Anthropic Traces Six Weeks of Claude Code Quality Complaints to Three Overlapping Product Changes
Six weeks of Claude Code complaints traced to three overlapping bugs—not a model problem, a product one. Here's what broke.
Why it matters
Anthropic's postmortem reveals how product-layer changes (reasoning effort, caching, system prompt limits) can degrade user experience without touching the underlying model. Critical for teams relying on Claude Code for production workflows—and a blueprint for how to diagnose quality regressions at scale.
The key facts
6 to knowSix-week quality complaint window
Three root causes: reasoning effort downgrade, caching bug that erased model thinking, system prompt verbosity limit
3% quality drop from prompt limit alone
API and model weights were unaffected
All issues resolved April 20, 2026
Published postmortem (transparency signal)
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Anthropic published a postmortem tracing six weeks of Claude Code quality complaints to three overlapping product-layer changes: a reasoning effort downgrade, a caching bug that progressively erased the model's own thinking, and a system prompt verbosity limit that caused a 3% quality drop. The API…