I think its a meta comment about LLMs overconfidence. Like, htf does Astra, the big frontier model, still hallucinate confidence scores like this? Reliability and interpretability are still over the horizon.