
  <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>MilkLI Blog</title>
      <link>https://lmxbi.online/blog</link>
      <description>关于 AI Agent、自动化工作流与可落地产品实践的博客</description>
      <language>zh-CN</language>
      <managingEditor>li962533@163.com (Milkli)</managingEditor>
      <webMaster>li962533@163.com (Milkli)</webMaster>
      <lastBuildDate>Fri, 10 Apr 2026 00:00:00 GMT</lastBuildDate>
      <atom:link href="https://lmxbi.online/feed.xml" rel="self" type="application/rss+xml"/>
      
  <item>
    <guid>https://lmxbi.online/blog/2026-ai-coding-agent-harness</guid>
    <title>2026 年 AI 编程 Agent 的真正分水岭，不是模型，而是 Harness</title>
    <link>https://lmxbi.online/blog/2026-ai-coding-agent-harness</link>
    <description>大家都在争论哪个模型更会写代码，但真正拉开 AI 编程 Agent 差距的，往往不是模型本身，而是模型之外那套决定它如何读取、修改、验证与纠错的 Harness。</description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>AI Agent</category><category>Harness</category><category>Coding Agent</category><category>Context Engineering</category><category>Software Engineering</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/ai-agent-management-platform-comparison</guid>
    <title>AI Agent 管理平台两条路线之争：Anthropic 官方 Harness vs 网易帝王蟹 ClawHive</title>
    <link>https://lmxbi.online/blog/ai-agent-management-platform-comparison</link>
    <description>单兵作战的 AI 助手已触及天花板。Anthropic 与网易智企同一天发布两款 Agent 管理平台，代表了云端托管与本地纵深防御两条截然不同的路线。</description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>AI Agent</category><category>Harness</category><category>ClawHive</category><category>企业级AI</category><category>多智能体</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/harness-enterprise-agent-productization</guid>
    <title>当 Harness 开始产品化，企业级 Agent 才算真正进入正赛</title>
    <link>https://lmxbi.online/blog/harness-enterprise-agent-productization</link>
    <description>模型只是大脑，Harness 才是骨架、神经系统和组织纪律。网易智企 ClawHive 和 Anthropic Claude Managed Agents 同时交卷，答案的核心逻辑高度一致：企业级 Agent 的竞争，已经从模型能力切换到了 Harness Engineering。</description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>AI Agent</category><category>Harness</category><category>Enterprise AI</category><category>Claude</category><category>Anthropic</category><category>ClawHive</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/llm-rag-personal-knowledge-base</guid>
    <title>Karpathy 强推的 AI 知识库：用最简单的方式让 LLM 成为你的第二大脑</title>
    <link>https://lmxbi.online/blog/llm-rag-personal-knowledge-base</link>
    <description>从 Karpathy 的个人知识库工作流出发，详解 LLMRAG 范式——它与传统 RAG、LightRAG、GraphRAG 的本质区别在哪里，为什么普通人的知识管理应该用它。</description>
    <pubDate>Tue, 07 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>AI知识库</category><category>RAG</category><category>LLM</category><category>个人效率</category><category>Karpathy</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/lossless-claw-enhanced-fully-guide</guid>
    <title>🦞OpenClaw 第二大脑：lossless-claw-enhanced 精准记忆插件完全指南</title>
    <link>https://lmxbi.online/blog/lossless-claw-enhanced-fully-guide</link>
    <description>详解 lossless-claw-enhanced：解决 CJK Token 估算不准的核心问题，详解 DAG 摘要系统与 legacy 滑动窗口的区别，完整安装配置指南与常见坑排查。</description>
    <pubDate>Mon, 06 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>tool</category><category>AI助手</category><category>OpenClaw</category><category>lossless-claw</category><category>上下文管理</category><category>记忆系统</category><category>LLM</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/puaclaws-ai-persuasion-techniques</guid>
    <title>我用PUA话术让AI效率提升34%？亲测这90+种提示词说服技术</title>
    <link>https://lmxbi.online/blog/puaclaws-ai-persuasion-techniques</link>
    <description>PUAClaw 项目实测：把对 AI 的压力测试话术系统化，从彩虹屁到死亡威胁，90多种技术用「龙虾评级」衡量强度，真的有效吗？</description>
    <pubDate>Mon, 06 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>AI助手</category><category>提示词</category><category>PUAClaw</category><category>Prompt Engineering</category><category>AI安全</category><category>LLM</category>
  </item>

  <item>
    <guid>https://lmxbi.online/blog/hermes-agent-full-platform-deploy</guid>
    <title>Hermes Agent 实战: Telegram/飞书/企业微信全平台 AI 助手搭建</title>
    <link>https://lmxbi.online/blog/hermes-agent-full-platform-deploy</link>
    <description>Hermes Agent 实战教程，通过一个进程同时接入 Telegram、飞书和企业微信三大平台，掌握全平台 AI 助手搭建核心配置。</description>
    <pubDate>Wed, 01 Apr 2026 00:00:00 GMT</pubDate>
    <author>li962533@163.com (Milkli)</author>
    <category>tool</category><category>AI助手</category><category>Hermes</category><category>Telegram</category><category>飞书</category><category>企业微信</category>
  </item>

    </channel>
  </rss>
