Anthropic: Don't Rush to Split Out Sub-Agents

wyr95626@gmail.com (Leafw) — Sun, 25 Oct 2026 00:00:00 +0000

A few days ago I wrote a breakdown of a paper on parallel sub-agents. The takeaway: turning on concurrency generally hurt scores on short tasks, while on long tasks the results were mixed. On October 22, Anthropic published an article on the Claude website, “Building agents: When to fold subagents into your main agent loop”, written by Anthropic’s Diane Sarkis. It’s about when you should fold sub-agents back into the main loop, that is, merge a sub-agent’s prompt and capabilities into the main agent so they become its skills and tools.

Do More Sub-Agents Make Better Code?

wyr95626@gmail.com (Leafw) — Tue, 20 Oct 2026 00:00:00 +0000

Many agent products now let a main agent spin up a batch of sub-agents to work in parallel. But how well does multi-agent collaboration actually work, and is it always better than a single agent? Few people bother to compare.

I recently read a paper that does exactly that: When Sub-Agents Work in Parallel: The Promises and Pitfalls of Dynamic Concurrency in Long-Horizon Coding Tasks, https://arxiv.org/abs/2610.10263 . The work comes from Nanjing University, Central University of Finance and Economics, and the University of Illinois Urbana-Champaign, by Han Li, HanHaoNing Li, Ziqian Jiang, and Yiling Lou. They turned the built-in sub-agent concurrency of three coding agents, Codex, Claude Code, and Kimi Code, on and off, and ran each setting on 354 tasks: 2,124 runs in total, with 11.2 billion tokens logged, which comes to $20,780.02 at standard API prices.

Multi-Agent on Leafw's Blog

Anthropic: Don't Rush to Split Out Sub-Agents

Do More Sub-Agents Make Better Code?