An update on recent Claude Code quality reports

Anthropic has identified and resolved three separate issues that degraded Claude Code quality over the past month. First, the default reasoning effort was downgraded from 'high' to 'medium' on March 4 to reduce latency, but was reverted on April 7 after user complaints about reduced intelligence. Second, a caching optimization shipped March 26 contained a bug that continuously dropped prior reasoning history each turn instead of just once after idle sessions, causing forgetfulness and faster usage limit drain — fixed April 10. Third, a system prompt instruction added April 16 to limit response verbosity caused a measurable drop in coding quality and was reverted April 20. All fixes are live as of v2.1.116. Going forward, Anthropic plans broader eval suites for system prompt changes, tighter controls, gradual rollouts, and improved internal dogfooding on public builds. Usage limits are being reset for all subscribers.