80% of Anthropic's Production Code Is Now Written by Claude
In May 2026, most code merged at Anthropic was AI-authored, not human. The surprise isn't the volume — it's what didn't change: every line still ships through review, tests, and a merge gate.
NeuroX AI · June 5, 2026
In May 2026, more than 80% of the code merged into Anthropic's production codebase was authored by Claude, not by humans. The company building the model now ships mostly AI-written code — straight into the systems that serve every Claude request.
This isn't a demo metric. Boris Cherny, who heads Claude Code, hasn't hand-edited a line since November 2025 — he reviews and ships 10–30 pull requests a day, all AI-authored. Anthropic reports engineering output up 200% per engineer against its prior baseline.
The capability caught up fast. On complex, open-ended problems with no clear spec, Claude's success rate hit 76% in May — a 50-point jump in six months. Across the broader ecosystem, 4% of all public GitHub commits now come from Claude Code, with Anthropic projecting 20% by year-end.
Here's the part teams miss: 80% authored does not mean 80% unsupervised. Every one of those PRs still passes human review, a test suite, and a merge gate. The bottleneck moved from typing code to verifying it. The teams shipping fastest are the ones who rebuilt their review and CI layer to keep pace with an agent that writes faster than anyone can read.
Authoring is solved. The discipline around it is the new moat.