In the normal Ancient Savo game play every map tile is simulated in detail; soil fertility and moisture, the growth of grass and tree leaves, and how all of this is affected by the local geography (ie, the tile being in a bottom of a valley, or on a slope of a hill, and if that slope if facing north or south). And once the player send their adult offspring to start homesteads of their own, there needs to be an AI to run the other family units while the player is controlling their chosen family unit. For a few AI-simulated families it probably would be possible to handle their respective map regions in full detail, but I suspect that somewhere there is a limit when too much simulated map tiles would start to impair the performance, making the game laggy. And, generally speaking the idea is simple; for the AI families instead of a fully detailed map there needs to be some sort of summary in less detail; it is enough if the AI knows "we have this many tiles of field, and that many tiles of lakes with limonite at the lake bottom", and then there just needs to be some simulation maintaining that summarized map, how the AI actions affect the number of field tiles, and how harvested tiles yield food and so on. But, to figure out exact details of how this map summary should and would work, that isn't a trivial task. It is a coding task I greatly enjoy. Like, for a task like path finding I can just google and find a bunch of already proven approaches (which is fine, too). Here I'm on my own, and I need to exercise my brain to find answers to my coding questions.