Anthropic publishes April 23 postmortem on Claude Code quality issues
Anthropic has published a detailed postmortem explaining recent Claude Code quality issues that affected some users over the past month.
The company identified three separate changes that impacted Claude Code, the Claude Agent SDK, and Claude Cowork (the API was not affected):
- March 4: Changed Claude Code’s default reasoning effort from high to medium to reduce latency, but this was the wrong tradeoff. Reverted April 7.
- March 26: A caching optimization bug that cleared Claude’s thinking history every turn after idle sessions, making Claude seem forgetful and repetitive. Fixed April 10.
- April 16: A system prompt change to reduce verbosity that hurt coding quality. Reverted April 20.
All issues have been resolved as of April 20 (v2.1.116), and Anthropic is resetting usage limits for all subscribers.
The postmortem is a pretty decent write-up that transparently explains what went wrong, how it was fixed, and what Anthropic will do differently going forward.