company to use AI for 80% of all copywriting. - Who: Klarna - Source: Klarna press releases, May 28, 2024. https://www.klarna.com/international/press/ai-helps-klarna-cut-marketing-agency-spend-by-25-and-run-more-campaigns/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-klarna-built-ai-powered-copywriting-tool-copy-2024 > We continue to enhance our toolkit for ad buyers, including AI features that help to dynamically adjust content mix to optimize campaign performance - Who: Walmart - Source: earnings call transcript via MarketBeat (third-party transcript site), May 21, 2026. https://www.marketbeat.com/earnings/reports/2026-5-21-walmart-inc-stock/ - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/walmart-continue-enhance-toolkit-ad-buyers-including-ai-2026 > In Acquiring, AI tools continue to drive significant improvements in sales force productivity, contributing to the strong market share gains we are seeing across the region. - Who: Mercado Libre - Source: MercadoLibre, Inc. Q1 2026 results (Form 8-K, Exhibit 99.1), SEC EDGAR, May 7, 2026, at To our Shareholders, Artificial Intelligence. https://www.sec.gov/Archives/edgar/data/1099590/000109959026000014/meli-20260507xex991.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/mercado-libre-acquiring-ai-tools-continue-drive-significant-2026 > Or another question, sales. Asking questions about, what is this feature? How do I sell it? Is it available in Brazil? What's the use case? What's the price? All of these things are fully understood. And gadget, if it does not understand, routes it to the right person. - Who: Geoff Charles, Chief Product Officer (CPO), Ramp - Source: What product looks like when coding is solved | Geoff Charles (Ramp CPO), Lenny's Podcast (YouTube channel); Geoff Charles's talk at the Lenny and Friends Summit, recorded 2026-09-10, September 25, 2026. https://www.youtube.com/watch?v=ZG8Mf3P9xzI - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-another-question-sales-asking-questions-about-what-2026 > In aggregate, Kai has helped shift 25,000 hours per year from administrative work, to revenue generating work. - Who: Sharadh Krishnamurthy, Engineering manager, Agent Foundation team, Stripe - Source: Stripe developer blog, July 30, 2026. https://stripe.dev/blog/meet-stripes-knowledge-ai-platform - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-aggregate-kai-helped-shift-25-000-hours-2026 > Marketing web quality assurance: 2 weeks → 50 minutes. - Who: Praveen Neppalli Naga, Chief Technology Officer, Uber - Source: X post on Praveen Neppalli's own account (Uber's CTO), July 7, 2026. https://x.com/praveenTweets/status/2074605343439810922 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-marketing-web-quality-assurance-2-weeks-50-2026 > Generated over 1,000 images in the first three months of 2024 using genAI, reducing the image development cycle from 6 weeks to just 7 days. - Who: Klarna - Source: Klarna press releases, May 28, 2024. https://www.klarna.com/international/press/ai-helps-klarna-cut-marketing-agency-spend-by-25-and-run-more-campaigns/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/klarna-generated-over-1-000-images-first-three-2024 > Colleague usage of AI tools increased threefold during H1 2026 and we launched AI-powered client intelligence which converts advisor calls into actionable insights - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), July 31, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/31072026/nwg-h1-2026-results.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-colleague-usage-ai-tools-increased-threefold-during-2026 > I think there's almost 1,000 use cases today, though we'd say that the really important ones are 50 across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation, document reading. - Who: JPMorgan Chase - Source: earnings call transcript via Webull (third-party transcript site), July 14, 2026. https://www.webull.com/news/15271033854174208 - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/jpmorgan-chase-i-think-theres-almost-1-000-use-2026 > we're seeing uptake of these tools, whether it's our legal team or marketing team or developers. And we think it's creating employees with superpowers. - Who: Dara Khosrowshahi, Chief Executive Officer, Uber - Source: Uber Q1 2026 earnings call transcript (FactSet CallStreet corrected transcript, hosted on Uber's investor site), May 6, 2026. https://s23.q4cdn.com/407969754/files/doc_financials/2026/q1/Uber-Q1-26-Earnings-Call-Transcript.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/uber-were-seeing-uptake-these-tools-whether-its-2026 ## How does AI help keep a product secure? AI used to find vulnerabilities, review security, and the risks agents add. Topic page: https://the-ai-record.com/topics/security > At Atlassian, agent loops in Jira automatically resolve 51% of potential vulnerabilities - Who: Atlassian (the organization sells AI tools) - Source: Atlassian (own site), October 1, 2026. https://www.atlassian.com/blog/ai-at-work/ai-sdlc-transformation-playbook - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/atlassian-atlassian-agent-loops-jira-automatically-resolve-51-2026 > Since launch, security merges through our freshness-gated merge queue went from about 10% to 80%. - Who: Erin Son, Senior Engineer, Shopify - Source: How River takes security work from a fix to merge, Shopify engineering blog, September 2, 2026. https://shopify.engineering/river-vulnerability-remediation - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-since-launch-security-merges-through-freshness-gated-2026 > To get ahead of this problem, we put a million dollars towards a sandbox challenge. We told the world … we believe is one of the most secure sandboxes, if not the most secure sandbox in the world, and to strengthen it further, we want you to poke holes. And we invited, I mean, you could imagine this, right? We invited the entire world to attack us. Only one person with like an army of agents was able to make progress. And we're basically processing his contribution now and we're sharing it with industry peers. - Who: Guillermo Rauch, Founder and CEO of Vercel, Vercel (the organization sells AI tools) - Source: 60% of Vercel Deployments Are Now Shipped by AI Agents, Sourcery with Molly O'Shea (YouTube), October 16, 2026, at 47:56. https://www.youtube.com/watch?v=WuorvLfAFmA - Grade C: An interview or press quote. Checked against the source. - Context (the archive's note, not their words): Rauch is talking about securing code and sandboxes against attackers who use coding agents. - Passage page: https://the-ai-record.com/passages/vercel-put-a-million-dollars-sandbox-challenge-one-person-army-of-agents-2026 > All of our infrastructure is on Azure. … Primarily for this reason, because, like, when we started, law firms were like, AWS? No. … Basically, we had to host everything on Microsoft because they already have three sixty five. - Who: Jacob Lauritzen, CTO, Legora (the organization sells AI tools) - Source: Building AI for Lawyers - Jacob Lauritzen, CTO of Legora, Latent Space, November 2, 2026, at transcript, around 28:30. https://latent.space/p/building-ai-for-lawyers-jacob-lauritzen-cto-of-legora/ - Grade C: An interview or press quote. Checked against the source. - Also filed under: Where do agents run? - Passage page: https://the-ai-record.com/passages/legora-all-infrastructure-on-azure-law-firms-security-review-2026 ## Does AI make human support agents faster? AI that summarizes, guides and drafts for human support staff, and what organizations report it changes for them. Topic page: https://the-ai-record.com/topics/support-staff-assist > AI tooling has reduced call summarisation time by more than 70% - Who: NatWest Group - Source: NatWest Group results announcement (the company's own regulatory results report), February 13, 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/results-center/13022026/nwg-company-announcement.pdf - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-ai-tooling-reduced-call-summarisation-time-more-2026 > ~9,000 contact centre colleagues are using AI to transcribe customer conversations and reducing time spent to search relevant data. - Who: NatWest Group - Source: NatWest Group website (the company's own sustainability page), October 29, 2026. https://www.natwestgroup.com/sustainability/society/innovation-and-digitisation.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-9-000-contact-centre-colleagues-using-ai-2026 > saving over 70,000 hours through automated call and chat summaries and another 17,500 hours through faster, high- quality complaint responses. - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-saving-over-70-000-hours-through-automated-2026 > 18,000 customer service representatives receive real-time guidance in seconds, helping them resolve client needs faster - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), October 14, 2026. https://newsroom.bankofamerica.com/content/newsroom/company-overview/bank-of-america-fast-facts.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-18-000-customer-service-representatives-receive-real-2026 > EricaAssist works alongside employees during calls – summarizing and surfacing relevant guidance in real time - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-ericaassist-works-alongside-employees-during-calls-summarizing-2026 > summarize why a client is calling, pull together relevant information, and recommend next steps based on the employee's role and the client's relationship with the bank - Who: Bank of America - Source: Bank of America newsroom (the company's own press site), July 21, 2026. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/07/bank-of-america-enhances-ericaassist-with-generative-ai-to-help-.html - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/bank-of-america-summarize-why-client-calling-pull-together-relevant-2026 > To better support our colleagues, we use AI summaries to help triage customer needs. In 2025, the technology benefitted approximately 9,000 customer support agents - Who: NatWest Group - Source: NatWest Group annual report, strategic report (the company's own annual publication), February 2026. https://investors.natwestgroup.com/~/media/Files/R/RBS-IR-V2/annual-reports/ar-2025/strategic-report-2025.pdf - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/natwest-better-support-colleagues-use-ai-summaries-help-2026 > By automating simple and repetitive tasks, our support engineers are able to augment and scale their time in a more valuable way - Who: Salesforce (the organization sells AI tools) - Source: SEC filing (the company's own regulatory report), March 2, 2026. https://www.sec.gov/Archives/edgar/data/1108524/000110852426000060/crm-20260131.htm - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/salesforce-automating-simple-repetitive-tasks-support-engineers-able-2026 ## How is AI used for testing? AI that writes, runs or repairs tests, and what that does to quality. Topic page: https://the-ai-record.com/topics/testing > The bigger reason is headless execution: our 1,700+ behavior tests finish in about four minutes on CI, on both platforms. - Who: Thiago Magalhaes, Author, Shopify Engineering blog, Shopify - Source: Reflex: an architecture for building mobile apps with agents, Shopify Engineering, November 5, 2026, at A contract built on first principles. https://shopify.engineering/reflex - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-1700-behavior-tests-four-minutes-ci-2026 > Stripe’s enormous preexisting battery of tests—over three million of them—can provide this feedback. - Who: Alistair Gray, Software engineer, Leverage team, Stripe - Source: Minions: Stripe’s one-shot, end-to-end coding agents—Part 2, Stripe developer blog, February 19, 2026. https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents-part-2 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/stripe-stripes-enormous-preexisting-battery-tests-over-three-2026 > We built twins of Okta, Jira, Slack, Google Docs, Google Drive, and Google Sheets, replicating their APIs, edge cases, and observable behaviors. - Who: StrongDM - Source: How StrongDM's AI team build serious software without even looking at the code, Simon Willison's Weblog, February 7, 2026. https://simonwillison.net/2026/feb/7/software-factory/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/strongdm-we-built-twins-okta-jira-slack-2026 ## How do organizations train people to use AI? Training programs, champions and the new skills people are asked to learn. Topic page: https://the-ai-record.com/topics/training-and-skills > We're helping our associates build the skills they'll need to thrive in an AI-Powered workplace through things like embracing OpenAI Certifications and rolling out ChatGPT Enterprise Licenses. - Who: Walmart - Source: earnings call transcript via MarketBeat (third-party transcript site), November 20, 2025. https://www.marketbeat.com/earnings/reports/2025-11-20-walmart-inc-stock/ - Grade A: A filing or earnings call the organization answers for. Checked against the source. - Passage page: https://the-ai-record.com/passages/walmart-were-helping-associates-build-skills-theyll-need-2025 > So periodically at night or in off hours, we give River, “Here’s all the conversations you’ve had today. What went well? What did you struggle with? - Who: Tobi Lütke, Co-founder and Chief Executive Officer, Shopify - Source: The Knowledge Project (podcast by Shane Parrish of Farnam Street, independent), transcript on fs.blog, September 10, 2026. https://fs.blog/knowledge-project-podcast/tobi-lutke-3/ - Grade C: An interview or press quote. Checked against the source. - Passage page: https://the-ai-record.com/passages/shopify-so-periodically-night-off-hours-give-river-2026 ## What counts as resolved when AI answers customers? How each organization defines a resolution, containment or deflection, and how those definitions differ. Topic page: https://the-ai-record.com/topics/what-counts-as-resolved > Spam and routine exchanges are excluded. - Who: Zendesk (the organization sells AI tools) - Source: Zendesk newsroom (the company's own press release), May 19, 2026. https://www.zendesk.com/newsroom/press-releases/relate-2026/ - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/zendesk-spam-routine-exchanges-excluded-2026 ## Where do agents run? The environments agents work in: sandboxes, cloud machines, background jobs, chat tools. Topic page: https://the-ai-record.com/topics/where-agents-run > So I talked about how the agent has this computer, this Linux box. This is running a firecracker virtual machine, and the reason this is really helpful is because it's very easy then to essentially snapshot the state of the machine and shut it down and start it back up. - Who: Lee Robinson, Works on Grok Bot at SpaceX (formerly Cursor), SpaceX (SpaceXAI) (the organization sells AI tools) - Source: Stanford CS146S: How always-on agents work, leerob (YouTube), October 11, 2026, at 12:24. https://www.youtube.com/watch?v=L1OBRtfPS34 - Grade B: The organization's own post, paper or talk. Checked against the source. - Context (the archive's note, not their words): Robinson is describing the infrastructure behind Grok Bot, the always-on agent his team at SpaceX builds, and the computer each agent gets in the cloud. - Passage page: https://the-ai-record.com/passages/spacex-grok-bot-agent-computer-firecracker-vm-2026 > Many Inspect sessions never need a sandbox at all, and a support agent can answer plenty of questions before it has any reason to run code or process a file. - Who: Sam Kronick, Author of Ramp's "Building Laminar" engineering blog post, Ramp - Source: Building Laminar: Ramp's Managed Agents Platform, Ramp Builders blog, October 20, 2026. https://builders.ramp.com/post/building-laminar - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/ramp-many-inspect-sessions-never-need-sandbox-2026 > simply asking Ona or the bot on my research failed, can you my analysis failed, can you tell me why? And it will just go through the entire troubleshooting process. - Who: Xiucheng Quek, Software engineer, Genentech - Source: Background Agents for Genomics and Cloud Ops, YouTube video of a talk or interview (speaker in notes), May 7, 2026. https://www.youtube.com/watch?v=GXDtw0EmoX0 - Grade B: The organization's own post, paper or talk. Checked against the source. - Passage page: https://the-ai-record.com/passages/genentech-simply-asking-ona-bot-my-research-failed-2026