As AI has improved I’ve seen wave after wave of skeptics claiming that the models can’t do things which I have personally seen them do.

This initially baffled me, but after talking with them I realized what was happening:

1. The skeptics weren’t using the latest models