An update on recent Claude Code quality reports
- Source
- anthropic.com
- Author
- Anthropic Engineering
- Date
Context
A coding assistant can become forgetful because its surrounding software stops preserving information it previously used. In an April 23, 2026 postmortem, Anthropic described an optimization intended to clear older reasoning once when a session resumed after more than an hour idle. A bug instead repeated that clearing on every subsequent turn, making Claude appear forgetful and repetitive.
The company reported fixing that bug on April 10 and said its API and layer were unaffected by the investigated quality problems. This incident illustrates how conversation management can change a user's experience of intelligence even when the underlying inference service is functioning normally.
- inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
- system prompt — The standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.
“We’ve traced these reports to three separate changes that affected Claude Code, the Claude Agent SDK, and Claude Cowork.”
“We never intentionally degrade our models, and we were able to immediately confirm that our API and inference layer were unaffected.”
“A bug caused this to keep happening every turn for the rest of the session instead of just once, which made Claude seem forgetful and repetitive.”
“Because each change affected a different slice of traffic on a different schedule, the aggregate effect looked like broad, inconsistent degradation.”
“All users now default to xhigh effort for Opus 4.7, and high effort for all other models.”
Checking sign-in…
Loading comments…