I'm still thinking about the flavored civics thing ... I checked it out but I'm not 100% sold on it. Here's how it works ... the AI assigns a value to each civic in a particular category using the long, complicated formula in CvPlayerAI::AI_civicValue. Typical values for a civic the AI might pick vary a lot and depend highly on the situation and the number of cities the AI has, but values for a sizable empire in mid to late game are between 100 and 500 for good choices.
What Grave did was add a small piece to this civic value calculation based on flavors ... a civ with a strong Military flavor could be made more likely to run police state for example. Now, it could be somewhat cool in game to have AIs roleplaying a bit more with their civic selections and have Stalin more likely to run a Police State etc but for this flavors feature to have any effect it would have to cause Stalin to pick Police State over something else even though based purely on effects on cities etc AI_civicValue had determined that something else was better. Clearly this could be detrimental to the AI if the flavor value was too strong and cause the Stalin to pick a sub-optimal civic for his current situation.
What's the right middle ground? It would certainly be possible to add a fairly weak effect for flavor where the AI would only go one way based on flavor alone if two civics were otherwise nearly equal ... but how strong should the flavor effect be to be relevant but only tip close calculations? Or to look at it another way, what's the margin of error for AI_civicValue?
Given the range of values produced by the function, flavor would want to add/subtract a percentage from the civic value I think instead of just add a flat value as Grave has it. One possible way to pick a percentage is to go based on how much better a civic has to be for the AI to consider switching ... currently, a new civic has to be 6/5 times as good as the current if the AI will suffer anarchy, 16/15 if it will not. So somewhere between 20% and 7% could make a reasonable range ... but then it will be a fairly small effect and you won't really see it much in game, so is it worth adding?
As for a new version, I can put out a small update with a few small tweaks plus Solver's latest at the end of the week. I've been checking the Better BTS AI changes to make sure they play well since I'd be incorporating that as well. It will be a while before the next real Revolution update, but it will be worth the wait I think.