In theory, if AI was really crazy good, which I don't think will happen anytime soon or at all, it would be able to let you play the game, without any code being written other than the AI. Describe your game, connect inputs, and the model will make up some video and audio output through generative diffusion models. The challenge might be that it's way easier to do if you can let the AI have time to process each request, as opposed to having to react to many inputs per second.
It might require the entire power output of a medium-sized country to play a game, though.
