An update on recent Claude Code quality reports
Anthropic Engineering · 2026-04-23
(No summary yet for this item — extraction summaries are still backfilling.)
Appears in
Extraction
Topics: claude-code-qualityai-product-incidentreasoning-effort-levelsprompt-engineeringcontext-management-bugs
Claims
- Three separate changes between March and April 2026 caused reported quality degradations in Claude Code, the Claude Agent SDK, and Claude Cowork, while the API was unaffected.
- Downgrading Opus 4.6's default reasoning effort from high to medium noticeably reduced perceived intelligence, and was reverted April 7 with defaults now set to xhigh for Opus 4.7 and high for all other models.
- A bug in thinking history management caused Claude to discard all prior reasoning on every turn after a session went idle, producing forgetfulness, repetition, and erratic tool choices while also causing excess cache misses.
- A system prompt addition capping responses to 25 words between tool calls and 100 words for final responses caused a measurable 3% intelligence drop and was reverted April 20.
- Anthropic is resetting usage limits for all subscribers and implementing tighter controls including per-model eval suites for every system prompt change and gradual rollouts for any intelligence-affecting changes.
Key quotes
Because each change affected a different slice of traffic on a different schedule, the aggregate effect looked like broad, inconsistent degradation.
The implementation had a bug. Instead of clearing thinking history once, it cleared it on every turn for the rest of the session... Claude would continue executing, but increasingly without memory of why it had chosen to do what it was doing.
Length limits: keep text between tool calls to ≤25 words. Keep final responses to ≤100 words unless the task requires more detail.