9-tsiak wasn't simulating the Demiurge, ve was simulating a generic superintelligence. But in the world of the story, because they're superintelligent, all superintelligences will reach the same conclusions about potential threats from other superintelligences, and will resolve those problems in the same way. So the simulated intelligence might as well be the Demiurge as anyone else, AND the Demiurge's irl actions uphold the ongoing acausally negotiated bargain.
Iām not sure if that makes sense (the story, not your understanding). Even if all superintelligences use the same reasoning processes, they donāt all share the same values. (If they did, they wouldnāt be competing at all.) So when 9-tsiak realizes that thereās probably an older superintelligence than it, which probably does not share its values and is therefore an enemy, what makes it choose to simulate Elder Brotherās values in particular?
It could be that 9-tsiak is simulating a āgeneric superintelligenceā, without any specified values beyond obvious ones like self-preservation, but Iām not sure if thatās even a coherent concept. Just because two agents can be intelligent without sharing values, doesnāt mean that intelligence is an algorithm with a parameter for an arbitrary āvalue functionā. (There are generic algorithms that work like this, e.g. minimax or A* search, but those get used as components of entire intelligent programs/systems where the desired goals are baked into various other design decisions. Like how, since self-driving cars have the goal of legally driving from point A to point B, they have wheels and can understand traffic signs. A self-driving paperclipper would be built differently.)
