<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>nextToken</title>
    <link>https://frankstuch.de/en/</link>
    <description>A blog about AI in practice: putting new developments in context, trying them out myself and reporting honestly on what works and what doesn&#39;t.</description>
    <language>en</language>
    <atom:link href="https://frankstuch.de/en/feed.xml" rel="self" type="application/rss+xml"/>
    <item>
      <title>Vibe coding vs. spec-driven development: same result, better code</title>
      <link>https://frankstuch.de/en/articles/vibe-gegen-spec/</link>
      <guid>https://frankstuch.de/en/articles/vibe-gegen-spec/</guid>
      <pubDate>Tue, 06 Oct 2026 08:00:00 +0000</pubDate>
      <description>The same application built three times with Claude Code, once in conversation and twice from a specification. For this application of medium complexity, vibe coding was no worse in outcome. But the code of the spec version is split along domain boundaries, and that matters for maintenance and further development.</description>
    </item>
    <item>
      <title>JEV hands-on: not an LLM killer, but a fast specialist</title>
      <link>https://frankstuch.de/en/articles/jev-hands-on/</link>
      <guid>https://frankstuch.de/en/articles/jev-hands-on/</guid>
      <pubDate>Thu, 01 Oct 2026 08:00:00 +0000</pubDate>
      <description>JEV sorts business emails about as accurately as GPT-6 Luna, but answers four times faster and costs about 47% less. A hands-on test with 450 emails, open code and numbers you can check.</description>
    </item>
  </channel>
</rss>