Agensh: Microsoft scales boss-free agent organizations to 1,024 coding agents
rohanpaul_ai · x · 2026-10-05
The arXiv paper "Agensh: Scaling Organizational Intelligence to 1,024 Agents" (Microsoft; authors include Li Dong and Furu Wei) drops the central orchestrator from multi-agent coding harnesses.
Workers run an asynchronous cooperation loop: continuously gathering context, claiming and self-assigning sub-tasks, acting, sharing findings, verifying, and merging progress. The infrastructure has three parts: a shared workspace for proposed/ongoing/completed work, a message interface between workers, and shared context holding reusable findings and intentions.
Evaluated with GPT-5.6-sol (high) on the 5 hardest ProgramBench tasks: scaling 1→128 agents raised the mean test-pass rate from 19.31% to 28.78% (49% relative), with larger organizations reaching comparable scores sooner; on pandoc, 1,024 agents lifted the pass rate from 33.89% to 55.06%. Worker trajectories show emergent self-organized cooperation.
Related event: Microsoft's Agensh Scales Decentralized Coding Agents to 1,024(2 posts)→
More from coding & agent
- Statechart-based actor runtime simulates 1 million foraging and fighting ants in-process — sull · 2026-11-22
- Agent ran for hours playing Smash Ultimate but botched the replay-saving task — anpaure · 2026-11-22
- Teknium shares Hermes telemetry: 92.4% of tokens are cache reads — Teknium · 2026-11-22
- Elvis Saravia opens up his "knowledge factory": a daily-automated MD knowledge graph — omarsar0 · 2026-11-22
- Team says Grok Bot agents ran half the prep for a 5,000-person summit in months — GavinSBaker · 2026-11-22
- Dev shows off browser pinball game polished with Gemini 4 Neon's help — AI_Andrew · 2026-11-22