
  <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>Chaos and Order</title>
      <link>https://www.youngju.dev/blog</link>
      <description>천천히 올바르게. AI Researcher &amp; DevOps Engineer Youngju&#39;s blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology.</description>
      <language>ko</language>
      <managingEditor>fjvbn2003@gmail.com (Youngju Kim)</managingEditor>
      <webMaster>fjvbn2003@gmail.com (Youngju Kim)</webMaster>
      <lastBuildDate>Sun, 09 Aug 2026 00:00:00 GMT</lastBuildDate>
      <atom:link href="https://www.youngju.dev/tags/harness-engineering/feed.xml" rel="self" type="application/rss+xml"/>
      
  <item>
    <guid>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.en</guid>
    <title>A Harness Is Not Configuration but a Deployable — The Real Bottleneck of the Self-Improvement Loop</title>
    <link>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.en</link>
    <description>The harness engineering post Lilian Weng published in July 2026 gives a name to the whole system wrapped around a model and treats it as a single engineering object. Rather than restating that definition, this post covers what changes when you treat the harness as a versioned deployable instead of a configuration file: how to fingerprint a harness so regressions can be traced, how to treat context as a playbook rather than an ever-growing prompt, and why the real bottleneck in a self-improvement loop is the evaluator rather than the model.</description>
    <pubDate>Sun, 09 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>llm</category><category>agent</category><category>harness-engineering</category><category>context-engineering</category><category>evaluation</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.ja</guid>
    <title>ハーネスは設定ではなくデプロイ成果物です — 自己改善ループの本当のボトルネック</title>
    <link>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.ja</link>
    <description>Lilian Wengが2026年7月にまとめたハーネスエンジニアリングの記事は、モデルを包むシステム全体にひとつの工学対象としての名前を与えます。この記事はその定義を移すのではなく、ハーネスを設定ファイルではなくバージョンの付いたデプロイ成果物として扱うと何が変わるのかを扱います。ハーネスの指紋を取って回帰を捕まえる方法、コンテキストを伸び続けるプロンプトではなくプレイブックとして扱う方法、そして自己改善ループの本当のボトルネックがなぜモデルではなく評価者なのかを整理しました。</description>
    <pubDate>Sun, 09 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>llm</category><category>agent</category><category>harness-engineering</category><category>context-engineering</category><category>evaluation</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement</guid>
    <title>하네스는 설정이 아니라 배포물입니다 — 자기 개선 루프의 진짜 병목</title>
    <link>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement</link>
    <description>Lilian Weng이 2026년 7월에 정리한 하네스 엔지니어링 글은 모델을 감싸는 시스템 전체를 하나의 공학 대상으로 이름 붙입니다. 이 글은 그 정의를 옮기는 대신, 하네스를 설정 파일이 아니라 버전이 붙은 배포물로 다루면 무엇이 달라지는지를 다룹니다. 하네스 지문을 찍어 회귀를 잡는 법, 컨텍스트를 늘어나는 프롬프트가 아니라 플레이북으로 다루는 법, 그리고 자기 개선 루프에서 진짜 병목이 왜 모델이 아니라 평가자인지를 정리했습니다.</description>
    <pubDate>Sun, 09 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>llm</category><category>agent</category><category>harness-engineering</category><category>context-engineering</category><category>evaluation</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.zh</guid>
    <title>框架不是配置，而是交付物 —— 自我改进循环真正的瓶颈</title>
    <link>https://www.youngju.dev/blog/llm/2026-08-09-harness-engineering-self-improvement.zh</link>
    <description>Lilian Weng 在 2026 年 7 月整理的那篇框架工程（harness engineering）文章，给包裹着模型的整个系统起了名字，把它当成一个工程对象。本文不复述那个定义，而是讨论：当你把框架当作带版本的交付物而不是配置文件来对待时，会有什么不同。文中梳理了给框架打指纹以捕捉回归的做法、把上下文当成 playbook 而不是越写越长的提示词的做法，以及为什么自我改进循环真正的瓶颈不是模型而是评估者。</description>
    <pubDate>Sun, 09 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>llm</category><category>agent</category><category>harness-engineering</category><category>context-engineering</category><category>evaluation</category>
  </item>

    </channel>
  </rss>
