<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Multi-Agent on Jamie&#39;s Blog</title>
    <link>http://akjamie.github.io/tags/multi-agent/</link>
    <description>Recent content in Multi-Agent on Jamie&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Tue, 23 Jun 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="http://akjamie.github.io/tags/multi-agent/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Multi-Agent Architectures: The Diagrams the Numbers Are Hiding</title>
      <link>http://akjamie.github.io/post/2026-06-23-multi-agent-architecture-token-breakdown/</link>
      <pubDate>Tue, 23 Jun 2026 00:00:00 +0000</pubDate>
      <guid>http://akjamie.github.io/post/2026-06-23-multi-agent-architecture-token-breakdown/</guid>
      <description>&lt;h1 id=&#34;multi-agent-architectures-the-diagrams-the-numbers-are-hiding&#34;&gt;Multi-Agent Architectures: The Diagrams the Numbers Are Hiding&lt;/h1&gt;&#xA;&lt;p&gt;LangChain&amp;rsquo;s &lt;a href=&#34;https://www.langchain.com/blog/choosing-the-right-multi-agent-architecture&#34;&gt;Choosing the Right Multi-Agent Architecture&lt;/a&gt; lays out four patterns — &lt;strong&gt;Subagents&lt;/strong&gt;, &lt;strong&gt;Skills&lt;/strong&gt;, &lt;strong&gt;Handoffs&lt;/strong&gt;, and &lt;strong&gt;Router&lt;/strong&gt; — and benchmarks them across three scenarios with tables of model-call counts and token totals. It&amp;rsquo;s a genuinely useful framework. But there&amp;rsquo;s a gap: in every one of the three benchmark scenarios, only &lt;em&gt;one&lt;/em&gt; pattern gets a sequence diagram (almost always Subagents). The other three patterns get a number in a table and nothing to show how that number was produced.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
