💭 I wonder if you could train a sufficiently large cross game muzero such that it could more quickly learn new games by transferring share concepts like resource management, game theory, value of information, etc.