<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>ideius — articles</title>
    <link>https://www.ideius.com/articles/</link>
    <description>AI consulting practice based in Media, Pennsylvania, serving Philadelphia and the Greater Philadelphia region.</description>
    <language>en-us</language>
    <atom:link href="https://www.ideius.com/articles/feed.xml" rel="self" type="application/rss+xml" /><lastBuildDate>2026-05-07T00:00:00.000Z</lastBuildDate>
    
    <item>
      <title>Evals that stick: how to keep retrieval honest after launch</title>
      <link>https://www.ideius.com/articles/evals-that-stick/</link>
      <guid isPermaLink="true">https://www.ideius.com/articles/evals-that-stick/</guid>
      <pubDate>2026-05-07T00:00:00.000Z</pubDate>
      <description><![CDATA[Most LLM evaluation suites die in the first month. The ones that survive are tied to a release decision the team actually makes.]]></description>
      
    </item>
    
  </channel>
</rss>
