Coasean Bargaining at Scale (blog.cosmos-institute.org)

Seb Krier of DeepMind argues that instead of defaulting to heavy, top‑down regulation for AGI, we should consider how AGI agents could enable "Coasean bargaining at scale": personalized fiduciary agents that slashthe transaction costs (discovery, negotiation, enforcement) that historically forced centralized planners to resolve externalities. By creating an ecology of specialized, interoperable agents — a “virtual agent economy” — individuals’ agents could encode fine‑grained preferences (when noise is tolerable, how much to pay for cleaner air, emergency priorities) and negotiate instantly with others’ agents, producing correlated equilibria and hyper‑granular contracts that make many local tradeoffs mutually beneficial without blanket bans. Technically, the vision hinges on a hybrid AGI landscape (large foundations distilled into task‑specific agents), advances in human‑agent alignment and negotiation, and privacy/safety tooling (zero‑knowledge proofs, differential privacy, automated escrows, reputation systems) to verify deals without leaking sensitive data. For the AI/ML community this points to research priorities in multi‑agent coordination, mechanism design, scalable negotiation, verification/incentive primitives and alignment tailored to principled fiduciary behavior. Krier frames this as a governance opportunity: AGI could empower decentralized, market‑like solutions to externalities — though practical limits (enforcement costs, inequality, capture risks) still require careful technical and policy design.