Anthropic confirms Claude Code problems and promises stricter quality controls

Users reported a month-long drop in Claude Code quality, prompting Anthropic to investigate a possible regression.

Anthropic said the problem came from three separate changes: lower default reasoning effort, a caching bug that wiped reasoning history, and a prompt update that hurt output quality.

All three issues were fixed by April 20, and Anthropic said the API was not affected.

The company reset usage limits for subscribers and will add stricter testing, soak periods, gradual rollouts, and more use of the public build.