<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Shubham Tripathi | AI Engineering Blog</title>
    <link>https://shubham-tripathi.com/blog/</link>
    <atom:link href="https://shubham-tripathi.com/feed.xml" rel="self" type="application/rss+xml"/>
    <description>AI research explained in plain language, fact-checked against the sources, plus notes on building LLM systems.</description>
    <language>en-gb</language>
    <lastBuildDate>Thu, 05 Nov 2026 05:19:42 +0000</lastBuildDate>
    <image>
      <url>https://shubham-tripathi.com/assets/og-image.jpg</url>
      <title>Shubham Tripathi | AI Engineering Blog</title>
      <link>https://shubham-tripathi.com/blog/</link>
    </image>
    <item>
      <title>An AI edited another AI’s weights, and its math score jumped from 7.9% to 48.1%</title>
      <link>https://shubham-tripathi.com/blog/weight-space-programming-gpt-6-astra-qwen/</link>
      <guid isPermaLink="true">https://shubham-tripathi.com/blog/weight-space-programming-gpt-6-astra-qwen/</guid>
      <pubDate>Thu, 05 Nov 2026 02:00:00 +0000</pubDate>
      <dc:creator>Shubham Tripathi</dc:creator>
      <category>AI research</category>
      <description>Inside MIT CSAIL’s weight-space programming study, where GPT-6 Astra fixed Qwen3.5-0.8B by editing 0.016% of its weights, and what the result does and doesn’t show.</description>
      <enclosure url="https://shubham-tripathi.com/blog/weight-space-programming-gpt-6-astra-qwen/cover.jpg" type="image/jpeg" length="65778"/>
    </item>
  </channel>
</rss>
