Claude Code now runs daily maintenance on Anthropic's software with a 46 percent merge rate
Not a pilot. Claude Code filed 388 PRs on Anthropic's own codebase in weeks—46% merged without human rewrite.
Why it matters
Autonomous agents are moving from proof-of-concept to daily production work inside a frontier lab itself. A 46% merge rate on real maintenance tasks signals agents can handle multi-step software engineering workflows with human-in-the-loop review—and it's happening at scale inside the company building them.
The key facts
6 to knowClaude Code filed 388 pull requests over several weeks
46% merge rate after human review
Tasks: crash fuzzing, dead-code removal, daily maintenance
Running on Anthropic's own software stack (internal deployment)
Boris Cherny (Claude Code inventor) quotes 'early signs of life that this might be possible'
Human review gate remains in place (not fully autonomous)
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Anthropic is testing whether Claude Code can handle daily maintenance of the company's own apps, from crash fuzzing to dead-code removal. In a few weeks, the AI created 388 pull requests, and 46 percent were merged after human review. Claude Code inventor Boris Cherny sees this as "early signs of…