I think it would be fun to solve multiple Zork games by feeding them all the exact same input. I'm thinking you'd need to build a tree of all possible inputs and responses from each game, then try to find a path through the tree that completes each game. It would interesting to see how you'd probably have to put the player in one game into a 'safe' room for a time so specific input can be given to another game that moves it forward (while just turning into noise/junk in the other game). Would probably need some pretty sophisticated heuristics to help trim down the massive search space of possible commands.