<?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>Vulnerability-Research on Napat&#39;s Inverse Blog</title>
    <link>/tags/vulnerability-research/</link>
    <description>Recent content in Vulnerability-Research on Napat&#39;s Inverse Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Sat, 21 Mar 2026 18:48:00 +0700</lastBuildDate>
    <atom:link href="/tags/vulnerability-research/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>ZeroDayBench Replication: What Actually Holds Up in Practice</title>
      <link>/ai-analysis/zerodaybench-replication-field-notes/</link>
      <pubDate>Sat, 21 Mar 2026 18:48:00 +0700</pubDate>
      <guid>/ai-analysis/zerodaybench-replication-field-notes/</guid>
      <description>&lt;p&gt;One of the stranger things about AI security is how many people trust benchmark scores they would never trust anywhere else.&lt;/p&gt;
&lt;p&gt;If someone told you a new static analyzer catches 90% of vulnerabilities, your first question would be: 90% of what? In what code? Under what assumptions? What did it miss? But when an LLM benchmark shows a leaderboard, people often skip those questions and go straight to conclusions.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
