Agensh is a multi-agent system that removes the central orchestrator bottleneck. Workers self-organize by asynchronously gathering context, claiming sub-tasks, acting, sharing findings, and merging results via a shared workspace, message interface, and shared context. On the five hardest ProgramBench tasks, scaling from 1 to 128 agents lifted the mean test-pass rate from 19.31% to 28.78%, a 49% relative improvement. On the pandoc task, scaling from 1 to 1,024 agents raised the rate from 33.89% to 55.06%.
Self-organized agent teams improve coding results as they scale
Agensh replaces central task allocation with asynchronous coordination and scales programming tasks from one to 1,024 concurrent agents.
Big Tech
Zhihao Zhan · Ting Song · Li Dong · Shaohan Huang · Jianxun Lian · Yan Xia · +1 more
Microsoft Research
Research Digest··2 min read
Thread:Multi-Agent Coordination
In one line
A multi-agent harness without a central orchestrator scales to 1,024 agents, improving test-pass rates by up to 49%.
Before you read this
Three ideas this paper builds onmulti-agent architectureA system where multiple AI agents collaborate, each handling specialized roles, communicating and coordinating to complete complex tasks.110 papers →agent harnessA system that structures, verifies, monitors, or controls how an AI agent interacts with external tools, environments, or execution runs.176 papers →multi-agent coordinationThe collaborative operation of multiple AI agents to solve tasks more effectively or with greater capability.87 papers →
Read it as
Zhan et al. built a multi-agent harness in which workers independently claim tasks, exchange findings, verify results, and merge progress through shared infrastructure.
The authors developed Agensh, a decentralized harness without a central orchestrator.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.
What it released
Code