Before each action runs, a separate classifier model reviews the conversation and decides whether the action matches what you asked for: it blocks actions that escalate beyond the task scope, target infrastructure the classifier doesn’t recognize as trusted, or appear to be driven by hostile content encountered in a file or web page. […]
Is this how the brain evolved, with piecemeal models checking others and in some nondeterministic manner preventing dangerous behaviors?
— Read on simonwillison.net/2026/Mar/24/auto-mode-for-claude-code/
Leave a Reply