I've been hesitant to say much about the AI because I believe that Frogboy and the SD team have a handle on it and will get the issues resolved. I also have enough patience to wait for them to do their job. Having said that, here are some anecdotes from my current "fun" game. Immense, around 12 AI's all "Gifted", uncommon x 3, common minors, uncommon resources, rare black holes, uncommon anomalies, slow game pacing, very slow tech, no pirates, no tech trading.
First, the AI's are doing reasonably well, especially the custom ones that I gave patriotic to. By the time the colony rush was over, the majors seemed to all have about 12 -20 planets each. AI doesn't seem to prioritize the precursor resources instead of yet another Durantium mining starbase in the middle of my territory. It also seems to leave entirely too many artifact and capsule anomalies laying about. I even saw it run a survey ship right up to one I was racing it for when it stopped and took off a new direction.
Here is an example of something that seems like potential bad AI, but the reality is that it isn't. This in turn might lead around to what it really needs.
Seems the AI is good at picking off fleets(or ships) that it greatly outnumbers. If it sees a juicy target, like a Transport but with no people, sitting in range but unguarded, it will send a ship to turn it into ashes. However, it also seems to ignore that if they move into range of the bait, they will then be in range of my real ambush fleet. (Potential Bad AI)
I set up these honey pots sometimes to lure the AI across the range border so I can piece meal them. Seemingly this would be a bad move for the AI, however, IF they don't have the sensor range to see my ambush waiting then how could they know? In the context of what the AI knows, it looks like an easy kill. We would all complain if it didn't take the easy kills.
This makes sense the first couple of times, but a human player would soon catch on to being ambushed and no longer take the bait or would move with great force to take the bait or counter ambush.
As I think Brad has already indicated, he is working on the FOW issues. Most of those seem to be related to planets and resources. I also seem to recall either hearing or reading something about the AI API having a "check danger" routine that it uses (at least in the context I recall) for avoiding pirates. Assuming that the AI puts a reasonable number of sensors on their ships or learns to also build sensor boats, then one would think that "check danger" would also alert it to my pending ambush. If it's not, then that would be an area to look at.
It seems to have particular trouble with "over the horizon" type of attacks. This is understandable, but also exactly what the AI needs to do back to us. And if we get up a sensor net to see it coming, then clearly the AI needs to do the same.
I would think the AI could try to reverse the process. Instead of super guarding every transport by putting it into a fleet, hereby diminishing Fleet firepower, it could leave the transport out of the fleet but be ready with a counterattack if you take the bait.
Finally, I don't know about everyone else, but I rarely station ships in orbit to defend a world. If I can get to the threat, then I take my ships TO them rather than waiting for them to come to me. If I don't have enough movement this turn to get there, then I might orbit for a turn and then reform the fleet and attack the threat. And yes, during all of this I am check and counting hexes to see just who can get how far and by when. For sure, GC3 is much better than GC2 about not having the AI waste resources and time on building five defenders on every planet. We don't know what all is in the API. I know that I heard, and have seen, that the AI responds to the threat of invasion by stationing ships in orbit of the "threatened" worlds. My recollection and observation is that it seems to be a state machine type of system. Something like, "If invasion threat to planet, then station ships" and "If they have invasion tech, then build defenses on worlds closest to enemy".
What I noticed (by watching with long range sensor ships) is that when the threat abated the stationed ships moved off in the opposite direction of the one the threat retreated. Once that happened, I quickly attacked with very quick moving transports from well outside their sensor range and took the unguarded planet. Which brings up: No matter how many AI planets I took with the help of Biological weapons, they never developed the cure.
Which brings up my last finally, finally. As far as counter measures to invasion tactics, I should not be limited to just one. I could understand that each one isn't available to research until it's used against me or I develop the tactic myself, but if I counter Bio-Weapons, I shouldn't be prevented from building firewalls.
Regards and I'm a very happy customer at this point. Looking forward to the future.