﻿<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0">
  <channel>
    <title>Latent Heat</title>
    <link>https://latentheat.dev</link>
    <description>Small models, ordinary hardware, negative results, and useful quantities of waste heat.</description>
    <item>
      <title>Jaffabotomisation</title>
      <link>https://latentheat.dev/blog/jaffabotomisation</link>
      <guid>https://latentheat.dev/blog/jaffabotomisation</guid>
      <description>I lied to models about a concept-injection hook, then actually did the injection. The aligned model refused, the uncensored one sushi'd everything, and the concept I wanted turned out not to exist as one feature.</description>
      <pubDate>Sun, 19 Jul 2026 23:00:00 GMT</pubDate>
    </item>
    <item>
      <title>I gave an AI a memory. It immediately cheated.</title>
      <link>https://latentheat.dev/blog/memory-lab-verify-gap</link>
      <guid>https://latentheat.dev/blog/memory-lab-verify-gap</guid>
      <description>External memory gave a frozen model a real new capability. Unfortunately, I had to survive my own research methods to prove it.</description>
      <pubDate>Fri, 17 Jul 2026 23:00:00 GMT</pubDate>
    </item>
    <item>
      <title>My model learned to think. It never learned to listen.</title>
      <link>https://latentheat.dev/blog/latent-threads-negative-result</link>
      <guid>https://latentheat.dev/blog/latent-threads-negative-result</guid>
      <description>I gave a small transformer private thoughts. The metrics said it worked. Removing the thoughts made it better.</description>
      <pubDate>Fri, 17 Jul 2026 23:00:00 GMT</pubDate>
    </item>
  </channel>
</rss>