TalkChess.com

This may be of interest to the chess variant people:
https://arxiv.org/abs/1905.05013
http://ludii.games/

I recall there was a similar project from stanford some years ago http://ggp.stanford.edu/. The game rules are specified at runtime through some language (prolog?) that you have to learn and choose best algorithms form. At the time MCTS was considered a better general algorithm but with the advent of neural networks, GGP bots should now be more capable i think, but training NN is not possible at runtime so maybe the rules can be modified a bit to account for that.

My engine nebiyu plays many games and variants and has NN now, but the rules are baked into the program. It would be interesting to see a GGP engine with mcts+nn.