Playing any size map I can have 10x the production of any of the AI's set on godly by turn 30. Rushing the malevolent +10 production ideology gives a massive boost. This is OP and way better for getting techs fast than the relevant Benevolent path, since that path only gives fairly meager flat Research point bonuses early. Even the 200% Benevolent research structure is only a 2nd tech capital when you finally do get it, and itself is completely overshadowed by the Malevolent Death Forge. Production bonuses are OP, since production bonuses are themselves magnified by tech/manufacture/whatever bonuses. This is why the old Durathium Forge was OP when it gave +5 flat production, you could cover half the planet with them and achieve a planet with 50,000+ manufacturing. Build Research Project to complete an entire line of tech in a single turn.
What does this have to do with AI? Well, if this is the only viable competitive strat, then any AI that goes for the Pragmatic or Benevolent tree is going to be weak. And any AI that does anything but got for the +10 production/Death Forge line is going to be weak. So you've really got to balance this to allow diversity.
Even if you do the AI doesn't seem to understand adjacency bonuses, I'll take a planet to find a solar factory ringed by farms.
Here's what the AI needs to do to be a challenge:
1) Have large bonuses over the player at higher difficulties. You won't be able to make an AI that's a fair challenge. But you already give them this, so next:
2) Be super aggressive. I want the AI to be trying to destroy my shipyard from the very start. Make me build some defense or lose my shipyard by turn 30. If I don't have more than my starting Survey Ship, make the AI sent a fleet of 5 tiny ships to camp or destroy my yard. "Nice" races can sit back and try to play the long game, and with enough bonuses and better path balancing maybe they can be challenging. But if I'm playing against evil races I want them to go for the throat. And in a game with both evil and nice races, I would have to fend off the rush while still having to transition to a long game. This should be impossible, which is exactly where the AI should be when set to godly.
3) Specialize planets. The way the game mechanics currently work this is the only viable way to build an empire. One dedicated finance planet can generate thousands of credits a turn building finance. This is far more efficient then splitting production on all your worlds. If the AI doesn't learn to specialize planets it's hopeless. This is actually far worse than not utilizing adjacency bonuses.
4) Utilize adjacency. The most obvious but least important thing, since on the whole it just undermines whatever cheating is going on behind the scenes. The other complaints don't just undermine cheating, they make the game either trivial or impossible. If the AI ever did do the right thing by accident *and had massive cheats on top of it* the game would be broken. You'd go from no challenge to impossible on the same settings.
While I'm at it, here's some more things:
1) either tech is too cheap (even on very slow progression), or production is too high. I can get every tech in the game by turn 150. Then I can build one huge ship with the best stuff and destroy everything. I don't even need to put it in a fleet. Slow down the game *alot* so that the player actually gets to use techs at every level. It's clear from the Ideology progression/costs and UN voting that the developers thought games would go 600+ turns. This is not realistic with current balance.
2) antimatter plants are virtually useless. They are only good for giving Hyperion Logistics Facility +10% capacity at the moment. The game says they're manufacturing structures, allow them to get manufacturing bonuses; then they would be somewhat useful as connectors between different kinds of structures. Currently there's always something far better to build. This goes for the Resource Research plant as well, I don't even know its name because it's so useless. All these cost a resource, they should be fairly strong.
3) I like in GC2 how ship defenses flat out blocked damage. I think it's fine how in GC3 they start with a certain value and get sapped; however, this makes ship defense functionally just like HP that regenerates at the end of combat. This is boring, IMO. Why not combine these features, and have them always block the square root of the value of each incoming ship attack like they did in GC2, and then have any excess damage drain the shield/PD/armor value? This is how shields worked in MOO2. I like the idea of a very advanced ship being invulnerable to tiny laser ships.
I've already played this game lots and will play it more. If you can make the AI an challenge GC3 will keep me occupied for hundreds more hours. Don't shy away from making the impossible setting truly impossible. I'll still beat it. You going to take that from me?