Capabilities Can't See Your Agent's Objective(jlmr.dev)
jlmr.dev
Capabilities Can't See Your Agent's Objective
https://jlmr.dev/posts/capabilities-cant-see-your-agents-objective/
1 comments
Reconciling intent has a bootstrap problem: it's inferred from the same model you're constraining, so it rationalizes. Side-effect gates — spend, irreversible writes — can't be talked around.