> let an LLM note all of the possible questions that you can answer

What does this even mean? At what point do you know you have all of them?

Humans are quite ingenious coming up with new, unique questions in my observation, whereas LLMs have a hard time replicating those efficiently.