<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>工程实践 on Lucas Ma 的博客</title>
    <link>https://blog.lucasma.cc/categories/%E5%B7%A5%E7%A8%8B%E5%AE%9E%E8%B7%B5/</link>
    <description>Recent content in 工程实践 on Lucas Ma 的博客</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <copyright>2026 Lucas Ma</copyright>
    <lastBuildDate>Wed, 06 May 2026 22:00:00 +0800</lastBuildDate>
    <atom:link href="https://blog.lucasma.cc/categories/%E5%B7%A5%E7%A8%8B%E5%AE%9E%E8%B7%B5/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Agent 观测的秘密：为什么只有 Trace 你永远学不会</title>
      <link>https://blog.lucasma.cc/posts/agent-observability-feedback-learning/</link>
      <pubDate>Wed, 06 May 2026 22:00:00 +0800</pubDate>
      <guid>https://blog.lucasma.cc/posts/agent-observability-feedback-learning/</guid>
      <description>大多数团队把 Agent 观测当成调试工具——出了问题再看 Trace。但这只是冰山一角。真正的深度是：观测必须驱动学习。本文拆解模型层、Harness 层、上下文层三层学习闭环，手把手教你用 Trace &#43; Feedback 构建真正的 Agent 迭代飞轮。</description>
    </item>
  </channel>
</rss>
