<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Sglang on Greg Herlein</title>
    <link>https://blog.herlein.com/tags/sglang/</link>
    <description>Recent content in Sglang on Greg Herlein</description>
    <generator>Hugo</generator>
    <language>en</language>
    <copyright>&amp;copy; 2018-2026. All rights reserved.</copyright>
    <lastBuildDate>Sun, 23 Aug 2026 00:00:02 -0800</lastBuildDate>
    <atom:link href="https://blog.herlein.com/tags/sglang/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>A 27B Model That Reverse-Engineered a License Check (And How to Run It)</title>
      <link>https://blog.herlein.com/post/qwen38-27b-gx10-setup/</link>
      <pubDate>Sun, 23 Aug 2026 00:00:02 -0800</pubDate>
      <guid>https://blog.herlein.com/post/qwen38-27b-gx10-setup/</guid>
      <description>&lt;p&gt;XDA ran a piece that stopped me mid-scroll: &lt;a href=&#34;https://www.xda-developers.com/qwen-3-8-27b-reverse-engineering-job-frontier-model/&#34;&gt;a 27B open-weights model reverse-engineered a commercial application&amp;rsquo;s license check&lt;/a&gt; — recovered a deliberately obscured crypto key out of ARM64 assembly, caught and corrected its own mistake without being told, and produced a working bypass PoC. In about thirty minutes. On a desktop box. I have (a version of) that box. So I went and set it up.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
