"Mean people fail" seems like a reasonably testable hypothesis. Paul could have his wife meet a random sample of startup founders in as double-blind a manner a

"Mean people fail" seems like a reasonably testable hypothesis. Paul could have his wife meet a random sample of startup founders in as double-blind a manner as possible (not just at social events or where they are attempting to curry favor with YC and mixed in with plenty of non-founders) then track the results of them. I am not optimistic that the results would be what he'd like, but I'm always willing to consider new evidence.