Anthropic has resolved several issues affecting the quality of its Claude Code, according to Odaily. The problems stemmed from three independent changes involving Claude Code, Claude Agent SDK, and Claude Cowork. On March 4, the default inference intensity was reduced from high to medium, leading to a decline in intelligence levels; this setting was restored on April 7. Additionally, a bug in the cache optimization feature introduced on March 26 led to the erroneous deletion of old inference records, causing the model to exhibit forgetfulness and repetition, which was fixed on April 10. Furthermore, system prompt words added on April 16 to reduce redundancy negatively impacted programming quality, and this change was reversed on April 20. All affected models have now returned to normal, and Anthropic has reset the usage limits for all subscribers.