<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
        <title>Harness Engineering on Kd05</title>
        <link>https://kd05.com/tags/harness-engineering/</link>
        <description>Recent content in Harness Engineering on Kd05</description>
        <generator>Hugo -- gohugo.io</generator>
        <language>en</language>
        <lastBuildDate>Thu, 02 Apr 2026 08:30:00 +0800</lastBuildDate><atom:link href="https://kd05.com/tags/harness-engineering/index.xml" rel="self" type="application/rss+xml" /><item>
        <title>Vibe Coding: From Prompt Engineering to Harness Engineering Paradigm Shift</title>
        <link>https://kd05.com/p/vibe-coding-harness-engineering-paradigm-shift/</link>
        <pubDate>Thu, 02 Apr 2026 08:30:00 +0800</pubDate>
        
        <guid>https://kd05.com/p/vibe-coding-harness-engineering-paradigm-shift/</guid>
        <description>&lt;img src="https://kd05.com/p/vibe-coding-harness-engineering-paradigm-shift/cover.jpg" alt="Featured image of post Vibe Coding: From Prompt Engineering to Harness Engineering Paradigm Shift" /&gt;&lt;h2 id=&#34;introduction-the-problem-with-free-form-ai-coding&#34;&gt;Introduction: The Problem with &amp;ldquo;Free-form&amp;rdquo; AI Coding
&lt;/h2&gt;&lt;p&gt;I&amp;rsquo;ve been experimenting with various SDD development approaches in Vibe Coding, such as OpenSpec, SpecKit, and oh-my-openagent&amp;rsquo;s Prometheus.&lt;/p&gt;
&lt;p&gt;Andrej Karpathy introduced the concept of &lt;strong&gt;Vibe Coding&lt;/strong&gt; in February 2025—describing intent in natural language, letting LLMs generate code, &amp;ldquo;Accept All&amp;rdquo; without reviewing diffs, and just copy-pasting error messages for AI to fix when things go wrong.&lt;/p&gt;
&lt;p&gt;But once I actually started, I discovered a critical problem: &lt;strong&gt;AI keeps improvising beyond my intent&lt;/strong&gt;.&lt;/p&gt;</description>
        </item>
        
    </channel>
</rss>
