MIT paper shows inbox rules fix coordination for LLM agent societies of 16,000
omarsar0 · x · 2026-10-30
An MIT Media Lab paper, "The Inbox as a Coordination Primitive for LLM Agent Societies," studies the pull problem in large agent societies: the scarce resource isn't message delivery but reading—each message costs tokens and fills finite context.
- Push fails at scale: Once context fills, agents drop oldest replies; an agent leading 20 negotiations proposes already-rejected options 97% of the time.
- Inbox primitive: Messages queue outside context, and each agent's private rules decide what to read, keep, and reply to.
- Results: In simulations of 1,000–16,000 agents under noisy traffic, first-come-first-read misses every request, while an inbox prioritizing messages with refundable deposits still reaches 60% of needed agents at every size.
- Design: Four rules per inbox (what to read, keep, say as sender, say in replies); optimal rules depend on traffic and role.
- Network effects: Adoption compounds on real email topologies—half adoption recovers only 57% of coverage versus full adoption.
More from coding & agent
- exe.dev launches durable dev sandboxes with SSH root access and built-in coding agent Shelley — davidcrawshaw · 2026-10-30
- Solo dev building a game entirely with AI ships update 5: 494 medals, 28 vehicle types, Steam submission — Apprehensive-Law77 · 2026-10-30
- Dev ditches unanimous-vote stopping rule for a judge in multi-AI debate project — lluisantoni · 2026-10-30
- Replit can now run any GitHub repo from a link and auto-open PRs — amasad · 2026-10-30
- GitHub Octoverse: 34.3B commits in a year as agents start 40% of PRs — lemire · 2026-10-30
- Google Antigravity launches plugins marketplace with Jira, Figma, Stripe connectors — testingcatalog · 2026-10-30