<?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>Agents on lgcode</title>
    <link>https://blog.lgcode.me/tags/agents/</link>
    <description>Recent content in Agents on lgcode</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Sat, 22 Aug 2026 09:00:00 -0300</lastBuildDate>
    <atom:link href="https://blog.lgcode.me/tags/agents/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>One Substrate, Two Renderers</title>
      <link>https://blog.lgcode.me/posts/one-substrate-two-renderers/</link>
      <pubDate>Sat, 22 Aug 2026 09:00:00 -0300</pubDate>
      <guid>https://blog.lgcode.me/posts/one-substrate-two-renderers/</guid>
      <description>The pipeline that turns a production system into a demo video and the one that turns postmortems into blog posts are the same machine. Only the last step differs, and it&amp;#39;s the cheap step.</description>
    </item>
    <item>
      <title>The Demo Video Was the Easy Part</title>
      <link>https://blog.lgcode.me/posts/the-demo-video-was-the-easy-part/</link>
      <pubDate>Sat, 08 Aug 2026 09:00:00 -0300</pubDate>
      <guid>https://blog.lgcode.me/posts/the-demo-video-was-the-easy-part/</guid>
      <description>A margin that was arithmetically correct and substantively false, confidence scores from a model that wasn&amp;#39;t deployed, and one in five invoice lines failing its own arithmetic. Generating the video took minutes. Making it true took two days.</description>
    </item>
    <item>
      <title>What Dogfooding an MCP Server Taught Me About Tool Output</title>
      <link>https://blog.lgcode.me/posts/mcp-tool-output-design/</link>
      <pubDate>Mon, 03 Aug 2026 10:00:00 -0300</pubDate>
      <guid>https://blog.lgcode.me/posts/mcp-tool-output-design/</guid>
      <description>A filter that discarded the answer, a retry that couldn&amp;#39;t work, a partial result posing as a total, and cached prices with no freshness signal. All four found in one live session, all four the same class of mistake.</description>
    </item>
  </channel>
</rss>
