Some reflections on the most recent build:
(Standard/Prince/Tectonics/city flipping after conquest/must kill all units to conquer)
1.) I am noticing some good counterattacking with artillery and defensive units when the AI has an advantage. In my most recent game, the Zulus, despite being completely on the defensive, nearly destroyed a force of 9 tanks (I think they took down six and severely damaged the survivors) with a combination of artillery, infantry, and SAM Inf.
2.) I am still noticing, however, the AI blindly focusing on building some things that it shouldn't in an emergency. Despite my army bearing down on one of their last cities, the Zulus were still building the Apollo Program there.
3.) I also noticed an unusually large population of destroyers parked in that city, despite the fact that I was threatening to take it -- and also despite the fact that I was using my fleet and occasionally shipping troops back and forth. Now there HAD been some fast and furious naval combat at the start of the war, and my aircraft had dinged most of their ships even if I couldn't engage them in the water. It's possible that they were repairing, and it's also possible that, given my air superiority, the Zulus didn't want to sortie their fleet unless a ship moved within easy range of port, to prevent air attacks, but it seems pretty foolish to wait in port when they might be able to do some damage by going out on the high seas.
4.) I noticed that the AI, where appropriate, is aggressively building ships, and using them in a "picket line", off their coast, with typically two to three ships per picket. Very nicely implemented.
5.) I am not sure if the AI is handling play under the "city flipping after conquest" well. I noticed that in one game, Stalin was at war with Mansa. Although Stalin had reduced Mansa to one city, he wasn't making any progress. It looked as though he placed all of his offensive units in the most recently-taken city (presumably to prevent a culture flip,) and was not attempting to take Mansa's last city (which he probably could've done given his overall strength). I think the AI needs to learn that, in this circumstance, it may be wiser to press on the offensive to finish off the enemy. (Disclaimer: Mansa never capitulated, so it's possible he was able to chew up several attacks from Stalin while holed up in his capital. I did not look at this using Worldbuilder.)