Assume AI develops to the point where you can just tell it to make a game with your parameters or you're a billionaire who can self-fund a dev team without having to bend the knee to anyone. No limits or restrictions on what game you can make. What do you make?
I think I would do a GTA/Saint's Row style game, but with themes based on Death Wish. You've got a city with a gay hippie district, then various portions of the city are black neighborhoods, jeets, mohammedans, etc. You play as a White guy who has some tragedy that sends him into a vigilante quest Death Wish style where you do different missions in those neighborhoods, and as you play, you slowly begin to drive them out and gentrify the neighborhoods. Disrupt enough black gang activity in their portion of the map, burn down some drug houses, etc and they start leaving and White families move in and clean it up. Burn down enough Mosques and take down terror cells in the mohammedan section, and they are driving out and Whites move back in and so on.
Other White guys start joining your cause as well as you progress and turn your crusade into a movement. Could start out with the police being an antagonist too, as they would be, but as you start taking territory, maybe work in a political angle that sort of replaces a diverse city council with more and more White men as you play and take ground, and you see more White guys as cops who start being less hostile to your activities and start looking the other way. Missions where you have to rescue kids from the mohammedans or faggots, some missions outside of the city where you take out Chicom owned farmland where they have weapons and drones in secret storage and big drug farms. Hell, maybe there's a Little Tokyo or Little Koreatown section of the city that is neutral to you if you wanna go that way. Some kind of communist trained "community organizer" type mayor as the Big Bad you have to face near the end, in the heart of the Antifa held faggot downtown district to cap it off.
Obviously no company on earth would spend 5 seconds developing such a game, but if you could do it yourself and just release it out into the wild via torrents or something, I bet it would get some traction.
RE: the the theme of your game: when ME: Andromeda was in teaser mode after 2012 I wished for the game I just knew a creatively controlled Triple A couldn't make. An actual open ended, but narrative driven, colony simulator. Multiple options for coexisting with or exterminating native populations, with intriguing consequences. ME3 was a letdown, too, because they did not make branching narratives based off of player choices of previous game; think Witcher 2.
My dream game for 2 decades has been filling in the short comings of Elder Scrolls (3 and 4, actual emergent RPG mechanics) and Total War, and mixing their scales together. It's infeasible for one person to tackle all of that, but I have been working towards glaring deficits in the Total War formula.
One is tangible logistics. There's a blogpost somewhere on the web about the poor economic fundamentals of the series. There's also a lacking model of supply lines and terrain on the campaign maps. At best there are tabletop game tier abstractions and handwaves. Real warfare is centered around access to resources, which classic base building RTS sadly do better. Real world nations tend to form borders around natural terrain such as mountains and rivers. In the case of empires, ability to rapidly travel and communicate across a distance. Grand Strategy genre should explore this phenomenon. Not complacent developers thinking "funny, inconsequentially formed lines on the map after 200 turns!"
2nd is a robust application of social science into the philosophy and composition of groups. In the free expansion Empire Divided, there is a desperate need for a full system deciding if you can convert enemy Roman soldiers and cities over to your control. How many decades their culture has diverged, proximity/isolation, and your personal power are relevant reasons that the player can expand or reclaim territory. Then we have political factions; what makes someone a Julii or Populeres? Are they a family member or household retainer, a leading politician or disgruntled peasant to be swayed? Do factions merge in face of a common problem, then split when the enemy is conquered and the population pushes capacity?
These questions needing real answers every step of the campaign map game, not disconnected gimmick mini games. The data structures are so poorly architected that in Atilla/Charlemagne (I think Med2, too) when you marry off a daughter, that daughter is deleted and another woman is generated. This primitiveness does not allow convincing political intrigue to occur.
Again like with logistics, politics is a pillar of real warfare. Not single entity legendary lord with absurd numbers face tanks 5000 Saurus warriors complete with jank animations. My goal is to work out the right maths that can satisfy 'this German confederation has centuries of West Roman and other influences, fled the Huns through a dissolved Roman Empire to form a uniquely identified kingdom in Spain', or 'this crusader kingdom is recognizably White Catholic with native Semetic civilians, distinct from the Nubians who just got bussed 1000 km to be next door.
Edit: Economic threads from Reddit. If there's another blogpost, it's buried deep in Google. https://www.reddit.com/r/badeconomics/comments/em3l26/victoria_2s_flawed_economic_model_and_the_good/
https://www.reddit.com/r/Games/comments/4bmf7m/lets_talk_about_ingame_economies_and_why_they/
https://www.reddit.com/r/truegaming/comments/9qymie/why_do_many_games_with_an_economy_miss_with/
Thanks for reminding me how much i dislike the shallowness of contemporary grand strategies.