Anthropic: Claude wrote over 90% of its code contributions in September

Anthropic says more than 90% of September's code contributions to its codebase were written by Claude. That figure is a sharp marker of how far agentic coding has gone inside a frontier lab.

The company framed the number around a capability milestone: AI that meaningfully speeds up its own development. Anthropic was careful to say this is not the same as fully autonomous recursive self-improvement. Humans still direct, review and merge the work.

The disclosure arrives in the same week as the IPO filing. Elsewhere, Reddit threads note Anthropic now spends more on its "AI employees" than on human staff, and OpenAI and Google are pushing their own coding agents.

Skeptics point out that "contributions" can count lines or commits rather than value delivered. They also note that self-reported metrics are hard to verify. Readers should watch for whether Anthropic publishes a methodology.