<?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>First Proof on AI 实战派 · 从技术到赚钱</title><link>https://guijiagi.com/tags/first-proof/</link><description>Recent content in First Proof on AI 实战派 · 从技术到赚钱</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Fri, 09 Oct 2026 09:18:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/first-proof/index.xml" rel="self" type="application/rss+xml"/><item><title>它解出了第 6、9、10 题，却在第 7 题前停手：选择性暴露的「理解」假象</title><link>https://guijiagi.com/posts/2026-10-09-04-selective-proof-holes/</link><pubDate>Fri, 09 Oct 2026 09:18:00 +0800</pubDate><guid>https://guijiagi.com/posts/2026-10-09-04-selective-proof-holes/</guid><description>OpenAI 内部模型在 First Proof 十道题里只解出特定几题，Navier-Stokes 也是把模型对准每个开放难题后才命中。这种「挑得中、邻接却失手」的模式，暴露了当前 AI 推理离通才理解有多远。</description></item></channel></rss>