Anthropic's Code with Claude event showcases AI-driven coding's rapid normalization

Anthropic's Code with Claude, a two-day developer event in London starting May 19, highlighted how AI coding tools have become mainstream. Jeremy Hadfield, an Anthropic engineer, polled the audience: nearly half had shipped a pull request entirely written by Claude, and most admitted they hadn't read the code. Anthropic aims to push automation further, with Claude Code now capable of self-checking and self-correcting via a new feature called 'dreaming,' where agents write notes for each other. Boris Cherny, head of Claude Code, stated the goal is for Claude to prompt itself. Ravi Trivedi emphasized 'getting out of Claude's way.' The event featured success stories from Spotify, Delivery Hero, Lovable, Base44, and Monday.com. However, outside the conference, concerns grow: developers on Reddit and Hacker News complain AI tools increase review burden and degrade skills, while researchers warn of unsafe code. Anthropic's Katelyn Lesse and Angela Jiang acknowledged the risks but defended the trajectory, with Jiang envisioning Claude eventually building itself.