Anthropic Says Claude Now Writes Over 80% of Its Own Code — and Warns the Industry Isn't Ready for What Comes Next
Anthropic disclosed that Claude generates more than 80% of the code shipped internally, while simultaneously publishing a warning about recursive self-improvement timelines. The company is simultaneously the loudest booster and the loudest alarm bell.
Anthropic made two disclosures on Friday that, taken together, paint a striking picture of a company sprinting toward a capability it publicly worries may be dangerous. As summarized by @oscarlau, Claude now generates over 80% of the code used internally at Anthropic. Separately, the company issued what multiple briefings describe as a call for lab coordination on self-improving AI, with @marcopapa99 noting that Anthropic "warns about self-improving AI" and @ainunnajib characterizing it as the company putting "RSI on the timeline."
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.