AI Matters
Why it matters
Anthropic tests AI agent on its own software maintenance
The Decoder Claude Code now runs daily maintenance on Anthropic's software with a 46 percent merge rate
Anthropic revealed that its Claude Code agent successfully resolved routine software maintenance tasks, achieving a 46 percent merge rate for its code updates.
Why it matters
The experiment provides real-world data on how effectively autonomous AI agents can handle routine coding tasks under human supervision.
Latest brief · Aug 14, 2026