<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Control Theory on Andrew Hunter — Systems Architect &amp; Builder</title>
    <link>https://andrewphunter.com/tags/control-theory/</link>
    <description>Recent content in Control Theory on Andrew Hunter — Systems Architect &amp; Builder</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Tue, 22 Sep 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://andrewphunter.com/tags/control-theory/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Reasoning Tokens Steer the Trajectory</title>
      <link>https://andrewphunter.com/writing/reasoning-tokens-control-signal/</link>
      <pubDate>Tue, 22 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://andrewphunter.com/writing/reasoning-tokens-control-signal/</guid>
      <description>&lt;p&gt;The argument over reasoning models is loud and stuck at the wrong level. One camp reads the reasoning trace as a window into the model&amp;rsquo;s mind. The other reads it as after-the-fact theater, tokens emitted to look like thinking. Both camps do the same thing: they read the tokens as language and ask what the tokens say. The mechanism does not care what they say.&lt;/p&gt;&#xA;&lt;p&gt;A reasoning token is a step that steers a trajectory through the model&amp;rsquo;s probability field. Ask what it does to the distribution instead of what it means, and the whole &amp;ldquo;do the tokens mean anything&amp;rdquo; question dissolves into something you can measure.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
