Which is funny because the thing that LLM AI is not good at is coming up with new ideas and confirming they’re correct. It is good for making language (or other knowledge objects) based on existing input.