GPT recently retook Caplan's midterm and got an A: https://betonit.substack.com/p/gpt-retakes-my-midterm-and-gets-an

This is a fair test if GPT wasn't explicitly trained on the midterm (e.g. by reading Caplan's last post), and not otherwise. This market resolves YES if it is revealed that GPT somehow had access to material making the comparison unfair; otherwise the market resolves NO at the end of the year.