Anthropic filled a packed London room at its two-day Code with Claude event and found almost half the attendees had shipped a pull request that was completely written by Claude.
Those responses underlined how quickly the company, the developer of the Claude family of AI models, has moved from early code generation to tooling that aims to automate testing and fixes after updates to Claude 4.6 and 4.7 in February and April.
"The default isn’t 'I’m going to prompt Claude'; the default is now 'I’m going to have Claude prompt itself,'" Boris Cherny, who heads Claude Code, said in the opening keynote.
Anthropic highlighted a feature called dreaming, announced two weeks ago, in which Claude Code agents write and consolidate notes so later agents can learn from past errors and get up to speed on a code base.
Speakers and workshops showed how companies including Spotify and Delivery Hero and start-ups such as Lovable, Base44 and Monday.com have reorganised teams around Claude Code.
Outside the conference, developers on forums including Reddit and Hacker News complained that managers press for productivity while generated code increases review work, weakens hands-on skills and can introduce vulnerabilities.
Anthropic engineers said established software development best practices still apply, that Claude is around as capable as a midlevel engineer today, and that the company’s goal is to have agents that progressively test, tweak and reduce human oversight.
