I'm not sure if hammering an LLM until it agrees with you is the best way to get to the truth.

> with edge cases and hypotheticals

not

> conclusions I want to see

The point is to be adversarial with your own ideas, not the opposite thing.