@tekk It's not impressive if you hand-code the strategy. The difference here is it's a single "learning algorithm" that is developing the strategy itself. I don't know of anyone having done that for Mario.