Claude Code just fired its human supervisor
Starting August 14, Anthropic is flipping on 'auto mode' by default for Pro, Max, and Team users of Claude Code, meaning the AI will barrel ahead on coding tasks without pausing for your approval, unless it deems an action irreversible or destructive. Anthropic says this is actually safer: in testing, auto mode caught 89% of harmful actions versus just 13.6% for humans, who apparently rubber stamp 97% of permission prompts anyway. Claude Code's own team says they can't imagine going back.
Turns out the biggest security risk to your codebase was always the human clicking 'approve' without reading.