Anthropic has published a detailed postmortem explaining a six-week period of code quality complaints for Claude. The root cause: three overlapping product changes (context window expansion, instruction-tuning adjustments, and token accounting fixes) that collectively degraded reasoning performance on coding tasks without immediate visibility into the compound effect.
The transparency is a best-practice signal for enterprise customers relying on Claude for code generation. Anthropic's ability to isolate and correct the issues, then communicate the findings publicly, sets a reliability bar that matters for CTOs evaluating models for production AI engineering workflows.