Claude Code handles daily maintenance at Anthropic, 46% merge rate

Anthropic is testing Claude Code for daily maintenance of its own software, including crash fuzzing and dead-code removal, according to The Decoder. Over a few weeks the AI created 388 pull requests, with 46% merged after human review.

Why it matters: A 46% human-approved merge rate on autonomously generated maintenance PRs is a concrete, measurable data point for agentic coding reliability, moving the conversation from benchmark scores to actual production usage and signaling Anthropic's confidence in deploying its own agents on its own codebase.