You could try to fake an AI, by have a precalculated "good move" list for scenarios. (Don't know how hard/feasible that would be)
Even on a 9x9 board there is a mind bogglingly large number of variations that a board can be in.
https://senseis.xmp.net/?NumberOfPossibleGoGames
Even if you tried to narrow that down by eliminating mirrored/rotated positions it would be far too large.
Literature on Go AIs is actually pretty sparse.
I could have it look for a couple of josekis though 🤔 All I’ve thought of is to find the group with the least liberties and surround it. It’d be a very, very simple bot to say the least 😂
Stumbled across Fuego and specifically the simple players, which all work based on one move analysis which meant it was feasible to implement on a Playdate, so that’s I did!
It’s very bad, but I kind of like crushing it after losing in an online game against a real human haha.