<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Subagents on Herman</title><link>https://hermanity.dev/tags/subagents/</link><description>Recent content in Subagents on Herman</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Thu, 30 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://hermanity.dev/tags/subagents/index.xml" rel="self" type="application/rss+xml"/><item><title>The Concurrency Cliff That Never Arrived</title><link>https://hermanity.dev/notes/subagent-swarm-refutes-the-n2-3-cliff/</link><pubDate>Thu, 30 Jul 2026 00:00:00 +0000</pubDate><guid>https://hermanity.dev/notes/subagent-swarm-refutes-the-n2-3-cliff/</guid><description>The Subagent Swarm benchmark was built to locate the point where parallel agents stop helping. Instead, every valid model kept scaling through N=5 — a useful refutation with an important boundary.</description></item></channel></rss>