Exactly! During the process, it seemed like if there were like two GPTs self-playing to both generate the proper prompts iteratively and the other generates lik

Exactly! During the process, it seemed like if there were like two GPTs self-playing to both generate the proper prompts iteratively and the other generates like the output, all triggered by one concise command from a human - say write tests and dont stop iterating till the tests pass - basically automating the human out of the loop - could get rid of the loops fixing tests, but also take away control.