His main thesis is that players constantly demand stronger AI (that doesn’t cheat) but when they try it they hate it. The issue is that strong AI doesn’t role-play like an actual historical leader, it plays like a “gamer” who will stop at nothing to win.
That is, strong AI opponents treat Civ like a game of poker and they’ll use every possible means of defeating you. They’re not reliable allies or trading partners, they’re bluffing, duplicitous liars.
Human players who play against such AIs report a very negative experience. Many of the diplomacy functions in the game become rather useless against such an untrustworthy AI, and the whole situation devolves into something more akin to “turn-based Warcraft” rather than Civilization.
Discworld is amazing but not really a great setting for RPGs. The world is just too zany and hodge-podge. Everything I know about fantasy RPG fans tells me that they demand a “serious, rules-based” world.
There was a Discworld point and click adventure game though. The classic roguelike NetHack also has a ton of references to Discworld and a lot of humour and weirdness in general, though that also happens to be one of the things it gets criticized for the most. A Discworld RPG (which is at all faithful to the setting) would basically be NetHack on steroids.