<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>dataeng.ai</title>
    <link>https://dataengai.com/</link>
    <description>AI for data engineers. No hype, just what works in production.</description>
    <language>en-us</language>
    <item>
      <title>Your pipeline samples 2%. Judge all of them.</title>
      <link>https://dataengai.com/posts/jev-ep1-judge-all-rows/</link>
      <guid>https://dataengai.com/posts/jev-ep1-judge-all-rows/</guid>
      <pubDate>Thu, 24 Sep 2026 00:00:00 GMT</pubDate>
      <description>MotherDuck ran the numbers on judging all 100,000 rows: 89% accuracy, 40 seconds, fifty cents. The blocker was never accuracy. It was price.</description>
    </item>
    <item>
      <title>Airflow 3.3.2 shipped an upgrade trap, and the error message lies.</title>
      <link>https://dataengai.com/posts/airflow-332-one-way-door/</link>
      <guid>https://dataengai.com/posts/airflow-332-one-way-door/</guid>
      <pubDate>Wed, 23 Sep 2026 00:00:00 GMT</pubDate>
      <description>pandas 3 renames the DataFrame class path and XComs record the old one. Upgrade Airflow on every worker first, or the error message will send you chasing your config.</description>
    </item>
    <item>
      <title>dbt turned its AI on by default. Check the toggle.</title>
      <link>https://dataengai.com/posts/dbt-ai-default-on/</link>
      <guid>https://dataengai.com/posts/dbt-ai-default-on/</guid>
      <pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
      <description>From September 1, Wizard and Copilot ship enabled and every future dbt AI feature auto-enables. The meter is running, and the pool is shared.</description>
    </item>
    <item>
      <title>Your demo worked. Your AI product won't.</title>
      <link>https://dataengai.com/posts/demo-worked-product-wont/</link>
      <guid>https://dataengai.com/posts/demo-worked-product-wont/</guid>
      <pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate>
      <description>Your demo ran on ten hand-picked examples. Production is the long tail of real inputs. Four things break, and none of them are the model.</description>
    </item>
    <item>
      <title>Your support bot is hallucinating. Blame the dumpster, not the model.</title>
      <link>https://dataengai.com/posts/long-context-rot/</link>
      <guid>https://dataengai.com/posts/long-context-rot/</guid>
      <pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate>
      <description>2:14 AM. The bot is inventing refund policies. The retriever is fine and the prompt is fine. The problem is the fourteen thousand tokens of dumpster you fed it.</description>
    </item>
    <item>
      <title>RAG is just a data pipeline</title>
      <link>https://dataengai.com/posts/rag-is-just-a-data-pipeline/</link>
      <guid>https://dataengai.com/posts/rag-is-just-a-data-pipeline/</guid>
      <pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate>
      <description>Ingest, chunk, embed, index, retrieve, generate. The model is the last box in the diagram, and everything before it is your world.</description>
    </item>
  </channel>
</rss>