{ "apiVersion": "v1", "topic": { "id": "ai-review-quality", "title": "Does AI code review catch real problems?", "group": "Reviewing agent code", "summary": "How often AI review comments lead to a change, what kinds of problems they find, and where reviewers say the comments are noise.", "passages": 38, "url": "https://the-ai-record.com/topics/ai-review-quality", "json": "https://the-ai-record.com/api/v1/topics/ai-review-quality.json" }, "passages": [ { "id": "cloudflare-changes-authors-acted-34-code-quality-2026", "text": "Before the changes, authors acted on 34% of the code quality reviewer's serious findings. Now it's 59%.", "context": null, "speaker": "Ryan Skidmore", "speakerId": "ryan-skidmore", "speakerUrl": "https://the-ai-record.com/people/ryan-skidmore", "speakerLevel": "practitioner", "role": "Author of Cloudflare's \"Orchestrating AI Code Review at scale\" blog post", "voice": "person", "organization": "Cloudflare", "organizationId": "cloudflare", "organizationUrl": "https://the-ai-record.com/organizations/cloudflare", "aiVendor": true, "work": { "title": "Cutting our inference bill by 68% by being obsessive about optimization", "type": "post", "date": "2026-10-19", "publisher": "Cloudflare blog", "locator": "Review quality", "url": "https://blog.cloudflare.com/obsessive-inference-optimization/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "findings-acted-on-before", "value": 0.34, "unit": "share", "period": "before the September to October 2026 changes", "definition": "Share of the code quality reviewer's critical and warning findings where the author changed the flagged lines in their next push", "level": "n/a" }, { "metric": "findings-acted-on-after", "value": 0.59, "unit": "share", "period": "after the changes, one-day window", "definition": "Same measure, after the changes", "level": "n/a" } ], "url": "https://the-ai-record.com/passages/cloudflare-changes-authors-acted-34-code-quality-2026", "json": "https://the-ai-record.com/api/v1/passages/cloudflare-changes-authors-acted-34-code-quality-2026.json" }, { "id": "anthropic-measure-reviewer-noise-by-author-responses-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "How we make the reviewer better", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "measuring-impact" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/anthropic-measure-reviewer-noise-by-author-responses-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-measure-reviewer-noise-by-author-responses-2026.json" }, { "id": "anthropic-separate-agents-vote-on-each-review-finding-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "Layer by layer: 2. AI review on every pull request", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "code-review" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/anthropic-separate-agents-vote-on-each-review-finding-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-separate-agents-vote-on-each-review-finding-2026.json" }, { "id": "coval-reviews-capped-three-findings-silence-2026", "text": "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.", "context": null, "speaker": "Dana Dzik", "speakerId": "dana-dzik", "speakerUrl": "https://the-ai-record.com/people/dana-dzik", "speakerLevel": "practitioner", "role": "Engineer, Coval (author of the Coval blog post on its code reviewer)", "voice": "person", "organization": "Coval", "organizationId": "coval", "organizationUrl": "https://the-ai-record.com/organizations/coval", "aiVendor": true, "work": { "title": "Notes on Building a Code Reviewer", "type": "post", "date": "2026-09-10", "publisher": "Coval blog", "locator": "The rebuild", "url": "https://www.coval.ai/blog/how-we-built-our-code-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "review-runs-publishing-nothing", "value": 922, "unit": "count", "period": "1,732 recorded runs of the current design", "definition": "Review runs, of 1,732 recorded, that published no findings", "level": "human-approves" } ], "url": "https://the-ai-record.com/passages/coval-reviews-capped-three-findings-silence-2026", "json": "https://the-ai-record.com/api/v1/passages/coval-reviews-capped-three-findings-silence-2026.json" }, { "id": "uber-ureview-does-around-25-000-comments-2026", "text": "[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", "context": "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.", "speaker": "Ameya Ketkar", "speakerId": "ameya-ketkar", "speakerUrl": "https://the-ai-record.com/people/ameya-ketkar", "speakerLevel": "practitioner", "role": "Speaker, AI Engineer World's Fair 2026 (job title not stated)", "voice": "person", "organization": "Uber", "organizationId": "uber", "organizationUrl": "https://the-ai-record.com/organizations/uber", "aiVendor": false, "work": { "title": "Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026)", "type": "talk", "date": "2026-08-28", "publisher": "AI Engineer (YouTube)", "locator": "10:13", "url": "https://www.youtube.com/watch?v=EL123UNokkI" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-03" }, "topics": [ "code-review", "measuring-impact", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "ai-review-comments", "value": 25000, "unit": "comments", "period": "per week", "definition": "comments posted by uReview, Uber's AI code-review system", "level": "assisted" }, { "metric": "review-comment-address-rate", "value": 67, "unit": "percent", "period": "as of the talk", "definition": "share of uReview comments that the developer went and addressed, overall", "level": "assisted" } ], "url": "https://the-ai-record.com/passages/uber-ureview-does-around-25-000-comments-2026", "json": "https://the-ai-record.com/api/v1/passages/uber-ureview-does-around-25-000-comments-2026.json" }, { "id": "ionix-same-claude-session-review-missed-critical-bugs-2026", "text": "A lot of our code is now written with an AI assistant in the loop. The first thing we tried was the obvious thing: ask the same Claude session that wrote the change to review it. That does not work. It missed critical bugs, and it missed them confidently, because it had already reasoned its way to believing the code was correct, and re-reading its own work only confirmed the conclusion.", "context": null, "speaker": "Arthur Vinberg", "speakerId": "arthur-vinberg", "speakerUrl": "https://the-ai-record.com/people/arthur-vinberg", "speakerLevel": "leader", "role": "Head of DevOps & IT", "voice": "person", "organization": "Ionix", "organizationId": "ionix", "organizationUrl": "https://the-ai-record.com/organizations/ionix", "aiVendor": false, "work": { "title": "We built an AI PR reviewer. The hard part was teaching it to say nothing.", "type": "post", "date": "2026-08-13", "publisher": "Ionix blog", "locator": "A reviewer that didn’t write the code", "url": "https://www.ionix.io/blog/we-built-an-ai-pr-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "ai-review-quality", "code-review" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/ionix-same-claude-session-review-missed-critical-bugs-2026", "json": "https://the-ai-record.com/api/v1/passages/ionix-same-claude-session-review-missed-critical-bugs-2026.json" }, { "id": "ona-already-code-review-automation-running-every-pr-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Ona", "organizationId": "ona", "organizationUrl": "https://the-ai-record.com/organizations/ona", "aiVendor": true, "work": { "title": "How auto-approving low-risk PRs with AI cut our lead time by 74%", "type": "post", "date": "2026-04-09", "publisher": "Ona blog", "locator": "The experiment", "url": "https://ona.com/stories/auto-approving-low-risk-prs" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "measuring-impact", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "review-comment-acceptance-rate", "value": 0.88, "unit": "share", "period": "before auto-approval, 2026", "definition": "Share of the agent's inline review comments that human reviewers accepted", "level": "assisted" }, { "metric": "time-spent-on-pull-requests-reduction", "value": 0.5, "unit": "share", "period": "before auto-approval, 2026", "definition": "Reduction in time spent on pull requests, as the post states it", "level": "assisted" } ], "url": "https://the-ai-record.com/passages/ona-already-code-review-automation-running-every-pr-2026", "json": "https://the-ai-record.com/api/v1/passages/ona-already-code-review-automation-running-every-pr-2026.json" }, { "id": "anthropic-large-prs-84-get-findings-under-1-percent-incorrect-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "Bringing Code Review to Claude Code", "type": "post", "date": "2026-03-09", "publisher": "claude.com (Anthropic)", "locator": "Code Review in action", "url": "https://claude.com/resources/articles/code-review" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "code-review" ], "functions": [ "engineering" ], "figures": [ { "metric": "findings-marked-incorrect", "value": 0.01, "unit": "share", "period": "internal use for months before March 2026 (upper bound, 'less than 1%')", "definition": "share of Code Review findings that Anthropic engineers marked incorrect", "level": "human-approves" } ], "url": "https://the-ai-record.com/passages/anthropic-large-prs-84-get-findings-under-1-percent-incorrect-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-large-prs-84-get-findings-under-1-percent-incorrect-2026.json" }, { "id": "atlassian-rovodev-code-reviewer-effective-generating-comments-resolution-2026", "text": "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)", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Atlassian", "organizationId": "atlassian", "organizationUrl": "https://the-ai-record.com/organizations/atlassian", "aiVendor": true, "work": { "title": "RovoDev Code Reviewer: A Large-Scale Online Evaluation of LLM-based Code Review Automation at Atlassian", "type": "paper", "date": "2026-01-05", "publisher": "arXiv (Atlassian researchers; accepted at ICSE 2026, SEIP track)", "locator": "Abstract", "url": "https://arxiv.org/abs/2601.01129" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-06" }, "topics": [ "code-review", "measuring-impact", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "code-review-comment-resolution-rate", "value": 0.387, "unit": "share", "period": "one-year evaluation", "definition": "RovoDev Code Reviewer comments that triggered code changes in subsequent commits", "level": "human-approves" } ], "url": "https://the-ai-record.com/passages/atlassian-rovodev-code-reviewer-effective-generating-comments-resolution-2026", "json": "https://the-ai-record.com/api/v1/passages/atlassian-rovodev-code-reviewer-effective-generating-comments-resolution-2026.json" }, { "id": "atlassian-developers-most-likely-resolve-comments-readability-2025", "text": "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%.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Atlassian", "organizationId": "atlassian", "organizationUrl": "https://the-ai-record.com/organizations/atlassian", "aiVendor": true, "work": { "title": "Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?", "type": "post", "date": "2025-11-10", "publisher": "Inside Atlassian (Atlassian engineering blog)", "locator": "Which Comments Actually Get Resolved?", "url": "https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-06" }, "topics": [ "code-review", "measuring-impact", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/atlassian-developers-most-likely-resolve-comments-readability-2025", "json": "https://the-ai-record.com/api/v1/passages/atlassian-developers-most-likely-resolve-comments-readability-2025.json" }, { "id": "uber-ureview-today-analyzes-over-90-weekly-65-2025-2", "text": "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", "context": null, "speaker": "Sonal Mahajan", "speakerId": "sonal-mahajan", "speakerUrl": "https://the-ai-record.com/people/sonal-mahajan", "speakerLevel": "practitioner", "role": "Staff Engineer, Programming Systems team", "voice": "person", "organization": "Uber", "organizationId": "uber", "organizationUrl": "https://the-ai-record.com/organizations/uber", "aiVendor": false, "work": { "title": "uReview: Scalable, Trustworthy GenAI for Code Review at Uber", "type": "post", "date": "2025-08-12", "publisher": "Uber engineering blog", "url": "https://www.uber.com/us/en/blog/ureview/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-10-31" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/uber-ureview-today-analyzes-over-90-weekly-65-2025-2", "json": "https://the-ai-record.com/api/v1/passages/uber-ureview-today-analyzes-over-90-weekly-65-2025-2.json" }, { "id": "cloudflare-important-findings-per-review-went-028-2026", "text": "Important findings per review went from 0.28 before to 0.30 now, and the intervals overlap.", "context": null, "speaker": "Ryan Skidmore", "speakerId": "ryan-skidmore", "speakerUrl": "https://the-ai-record.com/people/ryan-skidmore", "speakerLevel": "practitioner", "role": "Author of Cloudflare's \"Orchestrating AI Code Review at scale\" blog post", "voice": "person", "organization": "Cloudflare", "organizationId": "cloudflare", "organizationUrl": "https://the-ai-record.com/organizations/cloudflare", "aiVendor": true, "work": { "title": "Cutting our inference bill by 68% by being obsessive about optimization", "type": "post", "date": "2026-10-19", "publisher": "Cloudflare blog", "locator": "Review quality", "url": "https://blog.cloudflare.com/obsessive-inference-optimization/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "measuring-impact" ], "functions": [ "engineering" ], "figures": [ { "metric": "important-findings-per-review-before", "value": 0.28, "unit": "count", "period": "before the changes", "definition": "Findings per review that two model judges from different providers labelled important, on a blinded sample", "level": "n/a" }, { "metric": "important-findings-per-review-after", "value": 0.3, "unit": "count", "period": "after the changes", "definition": "Same measure, after the changes", "level": "n/a" } ], "url": "https://the-ai-record.com/passages/cloudflare-important-findings-per-review-went-028-2026", "json": "https://the-ai-record.com/api/v1/passages/cloudflare-important-findings-per-review-went-028-2026.json" }, { "id": "cloudflare-medium-effort-did-lose-real-findings-2026", "text": "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.", "context": null, "speaker": "Ryan Skidmore", "speakerId": "ryan-skidmore", "speakerUrl": "https://the-ai-record.com/people/ryan-skidmore", "speakerLevel": "practitioner", "role": "Author of Cloudflare's \"Orchestrating AI Code Review at scale\" blog post", "voice": "person", "organization": "Cloudflare", "organizationId": "cloudflare", "organizationUrl": "https://the-ai-record.com/organizations/cloudflare", "aiVendor": true, "work": { "title": "Cutting our inference bill by 68% by being obsessive about optimization", "type": "post", "date": "2026-10-19", "publisher": "Cloudflare blog", "locator": "Review quality", "url": "https://blog.cloudflare.com/obsessive-inference-optimization/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "ai-review-cost" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/cloudflare-medium-effort-did-lose-real-findings-2026", "json": "https://the-ai-record.com/api/v1/passages/cloudflare-medium-effort-did-lose-real-findings-2026.json" }, { "id": "anthropic-one-pr-fifty-review-rounds-overnight-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "How we make the reviewer better", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "ai-review-cost" ], "functions": [ "engineering" ], "figures": [ { "metric": "review-rounds-single-pr", "value": 50, "unit": "rounds", "period": "one night, one Claude Code CLI pull request", "definition": "Rounds of AI review and agent fixes on a single pull request before its author stopped it (\"more than fifty\"; lower bound)", "level": "human-out" }, { "metric": "repeat-findings-share", "value": 0.59, "unit": "share", "definition": "Share of what the AI reviewer kept flagging that was code present since its first pass, found with git blame (\"about 59%\")", "level": "n/a" } ], "url": "https://the-ai-record.com/passages/anthropic-one-pr-fifty-review-rounds-overnight-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-one-pr-fifty-review-rounds-overnight-2026.json" }, { "id": "anthropic-past-incidents-become-daily-eval-for-reviewer-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "How we make the reviewer better", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-evals", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/anthropic-past-incidents-become-daily-eval-for-reviewer-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-past-incidents-become-daily-eval-for-reviewer-2026.json" }, { "id": "anthropic-prs-passing-after-one-review-round-65-to-78-2026", "text": "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.", "context": "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.", "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "How we make the reviewer better: Figure D, Review that converges (the chart's text description)", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-cost", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "prs-pass-after-one-review-round", "value": 0.78, "unit": "share", "period": "internal pull requests, after the convergence changes (0.65 before)", "definition": "Share of pull requests passing after one AI review and fix round", "level": "unclear" }, { "metric": "prs-needing-five-plus-review-rounds", "value": 0.015, "unit": "share", "period": "internal pull requests, after the convergence changes (0.089 before)", "definition": "Share of pull requests needing five or more review rounds", "level": "unclear" }, { "metric": "avg-review-iterations", "value": 1.32, "unit": "iterations per PR", "period": "internal pull requests, after the convergence changes (2.08 before)", "definition": "Mean reviews to pass", "level": "unclear" } ], "url": "https://the-ai-record.com/passages/anthropic-prs-passing-after-one-review-round-65-to-78-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-prs-passing-after-one-review-round-65-to-78-2026.json" }, { "id": "anthropic-review-noise-costs-more-when-author-agent-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "How we make the reviewer better", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "ai-review-cost" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/anthropic-review-noise-costs-more-when-author-agent-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-review-noise-costs-more-when-author-agent-2026.json" }, { "id": "anthropic-substantive-review-comments-16-to-54-percent-2026", "text": "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.", "context": "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.", "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Anthropic", "organizationId": "anthropic", "organizationUrl": "https://the-ai-record.com/organizations/anthropic", "aiVendor": true, "work": { "title": "How the Claude Code team does code review", "type": "post", "date": "2026-10-15", "publisher": "claude.dev Blog (Anthropic)", "locator": "Layer by layer: 2. AI review on every pull request", "url": "https://claude.dev/blog/solving-code-review-at-anthropic/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "code-review" ], "functions": [ "engineering" ], "figures": [ { "metric": "prs-with-substantive-review-comments", "value": 0.54, "unit": "share", "period": "internal pull requests, after the Code Review research preview launched (0.16 before)", "definition": "Share of internal pull requests getting substantive review comments", "level": "assisted" } ], "url": "https://the-ai-record.com/passages/anthropic-substantive-review-comments-16-to-54-percent-2026", "json": "https://the-ai-record.com/api/v1/passages/anthropic-substantive-review-comments-16-to-54-percent-2026.json" }, { "id": "traderspost-review-bot-posts-every-pull-request-waste-2026", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "TradersPost", "organizationId": "traderspost", "organizationUrl": "https://the-ai-record.com/organizations/traderspost", "aiVendor": false, "work": { "title": "The AI Review Bot We Reverted", "type": "post", "date": "2026-10-13", "publisher": "TradersPost blog", "locator": "A noisy bot costs attention", "url": "https://blog.traderspost.io/article/the-ai-review-bot-we-built-and-reverted" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-05" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/traderspost-review-bot-posts-every-pull-request-waste-2026", "json": "https://the-ai-record.com/api/v1/passages/traderspost-review-bot-posts-every-pull-request-waste-2026.json" }, { "id": "coval-asked-name-sensitive-risk-domains-2026", "text": "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.", "context": null, "speaker": "Dana Dzik", "speakerId": "dana-dzik", "speakerUrl": "https://the-ai-record.com/people/dana-dzik", "speakerLevel": "practitioner", "role": "Engineer, Coval (author of the Coval blog post on its code reviewer)", "voice": "person", "organization": "Coval", "organizationId": "coval", "organizationUrl": "https://the-ai-record.com/organizations/coval", "aiVendor": true, "work": { "title": "Notes on Building a Code Reviewer", "type": "post", "date": "2026-09-10", "publisher": "Coval blog", "locator": "The rebuild", "url": "https://www.coval.ai/blog/how-we-built-our-code-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "claims-resolving-to-real-code", "value": 0.98, "unit": "share", "period": "570 claims when the reviewer must quote the changed line", "definition": "Share of 570 review claims that resolved to real code when each claim had to quote a changed line", "level": "human-approves" } ], "url": "https://the-ai-record.com/passages/coval-asked-name-sensitive-risk-domains-2026", "json": "https://the-ai-record.com/api/v1/passages/coval-asked-name-sensitive-risk-domains-2026.json" }, { "id": "coval-configuration-catches-most-what-our-2026", "text": "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.", "context": null, "speaker": "Dana Dzik", "speakerId": "dana-dzik", "speakerUrl": "https://the-ai-record.com/people/dana-dzik", "speakerLevel": "practitioner", "role": "Engineer, Coval (author of the Coval blog post on its code reviewer)", "voice": "person", "organization": "Coval", "organizationId": "coval", "organizationUrl": "https://the-ai-record.com/organizations/coval", "aiVendor": true, "work": { "title": "Notes on Building a Code Reviewer", "type": "post", "date": "2026-09-10", "publisher": "Coval blog", "locator": "Where it goes", "url": "https://www.coval.ai/blog/how-we-built-our-code-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-cost", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/coval-configuration-catches-most-what-our-2026", "json": "https://the-ai-record.com/api/v1/passages/coval-configuration-catches-most-what-our-2026.json" }, { "id": "uber-one-biggest-learnings-process-like-model-2026", "text": "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.", "context": "Ameya Ketkar is describing what Uber's team learned while adding observability and evaluation to its uReview code-review agent.", "speaker": "Ameya Ketkar", "speakerId": "ameya-ketkar", "speakerUrl": "https://the-ai-record.com/people/ameya-ketkar", "speakerLevel": "practitioner", "role": "Speaker, AI Engineer World's Fair 2026 (job title not stated)", "voice": "person", "organization": "Uber", "organizationId": "uber", "organizationUrl": "https://the-ai-record.com/organizations/uber", "aiVendor": false, "work": { "title": "Building uReview, Uber's Multi-Agent Code Review Engine (talk, AI Engineer World's Fair 2026)", "type": "talk", "date": "2026-08-28", "publisher": "AI Engineer (YouTube)", "locator": "06:35", "url": "https://www.youtube.com/watch?v=EL123UNokkI" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-03" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/uber-one-biggest-learnings-process-like-model-2026", "json": "https://the-ai-record.com/api/v1/passages/uber-one-biggest-learnings-process-like-model-2026.json" }, { "id": "ionix-most-days-one-sentence-change-looks-fine-2026", "text": "Most days the entire output is one sentence saying the change looks fine, which is what we built it to do. On the days it leaves an inline comment, someone reads it.", "context": null, "speaker": "Arthur Vinberg", "speakerId": "arthur-vinberg", "speakerUrl": "https://the-ai-record.com/people/arthur-vinberg", "speakerLevel": "leader", "role": "Head of DevOps & IT", "voice": "person", "organization": "Ionix", "organizationId": "ionix", "organizationUrl": "https://the-ai-record.com/organizations/ionix", "aiVendor": false, "work": { "title": "We built an AI PR reviewer. The hard part was teaching it to say nothing.", "type": "post", "date": "2026-08-13", "publisher": "Ionix blog", "locator": "If you’re building one of these", "url": "https://www.ionix.io/blog/we-built-an-ai-pr-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "ai-review-quality", "code-review" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/ionix-most-days-one-sentence-change-looks-fine-2026", "json": "https://the-ai-record.com/api/v1/passages/ionix-most-days-one-sentence-change-looks-fine-2026.json" }, { "id": "ionix-permission-find-nothing-zero-findings-clean-2026", "text": "The second is permission to find nothing. The prompt says outright that returning zero findings is the correct and expected outcome for a clean PR, and that findings must never be manufactured to look thorough.", "context": null, "speaker": "Arthur Vinberg", "speakerId": "arthur-vinberg", "speakerUrl": "https://the-ai-record.com/people/arthur-vinberg", "speakerLevel": "leader", "role": "Head of DevOps & IT", "voice": "person", "organization": "Ionix", "organizationId": "ionix", "organizationUrl": "https://the-ai-record.com/organizations/ionix", "aiVendor": false, "work": { "title": "We built an AI PR reviewer. The hard part was teaching it to say nothing.", "type": "post", "date": "2026-08-13", "publisher": "Ionix blog", "locator": "The noise budget", "url": "https://www.ionix.io/blog/we-built-an-ai-pr-reviewer/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/ionix-permission-find-nothing-zero-findings-clean-2026", "json": "https://the-ai-record.com/api/v1/passages/ionix-permission-find-nothing-zero-findings-clean-2026.json" }, { "id": "sentry-security-scanning-five-dollars-a-pr-explicit-decision-worth-it-2026", "text": "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.", "context": "Pstrucha is answering the moderator's question: 'when does a loop pay for itself, and how often is this actually the case?'", "speaker": "Greg Pstrucha", "speakerId": "greg-pstrucha", "speakerUrl": "https://the-ai-record.com/people/greg-pstrucha", "speakerLevel": "practitioner", "role": "Developer, Sentry", "voice": "person", "organization": "Sentry", "organizationId": "sentry", "organizationUrl": "https://the-ai-record.com/organizations/sentry", "aiVendor": true, "work": { "title": "The Great Loops Debate", "type": "talk", "date": "2026-07-17", "publisher": "AI Engineer (YouTube)", "locator": "36:59", "url": "https://www.youtube.com/watch?v=c35YoMdnI78" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "security", "ai-cost-and-spend", "ai-review-quality", "ai-review-cost" ], "functions": [ "engineering", "security" ], "figures": [ { "metric": "cost-per-pr-checks", "value": 5, "unit": "usd", "period": "per pull request, as the speaker states it", "definition": "cost of running all the security scans he wants on one pull request, 'I think like five bucks a PR or something like that'", "level": "unclear" } ], "url": "https://the-ai-record.com/passages/sentry-security-scanning-five-dollars-a-pr-explicit-decision-worth-it-2026", "json": "https://the-ai-record.com/api/v1/passages/sentry-security-scanning-five-dollars-a-pr-explicit-decision-worth-it-2026.json" }, { "id": "cloudflare-engineers-dont-trust-ai-opinions-clear-findings-2026", "text": "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.", "context": "Describing early misses with Cloudflare's internal AI code reviewer, before the team tied its findings to machine-readable engineering standards.", "speaker": "Rajesh Bhatia", "speakerId": "rajesh-bhatia", "speakerUrl": "https://the-ai-record.com/people/rajesh-bhatia", "speakerLevel": "leader", "role": "Senior Director of Engineering, Cloudflare", "voice": "person", "organization": "Cloudflare", "organizationId": "cloudflare", "organizationUrl": "https://the-ai-record.com/organizations/cloudflare", "aiVendor": true, "work": { "title": "From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026)", "type": "talk", "date": "2026-05-06", "publisher": "Ona (YouTube)", "locator": "26:40", "url": "https://www.youtube.com/watch?v=MbLdrAZFQRs" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-03" }, "topics": [ "code-review", "autonomy-and-oversight", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/cloudflare-engineers-dont-trust-ai-opinions-clear-findings-2026", "json": "https://the-ai-record.com/api/v1/passages/cloudflare-engineers-dont-trust-ai-opinions-clear-findings-2026.json" }, { "id": "juspay-memory-matters-per-pr-conversation-context-review-quality-2026", "text": "Memory matters: Per-PR conversation context dramatically improved review quality over stateless approaches", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Juspay", "organizationId": "juspay", "organizationUrl": "https://the-ai-record.com/organizations/juspay", "aiVendor": true, "work": { "title": "How We Built an AI Code Reviewer with NeuroLink and Bitbucket", "type": "post", "date": "2026-04-06", "publisher": "dev.to (NeuroLink AI organization account, Juspay)", "locator": "Results & Lessons Learned > Key Lessons", "url": "https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/juspay-memory-matters-per-pr-conversation-context-review-quality-2026", "json": "https://the-ai-record.com/api/v1/passages/juspay-memory-matters-per-pr-conversation-context-review-quality-2026.json" }, { "id": "atlassian-rovo-dev-focusing-actionable-comments-only-2025", "text": "In addition, Rovo Dev is focusing only on actionable comments with only 0.6% classified as “No Issue”.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Atlassian", "organizationId": "atlassian", "organizationUrl": "https://the-ai-record.com/organizations/atlassian", "aiVendor": true, "work": { "title": "Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?", "type": "post", "date": "2025-11-10", "publisher": "Inside Atlassian (Atlassian engineering blog)", "locator": "Results", "url": "https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-06" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/atlassian-rovo-dev-focusing-actionable-comments-only-2025", "json": "https://the-ai-record.com/api/v1/passages/atlassian-rovo-dev-focusing-actionable-comments-only-2025.json" }, { "id": "atlassian-rovo-dev-generated-higher-proportion-bug-2025", "text": "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.", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Atlassian", "organizationId": "atlassian", "organizationUrl": "https://the-ai-record.com/organizations/atlassian", "aiVendor": true, "work": { "title": "Atlassian Rovo Dev Research: What Types of Code Review Comments Do Developers Most Frequently Resolve?", "type": "post", "date": "2025-11-10", "publisher": "Inside Atlassian (Atlassian engineering blog)", "locator": "Results", "url": "https://www.atlassian.com/blog/atlassian-engineering/atlassian-rovo-dev-research-what-types-of-code-review-comments-do-developers-most-frequently-resolve" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-06" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "bug-comment-proportion-ratio", "value": 2.8, "unit": "ratio", "period": "internal projects, ~4,000 comments", "definition": "Rovo Dev's proportion of bug comments relative to human reviewers' proportion", "level": "human-approves" } ], "url": "https://the-ai-record.com/passages/atlassian-rovo-dev-generated-higher-proportion-bug-2025", "json": "https://the-ai-record.com/api/v1/passages/atlassian-rovo-dev-generated-higher-proportion-bug-2025.json" }, { "id": "cognition-devin-fixes-own-comments-infinite-loop-high-signal-2026", "text": "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.”", "context": null, "speaker": "Walden Yan", "speakerId": "walden-yan", "speakerUrl": "https://the-ai-record.com/people/walden-yan", "speakerLevel": "executive", "role": "Co-founder and Chief Product Officer", "voice": "person", "organization": "Cognition", "organizationId": "cognition", "organizationUrl": "https://the-ai-record.com/organizations/cognition", "aiVendor": true, "work": { "title": "The Age of Async Agents: Cognition's Walden Yan and OpenInspect's Cole Murray", "type": "podcast", "date": "2026-05-28", "publisher": "Latent Space", "locator": "00:24:37", "url": "https://latent.space/p/cognition" }, "grade": "C", "gradeMeaning": "An interview or press quote.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-03" }, "topics": [ "code-review", "autonomy-and-oversight", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/cognition-devin-fixes-own-comments-infinite-loop-high-signal-2026", "json": "https://the-ai-record.com/api/v1/passages/cognition-devin-fixes-own-comments-infinite-loop-high-signal-2026.json" }, { "id": "block-bots-review-prs-auto-fix-loop-2026", "text": "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.", "context": "Jones is describing how Block handled code review once engineers were running several agents at once.", "speaker": "Angie Jones", "speakerId": "angie-jones", "speakerUrl": "https://the-ai-record.com/people/angie-jones", "speakerLevel": "leader", "role": "Led AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation", "voice": "person", "organization": "Block", "organizationId": "block", "organizationUrl": "https://the-ai-record.com/organizations/block", "aiVendor": false, "work": { "title": "Building an Autonomous Engineering Org", "type": "talk", "date": "2026-06-28", "publisher": "AI Engineer (YouTube)", "locator": "13:31", "url": "https://www.youtube.com/watch?v=whue9_YquGA" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "reported", "meaning": "Not yet confirmed against the source by an independent second reader." }, "topics": [ "code-review", "coding-agents", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/block-bots-review-prs-auto-fix-loop-2026", "json": "https://the-ai-record.com/api/v1/passages/block-bots-review-prs-auto-fix-loop-2026.json" }, { "id": "madewithlove-bugbot-ran-month-caught-around-ten-real-bugs-2025", "text": "In our case, Bugbot ran for a month and caught around ten real bugs. One of the best examples was a subtle mix-up between two different phone number lists in the code.", "context": null, "speaker": "Peter Eysermans", "speakerId": "peter-eysermans", "speakerUrl": "https://the-ai-record.com/people/peter-eysermans", "speakerLevel": "practitioner", "role": "Author, madewithlove blog (role not stated on the page)", "voice": "person", "organization": "madewithlove", "organizationId": "madewithlove", "organizationUrl": "https://the-ai-record.com/organizations/madewithlove", "aiVendor": false, "work": { "title": "Automatic pull request reviewing with Cursor's Bugbot", "type": "post", "date": "2025-09-02", "publisher": "madewithlove blog", "locator": "How good is it, really?", "url": "https://madewithlove.com/blog/automatic-pull-request-reviewing-with-cursors-bugbot" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [ { "metric": "real-bugs-caught-by-ai-reviewer", "value": 10, "unit": "count", "period": "one month, one team", "definition": "real bugs Bugbot found in pull requests, stated as 'around ten'; AI assists, a person decides", "level": "n/a" } ], "url": "https://the-ai-record.com/passages/madewithlove-bugbot-ran-month-caught-around-ten-real-bugs-2025", "json": "https://the-ai-record.com/api/v1/passages/madewithlove-bugbot-ran-month-caught-around-ten-real-bugs-2025.json" }, { "id": "madewithlove-it-occasionally-throws-false-positives-example-2025", "text": "It occasionally throws false positives. For instance, in the review above, it flagged swapped parameters that weren’t actually swapped. It’s also overly cautious about nullable types, warning about missing null checks even in safe contexts.", "context": null, "speaker": "Peter Eysermans", "speakerId": "peter-eysermans", "speakerUrl": "https://the-ai-record.com/people/peter-eysermans", "speakerLevel": "practitioner", "role": "Author, madewithlove blog (role not stated on the page)", "voice": "person", "organization": "madewithlove", "organizationId": "madewithlove", "organizationUrl": "https://the-ai-record.com/organizations/madewithlove", "aiVendor": false, "work": { "title": "Automatic pull request reviewing with Cursor's Bugbot", "type": "post", "date": "2025-09-02", "publisher": "madewithlove blog", "locator": "How good is it, really?", "url": "https://madewithlove.com/blog/automatic-pull-request-reviewing-with-cursors-bugbot" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/madewithlove-it-occasionally-throws-false-positives-example-2025", "json": "https://the-ai-record.com/api/v1/passages/madewithlove-it-occasionally-throws-false-positives-example-2025.json" }, { "id": "cloudflare-switching-all-reviews-v1-v2-ran-2026", "text": "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.", "context": null, "speaker": "Ryan Skidmore", "speakerId": "ryan-skidmore", "speakerUrl": "https://the-ai-record.com/people/ryan-skidmore", "speakerLevel": "practitioner", "role": "Author of Cloudflare's \"Orchestrating AI Code Review at scale\" blog post", "voice": "person", "organization": "Cloudflare", "organizationId": "cloudflare", "organizationUrl": "https://the-ai-record.com/organizations/cloudflare", "aiVendor": true, "work": { "title": "Cutting our inference bill by 68% by being obsessive about optimization", "type": "post", "date": "2026-10-19", "publisher": "Cloudflare blog", "locator": "What else has changed since our last post?", "url": "https://blog.cloudflare.com/obsessive-inference-optimization/" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-07" }, "topics": [ "ai-review-quality", "ai-evals" ], "functions": [ "engineering" ], "figures": [ { "metric": "v1-v2-approval-agreement", "value": 0.98, "unit": "share", "period": "shadow-mode comparison before early October 2026", "definition": "Share of tested merge requests where OpenCode v1 and v2 reviewers agreed on the approval decision", "level": "n/a" } ], "url": "https://the-ai-record.com/passages/cloudflare-switching-all-reviews-v1-v2-ran-2026", "json": "https://the-ai-record.com/api/v1/passages/cloudflare-switching-all-reviews-v1-v2-ran-2026.json" }, { "id": "nordic-corporate-bank-three-model-review-pool-2026", "text": "And then the self review, which is kind of interesting because we have this, we created this review pool. So we have one model, which is a Sonnet that does one review. … then we have kind of a … cheaper model that also does a review. And then we have the third model, … right now, it kind of consolidates this and verifies the findings in the reviews and decides if it should change something or not. And when this review pool is kind of finished and happy, … will assign it to a human.", "context": "The speaker describes how the agent reviews its own pull request with three models before handing it to a person.", "speaker": "Hallstein Brøtan", "speakerId": "hallstein-brotan", "speakerUrl": "https://the-ai-record.com/people/hallstein-brotan", "speakerLevel": "practitioner", "role": "Senior consultant, Novanet; developer on Nordic Corporate Bank's system", "voice": "person", "organization": "Nordic Corporate Bank", "organizationId": "nordic-corporate-bank", "organizationUrl": "https://the-ai-record.com/organizations/nordic-corporate-bank", "aiVendor": false, "work": { "title": "How We Built an Autonomous Development Workflow at Nordic Corporate Bank", "type": "talk", "date": "2026-06-29", "publisher": "NDC Conferences (YouTube); talk recorded at NDC in Oslo", "locator": "37:00", "url": "https://www.youtube.com/watch?v=p8M8--12h1Y" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-08" }, "topics": [ "code-review", "ai-review-quality", "choosing-models" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/nordic-corporate-bank-three-model-review-pool-2026", "json": "https://the-ai-record.com/api/v1/passages/nordic-corporate-bank-three-model-review-pool-2026.json" }, { "id": "cursor-bugbot-rule-never-use-foreign-keys-2026", "text": "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.", "context": "Zakariasson is giving an example of a rule Cursor keeps for its Bugbot code-review tool, in an answer about writing rules for agents.", "speaker": "Eric Zakariasson", "speakerId": "eric-zakariasson", "speakerUrl": "https://the-ai-record.com/people/eric-zakariasson", "speakerLevel": "practitioner", "role": "Engineer, Developer Experience and Product, Cursor", "voice": "person", "organization": "Cursor (Anysphere)", "organizationId": "cursor", "organizationUrl": "https://the-ai-record.com/organizations/cursor", "aiVendor": true, "work": { "title": "Building your own software factory", "type": "talk", "date": "2026-04-28", "publisher": "AI Engineer (YouTube)", "locator": "45:06", "url": "https://www.youtube.com/watch?v=rnDm57Py54A" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/cursor-bugbot-rule-never-use-foreign-keys-2026", "json": "https://the-ai-record.com/api/v1/passages/cursor-bugbot-rule-never-use-foreign-keys-2026.json" }, { "id": "juspay-off-the-shelf-ai-code-review-tools-generic-checks-haskell-2026", "text": "Off-the-shelf AI code review tools: Great for generic checks, but couldn't understand our Haskell payment systems or custom conventions", "context": null, "speaker": null, "speakerId": null, "speakerUrl": null, "speakerLevel": null, "role": null, "voice": "organization", "organization": "Juspay", "organizationId": "juspay", "organizationUrl": "https://the-ai-record.com/organizations/juspay", "aiVendor": true, "work": { "title": "How We Built an AI Code Reviewer with NeuroLink and Bitbucket", "type": "post", "date": "2026-04-06", "publisher": "dev.to (NeuroLink AI organization account, Juspay)", "locator": "The Architecture Decision", "url": "https://dev.to/neurolink/how-we-built-an-ai-code-reviewer-with-neurolink-and-bitbucket-3io9" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "internal-tools", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/juspay-off-the-shelf-ai-code-review-tools-generic-checks-haskell-2026", "json": "https://the-ai-record.com/api/v1/passages/juspay-off-the-shelf-ai-code-review-tools-generic-checks-haskell-2026.json" }, { "id": "wirelesscar-field-experiment-two-llm-assisted-code-review-2025", "text": "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.", "context": null, "speaker": "Fannar Steinn Aðalsteinsson", "speakerId": "fannar-steinn-adalsteinsson", "speakerUrl": "https://the-ai-record.com/people/fannar-steinn-adalsteinsson", "speakerLevel": "practitioner", "role": "Author, affiliated with WirelessCar Sweden AB and Chalmers University of Technology", "voice": "person", "organization": "WirelessCar", "organizationId": "wirelesscar", "organizationUrl": "https://the-ai-record.com/organizations/wirelesscar", "aiVendor": false, "work": { "title": "Rethinking Code Review Workflows with LLM Assistance: An Empirical Study", "type": "paper", "date": "2025-05-22", "publisher": "arXiv", "locator": "Abstract", "url": "https://arxiv.org/abs/2505.16339" }, "grade": "B", "gradeMeaning": "The organization's own post, paper or talk.", "verification": { "status": "verified", "meaning": "An independent second reader re-fetched the source and confirmed the wording.", "reviewedOn": "2026-11-04" }, "topics": [ "code-review", "autonomy-and-oversight", "ai-review-quality" ], "functions": [ "engineering" ], "figures": [], "url": "https://the-ai-record.com/passages/wirelesscar-field-experiment-two-llm-assisted-code-review-2025", "json": "https://the-ai-record.com/api/v1/passages/wirelesscar-field-experiment-two-llm-assisted-code-review-2025.json" } ], "notice": [ "Every passage is reproduced word for word from the cited source.", "Check the verification status: \"reported\" means not yet confirmed by an independent second reader.", "Organizations flagged aiVendor sell AI tools or services and have a reason to publish flattering numbers.", "An empty result means the archive holds nothing on it, not that nobody has said anything." ] }