>>2083998/thread
I participated in Patrician 4 playtests and you DON'T want AI that's either competent or proactive, unless the goal is to simply defeat the player and do so in a very quick order. And that's just a simple trading game. Machine by design is immensely more efficient at churning data and never gets distracted.
People who bitch about "good AI" usually have no idea what it means in practice when the computer isn't playing under artificial handicaps, but is using all the resources effectively and focuses on winning moves and only winning moves. Not even hyper-autists play like that, while machine can. Which will always result in the same outcome: AI beating your ass, so it will be accused of cheating. Especially when it is not cheating or even can't cheat to begin with.
Another aspect of bitching about AI are Paradrones, who are so used to the vanilla AI intentional idiocy, when you throw at them custom AI personalities (possible to be simply modded in), they get instantly overwhelmed, because AI no longer has a glass ceiling limiting its reach. Another good example, and Sid-related, is AI in SMAC/SMAX. Flip around few basic patterns and brace yourself for hyper-competent AI that won't need artificial resources to own your ass, because it will simply stop doing useless terraforming and plant its bases in far more efficient way, ALWAYS achieving maximum efficiency, and always needing for that the 5 seconds the turn cycles, not a conscious, 5 minute effort a human needs to plan and execute.