<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>Ci on David Pinkerton</title>
    <link>https://blog.dpinkerton.com/tags/ci/</link>
    <description>Recent content in Ci on David Pinkerton</description>
    <generator>Hugo -- 0.154.5</generator>
    <language>en-us</language>
    <lastBuildDate>Sun, 25 Oct 2026 17:00:00 +1000</lastBuildDate>
    <atom:link href="https://blog.dpinkerton.com/tags/ci/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Teaching My Local LLM to Get Out of the Way</title>
      <link>https://blog.dpinkerton.com/posts/teaching-my-local-llm-to-get-out-of-the-way/</link>
      <pubDate>Sun, 25 Oct 2026 17:00:00 +1000</pubDate>
      <guid>https://blog.dpinkerton.com/posts/teaching-my-local-llm-to-get-out-of-the-way/</guid>
      <description>My home server runs a local LLM and self-hosted CI runners on the same two-core CPU. The model was making CI up to 95% slower, and cpu_shares couldn&amp;rsquo;t fix it. A 40-line watcher that freezes the model whenever a CI job is running did.</description>
    </item>
  </channel>
</rss>
