<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Hackathon on 知识铺的博客</title>
    <link>https://index.zshipu.com/ai002/tags/Hackathon/</link>
    <description>Recent content in Hackathon on 知识铺的博客</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>zh-CN</language>
    <lastBuildDate>Mon, 31 Aug 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://index.zshipu.com/ai002/tags/Hackathon/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>开发者自编测试让AI Agent演示失去客观性</title>
      <link>https://index.zshipu.com/ai002/post/20260831/%E5%BC%80%E5%8F%91%E8%80%85%E8%87%AA%E7%BC%96%E6%B5%8B%E8%AF%95%E8%AE%A9AI-Agent%E6%BC%94%E7%A4%BA%E5%A4%B1%E5%8E%BB%E5%AE%A2%E8%A7%82%E6%80%A7/</link>
      <pubDate>Mon, 31 Aug 2026 00:00:00 +0000</pubDate>
      <guid>https://index.zshipu.com/ai002/post/20260831/%E5%BC%80%E5%8F%91%E8%80%85%E8%87%AA%E7%BC%96%E6%B5%8B%E8%AF%95%E8%AE%A9AI-Agent%E6%BC%94%E7%A4%BA%E5%A4%B1%E5%8E%BB%E5%AE%A2%E8%A7%82%E6%80%A7/</guid>
      <description>黑客松代理演示中，开发者自己编写测试用例，导致代理总能通过关卡。在All Things Agentic Hackathon上，一个运行于Cloud Run并使用Gemini 3.5 Flash的大麻track-and-trace合规代理同样存在这个问题。构建者同时也是测试制定者，演示环境自然围绕代理的优势设计，真实能</description>
    </item>
  </channel>
</rss>
