Bazillion-dollar idea: When an AI realizes that it made a mistake mid-response, it should have a mechanism to basically re-invoke itself on the same input and leave notes to its future self on what it got wrong

0 likes 0 replies