That's not completely right, but close enough to be dangerous. Indeed, quite a few researchers around LessWrong, MIRI, FHI etc. agree that the right way to form

That's not completely right, but close enough to be dangerous. Indeed, quite a few researchers around LessWrong, MIRI, FHI etc. agree that the right way to formalize decision-making should use the timeless view that you advocate. Decisions and precommitments are provable consequences of an agent's source code, agents with the same source code can use that fact to coordinate, and so on. I've done a lot of work on this, so feel free to ask :-)