<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Minutes of Dylan Patel on architecture</title>
    <link>https://minutesof.com/dylan-patel/on/architecture/</link>
    <description>Everything Dylan Patel has said on architecture: 4 verbatim quotes between February 2025 and July 2026, each with a timestamp and a link to the recording…</description>
    <language>en</language>
    <lastBuildDate>Sun, 30 Aug 2026 16:22:19 +0000</lastBuildDate>
    <atom:link href="https://minutesof.com/dylan-patel/on/architecture/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Patel says OpenAI models are much more sparse while Anthropic models are denser, with different architectural</title>
      <link>https://minutesof.com/q/d82add60-0d13-4045-a3c5-08b2e58a691e/</link>
      <guid isPermaLink="true">https://minutesof.com/q/d82add60-0d13-4045-a3c5-08b2e58a691e/</guid>
      <description>“Open eyes are much more sparse and that has benefits. Then anthropics are, they&#x27;re still sparse but more dense in general and that has different benefits.” — The AGI Post</description>
      <pubDate>Mon, 06 Jul 2026 22:07:54 +0000</pubDate>
      <category>architecture</category>
    </item>
    <item>
      <title>Patel reveals NVIDIA&#x27;s next generation will split inference into separate context processing and decode worklo</title>
      <link>https://minutesof.com/q/8896a5c5-5fee-4e45-86fb-6379c1c3eec2/</link>
      <guid isPermaLink="true">https://minutesof.com/q/8896a5c5-5fee-4e45-86fb-6379c1c3eec2/</guid>
      <description>“NVIDIA&#x27;s next generation actually has something very different. They&#x27;re not saying, Hey, there&#x27;s a training GPU and an inference GPU, right? Because either is fine.” — Together AI</description>
      <pubDate>Fri, 03 Oct 2025 20:23:30 +0000</pubDate>
      <category>architecture</category>
    </item>
    <item>
      <title>Patel explains AI hardware companies made architectural bets that failed when model architectures evolved in u</title>
      <link>https://minutesof.com/q/2245cf47-f50e-40dd-a9e2-a6cf6769af61/</link>
      <guid isPermaLink="true">https://minutesof.com/q/2245cf47-f50e-40dd-a9e2-a6cf6769af61/</guid>
      <description>“But still the model&#x27;s way too big to fit on it. This is, like, very simple. Right? You know, the same thing&#x27;s happening in the other direction.” — No Priors: AI, Machine Learning, Tech, &amp; Startups</description>
      <pubDate>Thu, 14 Aug 2025 10:01:35 +0000</pubDate>
      <category>architecture</category>
    </item>
    <item>
      <title>Patel notes DeepSeek&#x27;s routing innovation removing auxiliary loss represents compounding small improvements ov</title>
      <link>https://minutesof.com/q/fab5c10d-e2cd-4e52-b056-9586751d1efb/</link>
      <guid isPermaLink="true">https://minutesof.com/q/fab5c10d-e2cd-4e52-b056-9586751d1efb/</guid>
      <description>“this type of change can be big, it can be small, but they add up over time.” — Lex Fridman</description>
      <pubDate>Mon, 03 Feb 2025 00:12:13 +0000</pubDate>
      <category>architecture</category>
    </item>
  </channel>
</rss>
