Note that if the AI is weak, this could result in a more balanced game between weaker and stronger human players.
Do you think this chess variant would be compelling?
https://en.wikipedia.org/wiki/Monte_Carlo_tree_search
as an evaluation function, better yet it could critique me and help me learn to play better.
As a player, it depends on how good the AI is at chess. If it’s better than me, I’ll have no idea what’s going on. If it’s worse than me, that’ll leave me feeling frustrated. If it’s as strong as me, what’s the point? So… it doesn’t sound compelling either way.