<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Karpathy on Yuanchang&#39;s Blog</title>
    <link>https://yuanchang.org/en/tags/karpathy/</link>
    <description>Recent content in Karpathy on Yuanchang&#39;s Blog</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>en-us</language>
    <lastBuildDate>Wed, 25 Mar 2026 16:00:00 +0100</lastBuildDate><atom:link href="https://yuanchang.org/en/tags/karpathy/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>pip install Is the New Attack Surface: Three Lessons from the LiteLLM Incident for AI Founders</title>
      <link>https://yuanchang.org/en/posts/pip-install-new-attack-surface-litellm/</link>
      <pubDate>Wed, 25 Mar 2026 16:00:00 +0100</pubDate>
      
      <guid>https://yuanchang.org/en/posts/pip-install-new-attack-surface-litellm/</guid>
      
      <description>On March 24, 2026, LiteLLM — a Python package with 97 million monthly downloads — was backdoored on PyPI. A single pip install was enough to exfiltrate SSH keys, cloud credentials, and every API key on the machine. 500,000 machines were compromised, 300GB of credentials stolen. This post reconstructs the incident, provides a self-check prompt you can hand to your AI agent, and analyzes three lessons AI founders should take seriously.</description>
      
    </item>
    
  </channel>
</rss>
