<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>aiengineering.guide — The AI Engineering Roadmap</title>
    <link>https://aiengineering.guide/</link>
    <description>New lessons, notes, and case files from the zero-to-shipping AI engineering roadmap.</description>
    <language>en</language>
    <item>
      <title>What is an LLM, really?</title>
      <link>https://aiengineering.guide/lesson/what-is-an-llm</link>
      <guid>https://aiengineering.guide/lesson/what-is-an-llm</guid>
      <description>The mental model that will carry you through the rest of the roadmap. Not the history, not the maths — the mechanism, in the shape you&apos;ll actually reason about when things go wrong.</description>
      <category>lesson</category>
      <pubDate>Tue, 08 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Design an LLM gateway</title>
      <link>https://aiengineering.guide/lesson/llm-gateway</link>
      <guid>https://aiengineering.guide/lesson/llm-gateway</guid>
      <description>One door in front of every model call, where limits, retries, and fallbacks live once. Build the limiter from scratch, then see what a real gateway adds.</description>
      <category>lesson</category>
      <pubDate>Sun, 02 Aug 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Streaming responses aren&apos;t free</title>
      <link>https://aiengineering.guide/notes/streaming-is-not-free</link>
      <guid>https://aiengineering.guide/notes/streaming-is-not-free</guid>
      <description>Token streaming looks like a latency win, but it moves work into your error handling and your client. Here&apos;s what it actually costs.</description>
      <category>note</category>
      <pubDate>Sun, 02 Aug 2026 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>The eval pipeline that lied for a month</title>
      <link>https://aiengineering.guide/cases/eval-pipeline-that-lied</link>
      <guid>https://aiengineering.guide/cases/eval-pipeline-that-lied</guid>
      <description>A team shipped model changes against a green eval suite for a month while quality quietly dropped. Here&apos;s the decomposition — and the wrong turn that hid it.</description>
      <category>case</category>
      <pubDate>Sun, 02 Aug 2026 00:00:00 GMT</pubDate>
    </item>
  </channel>
</rss>