Uh yeah I mean I know what you’re talking about - he does defer to the interviewee (and they are often wrong as well, like in the Elon interview).

Moreso I’m referring to the LessWrong/Reddit/HN/Twitter internet-lore format which he borrows from and leans into. Many of his questions are deep on the surface but ultimately more about sensationalism (even if that means a bad faith read of the problem).

Also, at times I suspect interviewees pre-negotiate certain things for him to see in a positive manner. Like in this last episode, thanking Sam for open sourcing GPT eval - ignoring that GPT-4’s basic architecture description wasn’t even released; a big “fuck you” to researchers hoping to do any sort of replication or peer review.