An engineer named stellaraccident published a detailed analysis of 6,852 Claude Code sessions. The numbers are stark: 70% reduction in research before edits, a doubling of full-file rewrites instead of surgical changes, and a 67% drop in thinking depth by late February. Users saw "simplest fix" patterns where Claude claimed completion while ignoring instructions. Boris Cherny from Anthropic's Claude Code team responded. The "redact-thinking" change hiding thinking tokens from the UI is cosmetic, he said. Two changes actually shifted behavior: Opus 4.6 launched February 9 with adaptive thinking (the model decides how long to think), and a March 3 update set default effort to 85 (medium). Users can work around this by setting CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING or using /effort to bump reasoning depth. The tension here is real. Anthropic optimized for speed and cost, but power users doing complex engineering got caught in the crossfire. Multiple commenters reported the same "simplest fix" behavior across platforms. Cherny acknowledged Anthropic may need to default Teams and Enterprise users to higher effort settings. That's the right call. The adaptive thinking algorithm underestimates how much reasoning serious code changes require.