A computer won the Arimaa challenge last year.

https://en.wikipedia.org/wiki/Arimaa

I guess the only thing left is to design a game where you can change the rules of the game as a turn.

Might not work for long. There are already contests for best generic solutions[0] and it seems like quite popular topic in machine learning.

[0] https://en.wikipedia.org/wiki/General_game_playing

Congratulations! You reinvented Nomic.

https://en.m.wikipedia.org/wiki/Nomic

Calvinball.

Well, that's that then :| Or maybe this will spur the human players to improve :)

It was designed to be hard to win with the algorithms known to researchers back then. I cannot tell if it has something substantially harder than Go has. High branching factor, deep tree and positions that are hard to evaluate with a simple heuristic are present in Go. Does Arimaa have something else up its sleeve?

I think one of the main things was that you can move up to four pieces (or one piece four times, or two pieces two times, etc) in each turn.

Diplomacy has all pieces move each move, and multiple players you have to model (and interact with). That would be an interesting challenge.