<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Orchestration on the art of simplicity</title><link>https://naoko.github.io/tags/orchestration/</link><description>Recent content in Orchestration on the art of simplicity</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 08 Aug 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://naoko.github.io/tags/orchestration/index.xml" rel="self" type="application/rss+xml"/><item><title>Graph vs. Loop: Not a Fight, Two Different Layers</title><link>https://naoko.github.io/posts/2026-08-08-graph-vs-loop-agents/</link><pubDate>Sat, 08 Aug 2026 00:00:00 +0000</pubDate><guid>https://naoko.github.io/posts/2026-08-08-graph-vs-loop-agents/</guid><description>&lt;p&gt;A &amp;ldquo;graph engineering just changed how AI agents work — goodbye loops&amp;rdquo; post landed in my feed, and my first reaction was suspicion of the useful kind: &lt;em&gt;wait, is this different from the orchestrator/planner/coder/reviewer setup I already wrote about in the &lt;a href="https://naoko.github.io/posts/2026-04-28-harness-engineering-overview/"&gt;harness cheat sheet&lt;/a&gt;?&lt;/em&gt;&lt;/p&gt;&#10;&lt;p&gt;Short answer: it&amp;rsquo;s not a competing architecture. It&amp;rsquo;s a different layer.&lt;/p&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;&lt;strong&gt;Orchestrator–worker&lt;/strong&gt; (planner, coder, reviewer) answers &lt;strong&gt;who does the work&lt;/strong&gt; — the role topology.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Graph engineering&lt;/strong&gt; answers &lt;strong&gt;how control and state move between them&lt;/strong&gt; — the execution model.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;&lt;p&gt;You can run planner/coder/reviewer as an unconstrained loop, or as a graph. Same cast, different wiring. Nobody has to lose.&lt;/p&gt;</description></item></channel></rss>