This is a special post for quick takes by hereforYud. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
I don't have a link, but I think the basic idea is that sufficiently smart entities should be able to calculate the probabilities and costs of victory, and negotiate a better outcome.
Something like: if you and me are fighting for a bag of coins, and we both agree that my probability to win the bag is 60% and your probability is 40%, it would make more sense to stop fighting and split the bad 60:40. But perhaps you have a hidden knife I don't know about, because you planned to use it only if I am clearly winning the fight. If we both are smart enough, we might negotiate something like you saying "there is something you don't know about, can you credibly promise that if I disclose that information to you, you will take it into account when calculating the probabilities, rather than abusing the fact that now you know the information" and then we might split it 20:80 or something.
In summary: knowing the probabilities and costs, and having a way to precommit to handling previously unknown information in a fair way (to agree on the estimates of probabilities and costs had the information remained unknown).
There are a lot of scenarios where that makes sense, but the breadth of situations we're talking about is ridiculously wide. If 2 ASIs are made somewhat simultaneously 6 months from now, wouldn't the first one to exist take a lot of precautions against the emergence of a 2nd ASI and prepare to end it immediately? It seems like the resources needed to calculate through peace and then continuously check that it's being maintained from now to forever could very plausibly be greater than the resources needed to just end the other mind.
I definitely agree that ASI could keep the peace longer than humans have tended to, but even then, there could be weird motives in the ASI that encourage violence against things like it.