<?xml version="1.0" encoding="utf-8" standalone="yes" ?>
<rss version="2.0" 
  xmlns:content="http://purl.org/rss/1.0/modules/content/" 
  xmlns:dc="http://purl.org/dc/elements/1.1/" 
  xmlns:atom="http://www.w3.org/2005/Atom" 
  xmlns:sy="http://purl.org/rss/1.0/modules/syndication/" 
  xmlns:media="http://search.yahoo.com/mrss/">
  <channel>
    <title>AI Agent on ✏VictorHong&#39;s Blog</title>
    <link>https://hugo.jiahongw.com/tags/ai-agent/</link>
    <description>Recent content in AI Agent on ✏VictorHong&#39;s Blog</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>zh</language>
    <follow_challenge>
        <feedId>74349588616178697</feedId>
        <userId>68207800989737984</userId>
    </follow_challenge>
    <managingEditor>wujiahong2013@gmail.com (VictorHong)</managingEditor>
    <webMaster>wujiahong2013@gmail.com (VictorHong)</webMaster>
    <copyright>©2021-2026 | VictorHong</copyright>
    <lastBuildDate>Sun, 31 May 2026 22:35:33 +0800</lastBuildDate>
    
        <atom:link href="https://hugo.jiahongw.com/tags/ai-agent/index.xml" rel="self" type="application/rss+xml" />
    

      
      <item>
        <title>553 小时后，我对 AI 编程的一些反直觉观察</title>
        <link>https://hugo.jiahongw.com/20260531/553-hours-claude-code/</link>
        <pubDate>Sun, 31 May 2026 22:35:33 +0800</pubDate>
        <author>wujiahong2013@gmail.com (VictorHong)</author>
        <atom:modified>Sun, 31 May 2026 22:35:33 +0800</atom:modified>
        <guid>https://hugo.jiahongw.com/20260531/553-hours-claude-code/</guid>
        <description>不是「AI 多厉害」，而是「人该怎么配合」。 昨天刷到一条 Reddit 帖子，一位开发者在 Claude Code 里待了 553 小时，总结出了四条经验。数据很具体：89% 的缓存命中率</description>
        
        <dc:creator>VictorHong</dc:creator>
        <media:content url="https://hugo.jiahongw.com553-hours-claude-code-cover.png" medium="image"><media:title type="html">featured image</media:title></media:content>
        
        
        
          
            
              <category>AI 编程</category>
            
          
            
              <category>Claude Code</category>
            
          
            
              <category>开发效率</category>
            
          
            
              <category>AI Agent</category>
            
          
        
        
          
            
              <category>technology</category>
            
          
        
        
      </item>
      
      <item>
        <title>AI-Agent-进化论：从工具到系统的范式转移</title>
        <link>https://hugo.jiahongw.com/20260411/ai-agent-%E8%BF%9B%E5%8C%96%E8%AE%BA%E4%BB%8E%E5%B7%A5%E5%85%B7%E5%88%B0%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%8C%83%E5%BC%8F%E8%BD%AC%E7%A7%BB/</link>
        <pubDate>Sat, 11 Apr 2026 16:29:29 +0800</pubDate>
        <author>wujiahong2013@gmail.com (VictorHong)</author>
        <atom:modified>Sat, 11 Apr 2026 16:29:29 +0800</atom:modified>
        <guid>https://hugo.jiahongw.com/20260411/ai-agent-%E8%BF%9B%E5%8C%96%E8%AE%BA%E4%BB%8E%E5%B7%A5%E5%85%B7%E5%88%B0%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%8C%83%E5%BC%8F%E8%BD%AC%E7%A7%BB/</guid>
        <description>AI Agent 进化论：从工具到系统的范式转移 当 AI 不再只是回答问题，而是开始记住你、学习你、进化自己——我们正站在人机协作的新起点。 引言 2026 年 4 月，AI Agent</description>
        
        <dc:creator>VictorHong</dc:creator>
        
        
        
        
          
            
              <category>AI Agent</category>
            
          
            
              <category>Hermes Agent</category>
            
          
            
              <category>Claude</category>
            
          
            
              <category>Anthropic</category>
            
          
            
              <category>人工智能</category>
            
          
            
              <category>AI</category>
            
          
            
              <category>GPT</category>
            
          
            
              <category>LLM</category>
            
          
            
              <category>Prompt</category>
            
          
        
        
          
            
              <category>AI</category>
            
          
        
        
      </item>
      
      <item>
        <title>AutoCLI：把 55&#43; 网站变成你的命令行工具</title>
        <link>https://hugo.jiahongw.com/20260407/autocli-deep-research/</link>
        <pubDate>Tue, 07 Apr 2026 00:00:00 +0000</pubDate>
        <author>wujiahong2013@gmail.com (VictorHong)</author>
        <atom:modified>Tue, 07 Apr 2026 00:00:00 +0000</atom:modified>
        <guid>https://hugo.jiahongw.com/20260407/autocli-deep-research/</guid>
        <description>&lt;p&gt;AutoCLI 是一个用 Rust 重写的极速 CLI 工具，能把任何网站变成命令行接口。它复用 Chrome 登录态（无需 API Key），支持 55+ 网站（Twitter、Reddit、Bilibili、知乎、小红书等），单文件 4.7MB、零依赖。相比原 TypeScript 版本 OpenCLI，速度快 12 倍、内存省 10 倍。对 AI Agent 极其友好——在 AGENT.md 里配置 &lt;code&gt;autocli list&lt;/code&gt;，AI 就能自动发现并调用所有工具。&lt;/p&gt;</description>
        
        <dc:creator>VictorHong</dc:creator>
        <media:content url="https://hugo.jiahongw.comhttps://cos.jiahongw.com/agent/20260411/autocli-cover.png" medium="image"><media:title type="html">featured image</media:title></media:content>
        
        
        
          
            
              <category>cli</category>
            
          
            
              <category>rust</category>
            
          
            
              <category>ai-agent</category>
            
          
            
              <category>web-scraping</category>
            
          
            
              <category>automation</category>
            
          
        
        
          
            
              <category>tech</category>
            
          
        
        
      </item>
      
      <item>
        <title>当顶级AI工程师停止写代码：Karpathy与AI Agent革命</title>
        <link>https://hugo.jiahongw.com/20260325/%E5%BD%93%E9%A1%B6%E7%BA%A7ai%E5%B7%A5%E7%A8%8B%E5%B8%88%E5%81%9C%E6%AD%A2%E5%86%99%E4%BB%A3%E7%A0%81-karpathy%E4%B8%8Eai-agent%E9%9D%A9%E5%91%BD/</link>
        <pubDate>Wed, 25 Mar 2026 10:30:00 +0800</pubDate>
        <author>wujiahong2013@gmail.com (VictorHong)</author>
        <atom:modified>Wed, 25 Mar 2026 10:30:00 +0800</atom:modified>
        <guid>https://hugo.jiahongw.com/20260325/%E5%BD%93%E9%A1%B6%E7%BA%A7ai%E5%B7%A5%E7%A8%8B%E5%B8%88%E5%81%9C%E6%AD%A2%E5%86%99%E4%BB%A3%E7%A0%81-karpathy%E4%B8%8Eai-agent%E9%9D%A9%E5%91%BD/</guid>
        <description>核心观点：从键盘到指挥台 Andrej Karpathy 从 2025 年 12 月起就再也没亲手写过一行代码。 这位 OpenAI 联合创始人、前 Tesla AI 总监，在 No Priors Podcast 上公开表示，他现在把八成以上的编码工</description>
        
        <dc:creator>VictorHong</dc:creator>
        <media:content url="https://hugo.jiahongw.comhttps://cos.jiahongw.com/rss-daily/20260325/cover.png" medium="image"><media:title type="html">featured image</media:title></media:content>
        
        
        
          
            
              <category>AI Agent</category>
            
          
            
              <category>Claude Code</category>
            
          
            
              <category>Vibe Coding</category>
            
          
            
              <category>软件工程</category>
            
          
            
              <category>OpenAI</category>
            
          
            
              <category>Karpathy</category>
            
          
        
        
          
            
              <category>技术</category>
            
          
        
        
          
            
              <category>RSS Daily</category>
            
          
        
      </item>
      
      <item>
        <title>Superset 最佳实践：把 AI 变成你的并行工程团队</title>
        <link>https://hugo.jiahongw.com/20260312/superset-best-practices/</link>
        <pubDate>Thu, 12 Mar 2026 09:43:00 +0000</pubDate>
        <author>wujiahong2013@gmail.com (VictorHong)</author>
        <atom:modified>Thu, 12 Mar 2026 09:43:00 +0000</atom:modified>
        <guid>https://hugo.jiahongw.com/20260312/superset-best-practices/</guid>
        <description>核心思想：你不是在用一个 AI 助手，而是在管理一支 10 人工程团队。 Superset 是一个基于 Git worktree 的 AI Agent 并行编排工具。与传统 AI 编程（任务 → 等待 Agent → Review → 下一个任务）</description>
        
        <dc:creator>VictorHong</dc:creator>
        <media:content url="https://hugo.jiahongw.comhttps://openclaw.cos.jiahongw.com/blog/superset-best-practices-cover.png" medium="image"><media:title type="html">featured image</media:title></media:content>
        
        
        
          
            
              <category>Superset</category>
            
          
            
              <category>AI Agent</category>
            
          
            
              <category>Claude Code</category>
            
          
            
              <category>Codex</category>
            
          
            
              <category>Git Worktree</category>
            
          
            
              <category>并行开发</category>
            
          
            
              <category>效率工具</category>
            
          
        
        
          
            
              <category>AI编程</category>
            
          
        
        
      </item>
      

    
  </channel>
</rss>