Economist's Dinner on Multi-Agent Swarms: Debating Whether Swarms Survive Smarter Models
JessicaHullman · x · 2026-11-13
Economist AndreyFradkin hosted a dinner on the economics of multi-agent swarms and shared three takeaways:
- No good taxonomy yet for when swarms beat single-threaded or scaffold-style agentic systems; attendees debated whether swarms offer value beyond speed/parallelization as models get smart enough to solve things alone.
- Self-organization is unsettled: even where swarms stay useful, it's unclear if current training can teach agents optimal self-organization, or whether a Darwinian path of competing, evolving agentic organizations is more likely.
- Economics lens: specialization and distributed information aggregation make markets powerful — a direct analogy for designing agent organizations.
More from AGI Musings
- AI wealth's 'third wave' of philanthropy: ~$37B a year, best spent catalyzing markets — RishiBommasani · 2026-11-13
- Jensen Huang shouldn't have resigned — he should have fought for AGI — jtoy · 2026-11-13
- Open source faces a blunt choice as zero-days surge 10x: accept AI patches or risk attacks — Paimaamu · 2026-11-13
- AI Shopping Agents Could Break the 20-25% Commission Marketplace Model — julianweisser · 2026-11-13
- a16z: 30M coders drive most AI revenue, 1.5B knowledge workers still untapped — robleclerc · 2026-11-13
- Could an AI agent go viral? A thought experiment on worms spreading through 1,000-agent networks — Both-Shape-3953 · 2026-11-13