<?xml version="1.0" encoding="UTF-8"?>
<!-- generated: thoughts-blog-markdown -->
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Sergi Meseguer Thoughts</title>
    <link>https://sergimeseguer.com/thoughts/</link>
    <description>Thoughts on engineering, products, and the craft of building well.</description>
    <item>
      <title>I don&apos;t want AI agents to start coding. I want them to understand work first.</title>
      <link>https://sergimeseguer.com/thoughts/pi-subagents-extension/</link>
      <guid isPermaLink="true">https://sergimeseguer.com/thoughts/pi-subagents-extension/</guid>
      <pubDate>Sat, 25 Jul 2026 00:00:00 GMT</pubDate>
      <dc:date>2026-07-25</dc:date>
      <description>pi-subagents is a workflow I built around a simple belief: the interface with AI is a spec, not a prompt. It makes agents understand, plan, build, and validate work — while keeping judgment where it belongs: with you.</description>
    </item>
    <item>
      <title>What it took to actually trust an AI agent harness.</title>
      <link>https://sergimeseguer.com/thoughts/trust-ai-agent-harness/</link>
      <guid isPermaLink="true">https://sergimeseguer.com/thoughts/trust-ai-agent-harness/</guid>
      <pubDate>Mon, 13 Apr 2026 00:00:00 GMT</pubDate>
      <dc:date>2026-04-13</dc:date>
      <description>I run three AI coding agents. I didn’t trust any of them with my machine, so I built a container sandbox that gives each one only the access it needs. This is what changed.</description>
    </item>
    <item>
      <title>The header isn&apos;t decoration.</title>
      <link>https://sergimeseguer.com/thoughts/header-is-not-decoration/</link>
      <guid isPermaLink="true">https://sergimeseguer.com/thoughts/header-is-not-decoration/</guid>
      <pubDate>Thu, 19 Mar 2026 00:00:00 GMT</pubDate>
      <dc:date>2026-03-19</dc:date>
      <description>The header on sergimeseguer.com isn&apos;t decoration. It&apos;s a statement.</description>
    </item>
  </channel>
</rss>
