I end every AI session with two questions

One is from Sam Altman, the other Claude suggested and it works extremely well. The first question I ask: What are you least confident about right now. The AI will list like 6 to 7 things that it didn’t properly investigate. I would say one out of four times one of the items is a huge deal and you’re shocked that the AI even took action without understanding this first, then just have the AI investigate thoroughly exhaustively finding the root cause of each issue or just understanding each point. The second