I Want to Pay Forecasters to Keep AI Scientists+Strategists Honest

I'm making an OOD bet that some of you are good at OOD bets, and at judging whether others' bets are good even with no ground truth to check them against.

We're forming a team building AIs that identify new science, technology, and the strategic ramifications that follow. The hard part isn't generating outputs; it's automatically filtering for the ones likely to be right, even when they concern wild technical domains no human or AI has seen before. That's an OOD forecasting problem in disguise. How would you do it? How would you get an AI to do it?

Half talk, half live exercise, half recruiting pitch.