Is Anthropic 'nerfing' Claude? Users increasingly report performance degradation as leaders push back

Users accuse Anthropic of degrading Claude Opus 4.6 and Claude Code, citing worse reasoning, more task abandonment, and higher token waste. An AMD director's analysis of 6,852 session files showed performance decline since February. Anthropic denies degradation, attributing changes to adaptive thinking defaults and effort level adjustments. Benchmark claims of a 67% drop were disputed as comparing different test scopes. Prompt caching changes from 1-hour to 5-minute TTL were confirmed but framed as optimization, not degradation. A trust gap persists between Anthropic and power users experiencing real friction.