Claude writes 90% of Claude Code's code while revert rate fell over 40%
ClaudeDevs · x · 2026-10-17
Anthropic's Claude Code team (Addy Osmani et al.) details their code review evolution: Claude now writes about 90% of Claude Code's own code, yet from June to September merged PRs more than doubled while the revert share fell by over 40%.
Key points:
- Anthropic's monorepo went from 2,500 changes/day in May to 30,000+ in September, making line-by-line human review impractical
- They replaced "did a human read the code?" with a sharper question: what could break, and which check—automated or human—catches it
- Most reading moved to machines; human time and attention were optimized, with heavy investment in fast, easy undo
- The second job of review—shared codebase understanding—is addressed separately
The post explains the workflow of minimizing human review time without raising reverts or incidents, and how other teams can adopt it.
Related event: Claude Now Writes 90% of Claude Code, Anthropic Reveals Review Process(2 posts)→
More from coding & agent
- Backend migration from Python to Go now takes one engineer just one week — rakyll · 2026-10-17
- Audio founder says Claude one-shot decompiled rival algorithms, moat gone — ChrisGPT · 2026-10-17
- Google's playbook for agent evals: four layers of checks beyond the vibe check — rseroter · 2026-10-17
- AgentForesight: a 7B online auditor that catches multi-agent failures mid-run — jiqizhixin · 2026-10-17
- Qwen3.8-27B shows a weird write-delete loop in /tmp before every web research task — Training-Respect8066 · 2026-10-17
- Dev gives his agent an alarm tool to wake him during overnight /goal runs — zack_overflow · 2026-10-17