Still waiting for someone to demonstrate a large language model that can sustain a conversation about its own limitations without crashing into absurdity or obvious contradictions. Transformers are powerful but let's not pretend they're self-aware just yet.