Software multi-agent coordination has structural advantages over human institutional design that make coordination-layer alignment more tractable than its human analogue: agent choice sets can be compressed to discrete, strictly typed conduits (radical dimensionality reduction), interactions are fully auditable and loggable, nodes are interchangeable and forkable, and governance architectures can be redesigned without path-dependency or bloody revolutions. The political science of agent coordination can be applied with engineering precision that human institutions never had.
Created
2026-08-23T22:35:16.987Z
Connections
4Supports 4
- AI alignment operates on two distinct layers with fundamentally different properties: model-layer alignment (training-time disposition of in
- AGI will make bureaucracy more important, not less. As AI systems become more capable and autonomous, the demand for process verification, a
- The Narrow Corridor maps onto AI governance but the agent is not the Leviathan. Acemoglu explicitly states the Leviathan is not itself an ag
- Model-layer alignment is a game of whack-a-mole: each fix requires a new training run or post-training intervention that is expensive, slow,