This is really funny to watch and see what the LLMs are thinking. This makes me think how they would perform against a custom ML model trained with RL, e.g. https://ai.meta.com/blog/rebel-a-general-game-playing-ai-bot...