How do coding agents open and merge changes?

“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.”

Geoff CharlesChief Product Officer (CPO) · Executive, Ramp

“More than 70% of pull requests are attributed to local or cloud agents.”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“Between 25-35% of the PRs we merge on a weekly basis are now authored by `ai-sdk-factory` agents.”

VercelPublished by the organization itself (post), no named author

“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.”

Jina YoonAuthor, PostHog newsletter · Practitioner, PostHog

“And today at Uber, one in nine of our PRs is fully production ready with zero human authoring from our background agents.”

Ty SmithPrincipal Engineer, AI & Developer Productivity · Practitioner, Uber

“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.”

BlockPublished by the organization itself (post), no named author

“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.”

BlockPublished by the organization itself (post), no named author

“we've merged more than 2.5 million automated maintenance PRs, the vast majority auto-merged with no human in the loop.”

SpotifyPublished by the organization itself (post), no named author

“About one in eight pull requests merged into our codebase last week was authored by River, reviewed by us.”

Tobi LütkeCo-founder and Chief Executive Officer · Executive, Shopify

“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.”

Tobi LütkeCo-founder and Chief Executive Officer · Executive, Shopify

“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.”

Praveen Neppalli NagaChief Technology Officer · Executive, Uber

“More than one-third of the PRs we merge are now created by agents that run on their own computers in the cloud.”

Cursor (Anysphere)Published by the organization itself (post), no named author

“Thirty-five percent of the PRs we merge internally at Cursor are now created by agents operating autonomously in cloud VMs.”

Cursor (Anysphere)Published by the organization itself (post), no named author

“Over 1,300 Stripe pull requests (up from 1,000 as of Part 1) merged each week are completely minion-produced”

Alistair GraySoftware engineer, Leverage team · Practitioner, Stripe

“They pull review feedback, respond inline, push updates, and often squash and merge their own pull requests.”

OpenAIPublished by the organization itself (post), no named author

“Over a thousand pull requests merged each week at Stripe are completely minion-produced”

Alistair GraySoftware engineer, Leverage team · Practitioner, Stripe

“we have merged over 7,000 fully AI-generated pull requests”

HubSpotPublished by the organization itself (post), no named author

“~30% of all pull requests merged to our frontend and backend repos are written by Inspect.”

Zach BruggemanPrincipal engineer; founding engineer of Inspect · Practitioner, Ramp

“our agents have generated more than 1,500 pull requests that teams across Spotify have merged into our production codebase.”

SpotifyPublished by the organization itself (post), no named author

“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”

Tobi LütkeCo-founder and Chief Executive Officer · Executive, Shopify

“one in three pull requests on GitHub now involves an agent”

Satya NadellaChairman and Chief Executive Officer · Executive, Microsoft

“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:”

WarpPublished by the organization itself (documentation), no named author

“A human intervenes at two points: approving the spec before implementation begins, and merging the PR after review. Everything else runs automatically.”

WarpPublished by the organization itself (documentation), no named author

“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.”

WarpPublished by the organization itself (documentation), no named author

“In May 2026, Anthropic averaged around 2,500 code changes merged per day; as of September, we merged over 30,000 per day.”

AnthropicPublished by the organization itself (post), no named author

“and 1,000 of those PRs over the last month was submitted by a non-engineer.”

Geoff CharlesChief Product Officer (CPO) · Executive, Ramp

“In the past four weeks, OCA assisted with 575 of the 1,220 PRs merged to address incidents.”

Richard MarmorsteinCo-author of Ramp's OCA post (job title not stated) · Practitioner, Ramp

“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.”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“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.”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“authoring ~10% of all merged PRs at Monzo”

MonzoPublished by the organization itself (post), no named author

“A human stays in control throughout the entire process, including merging every change.”

VercelPublished by the organization itself (post), no named author

“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.”

Noah BlauensteinerAuthor, Xebia articles · Practitioner, Xebia

“More than 70% of pull requests are attributed to local or cloud agents.”

Praveen Neppalli NagaChief Technology Officer · Executive, Uber

“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.”

Praveen Neppalli NagaChief Technology Officer · Executive, Uber

“accounting for roughly 1 in 12 pull requests at Dropbox today”

DropboxPublished by the organization itself (post), no named author

“3,536 River-coauthored pull requests merged in that window.”

Burke LibbeyPrincipal Engineer · Practitioner, Shopify

“Merged 160+ PRs using the automated upgrader”

Mahib HosainEngineer, Developer Productivity team · Practitioner, Stripe

“we are allowing the bot to merge that directly into our main branch because we have integration tests, unit tests, etc., canary tests.”

Xiucheng QuekSoftware engineer · Practitioner, Genentech

“1,800 code changes per week are now written entirely by Uber's internal background coding agent”

Praveen Neppalli NagaChief Technology Officer · Executive, Uber

“roughly 1,500 pull requests have been opened and merged with a small team of just three engineers driving Codex.”

OpenAIPublished by the organization itself (post), no named author

“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.”

OpenAIPublished by the organization itself (post), no named author

“while they're human-reviewed, they contain no human-written code.”

Alistair GraySoftware engineer, Leverage team · Practitioner, Stripe

“Codex reviews every internal pull request.”

OpenAIPublished by the organization itself (post), no named author

“It gets assigned issues by human engineers, opens pull requests, and does the work assigned.”

GitHubPublished by the organization itself (post), no named author

“All the surrounding Fleet Management infrastructure — targeting repositories, opening pull requests, getting reviews, and merging into production — remains exactly the same.”

SpotifyPublished by the organization itself (post), no named author

“every engineer must be able to explain, even if an AI generated the PR, must be able to explain what they built.”

Ahmad Al-DahleChief Technology Officer · Executive, Airbnb

“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.”

Venkat VenkataramaniVP of Engineering, Applied Infra · Leader, OpenAI

“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.”

Tobi LütkeCo-founder and Chief Executive Officer · Executive, Shopify

“about 30% of our PRs in that week came from Minions”

Will GaybrickPresident, Technology and Business (also styled President of Product and Business) · Executive, Stripe

“last week, I think 7,000 PRs came from minions”

Will GaybrickPresident, Technology and Business (also styled President of Product and Business) · Executive, Stripe

“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.”

Daksh GuptaCo-founder · Executive, Greptile

“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.”

Daksh GuptaCo-founder · Executive, Greptile

“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 …”

Eyal BlumSoftware Engineer · Practitioner, Figma

“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…”

Eyal BlumSoftware Engineer · Practitioner, Figma

“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.”

Eyal BlumSoftware Engineer · Practitioner, Figma

“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.”

Vaibhav GuptaCo-founder and Chief Executive · Executive, Boundary (BoundaryML)

“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?”

Andrew OrobatorSenior Software Engineer · Practitioner, Reddit

“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.”

Andrew OrobatorSenior Software Engineer · Practitioner, Reddit

“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.”

Andrew OrobatorSenior Software Engineer · Practitioner, Reddit

“Builderbot is an orchestration layer that coordinates multiple AI agents across our entire codebase. It works inside Slack: anyone can tag @builderbot with a short description of what they need, and it gets to work right there in the thread, whether it’s a bug fix, a migration across services, or a new feature.”

BlockPublished by the organization itself (post), no named author

“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.”

Brad AxenHead of AI Capabilities · Leader, Block

“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.”

FactoryPublished by the organization itself (post), no named author

“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?”

Walden YanCo-founder and Chief Product Officer · Executive, Cognition

“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.”

Walden YanCo-founder and Chief Product Officer · Executive, Cognition

“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 …”

Clare LiguoriSenior Principal Engineer, AWS · Practitioner, Amazon

“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.”

Clare LiguoriSenior Principal Engineer, AWS · Practitioner, Amazon

“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.”

Clare LiguoriSenior Principal Engineer, AWS · Practitioner, Amazon

“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.”

Ameya KetkarSpeaker, AI Engineer World's Fair 2026 (job title not stated) · Practitioner, Uber

“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.”

Rajesh BhatiaSenior Director of Engineering, Cloudflare · Leader, Cloudflare

“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.”

Rajesh BhatiaSenior Director of Engineering, Cloudflare · Leader, Cloudflare

“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,”

Geoff CharlesChief Product Officer (CPO) · Executive, Ramp

“It finds the right human reviewer to loop in.”

Geoff CharlesChief Product Officer (CPO) · Executive, Ramp

“Total merged changes more than doubled year over year in August, from roughly 8,100 to 17,000.”

SpotifyPublished by the organization itself (post), no named author

“the number of issues with a pull request attached by someone in an engineering, product, or design function has grown sevenfold”

LinearPublished by the organization itself (post), no named author

“Alert investigation: automatic investigations of paging alerts in production”

MonzoPublished by the organization itself (post), no named author

“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.”

Noah BlauensteinerAuthor, Xebia articles · Practitioner, Xebia

“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.”

Patrick DeboisProduct DevRel Lead · Practitioner, Tessl

“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.”

Patrick DeboisProduct DevRel Lead · Practitioner, Tessl

“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.”

Christopher SansonProduct Lead, AI Developer Experience · Leader, Airbnb

“People that were using agentic AI for four hours or more per day have seen PR output more than double from before AI.”

Christopher SansonProduct Lead, AI Developer Experience · Leader, Airbnb

“Our PR throughput inside of Airbnb is 65% higher than it was before we introduced agentic AI.”

Christopher SansonProduct Lead, AI Developer Experience · Leader, Airbnb

“more than 99% of our engineers use AI coding tools every week, 94% report that AI has made them more productive”

SpotifyPublished by the organization itself (post), no named author

“we're seeing a 76% increase in pull request frequency”

SpotifyPublished by the organization itself (post), no named author

“Mode · Automation: PR review. Durable session, woken by an external system, often no human in the loop.”

Burke LibbeyPrincipal Engineer · Practitioner, Shopify

“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”

Burke LibbeyPrincipal Engineer · Practitioner, Shopify

“Pull requests arrive faster than human-driven review can stay meaningful.”

Bence A. TóthAuthor on the Zendesk Engineering blog (role not stated on the pages read) · Practitioner, Zendesk

“Escalate to a human only when judgment is required”

OpenAIPublished by the organization itself (post), no named author

“code reviewed over 50,000 pull requests authored by humans”

HubSpotPublished by the organization itself (post), no named author

“engineers using it ship about 70% more merged pull requests”

OpenAIPublished by the organization itself (post), no named author

“roughly corroborates the 67% increase in merged pull requests”

AnthropicPublished by the organization itself (paper), no named author

“agents spin up to then triage and run your on-call”

Ahmad Al-DahleChief Technology Officer · Executive, Airbnb

“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.”

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

“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.”

Mike KriegerHead of Anthropic Labs · Leader, Anthropic