It looked to me like PrudentBot could be done in depth 1 - “(proof that X(PB) cooperates) AND (proof that X(DB) defects)”. https://www.lesswrong.com/posts/iQWk5jYeDg5ACCmpx/robust-cooperation-in-the-prisoner-s-dilemma (and the paper that’s about) tells me that’s wrong, if you only have the one layer of provability then PB doesn’t cooperate with itself. It also had an introduction to the math here that I found helpful.
This is great.
It looked to me like PrudentBot could be done in depth 1 - “(proof that X(PB) cooperates) AND (proof that X(DB) defects)”. https://www.lesswrong.com/posts/iQWk5jYeDg5ACCmpx/robust-cooperation-in-the-prisoner-s-dilemma (and the paper that’s about) tells me that’s wrong, if you only have the one layer of provability then PB doesn’t cooperate with itself. It also had an introduction to the math here that I found helpful.