<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Caged AI blog</title>
    <link>https://caged.ai/blog/</link>
    <description>Field notes on SD-WAN, WAN bonding, and AI traffic governance at the network edge.</description>
    <language>en-us</language>
    <lastBuildDate>Sun, 04 Oct 2026 00:00:00 GMT</lastBuildDate>
    <atom:link href="https://caged.ai/blog/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>How an email domain changed a privacy test result</title>
      <link>https://caged.ai/blog/privacy-test-data-lab-study/</link>
      <guid isPermaLink="true">https://caged.ai/blog/privacy-test-data-lab-study/</guid>
      <description>A synthetic Caged AI lab study shows why test fixtures, detector choices, and inspection latency must be evaluated together.</description>
      <pubDate>Sun, 04 Oct 2026 00:00:00 GMT</pubDate>
      <category>ai governance</category>
      <category>privacy testing</category>
      <category>lab evidence</category>
    </item>
    <item>
      <title>Open weights vs. frontier models: who actually holds the market in 2026</title>
      <link>https://caged.ai/blog/open-weight-vs-frontier-model-market-share-2026/</link>
      <guid isPermaLink="true">https://caged.ai/blog/open-weight-vs-frontier-model-market-share-2026/</guid>
      <description>Two credible datasets put open-weight AI models at 11% and about 60%. Both are right—and the gap matters for branch networks carrying AI traffic.</description>
      <pubDate>Fri, 24 Jul 2026 00:00:00 GMT</pubDate>
      <category>ai governance</category>
      <category>open weight models</category>
      <category>market data</category>
    </item>
  </channel>
</rss>
