This is my first post here, so I wish first of all to thank Frogboy and all the people from SD for their dedication to the gamers....
Unfortunately I should agree with the most of the comments here, the AI is definetively too passive, at least for a 40 y.o. Grognard
I managed to "OVER-leading" my first game (10 times stronger than any other AI "intelligent-level" opponents in all realms after a few years) because I red and applied a few "begginers-guides" found here, at wiki and at others forums.
So how to improve it ?
1) Create a sticky thread or even better a subforum where people can comment and suggest new improvement for AI algorithms (this works fairly well in Paradox games community)
2) Put us at work !!
For instance ask to skilled players (the ones leading the Metaverse ladder maybe) to write the "template building list" for the Race they know the best, this "template" should differentiate among PQ (PQ<6 has one, PQ<12 another, etc), Map conditions (dimensions, planetary abbundancy, numbers of opponents, etc) and among Victory conditions.
Then ask them to write similar "template tecnology research" which differs according to Map conditions (bigger maps = more sensor and engine) again and Victory conditions.
Add an adjustable Aggression setting which you already mentioned and adjust this according to your internal and/or beta testing....et voilà le tour c'ést joué !!
EDIT: it seems that map dimensions and starting abbundance of resources are the faremost "peacekeeping" factor which you should take care of.
Have a nice day and sorry for my English. Amadeus