<?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>Tool-Design on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/tool-design/</link><description>Recent content in Tool-Design on 兜忆轩 | Cedric's Compiler Notes</description><image><title>兜忆轩 | Cedric's Compiler Notes</title><url>https://douyixuan.github.io/images/papermod-cover.png</url><link>https://douyixuan.github.io/images/papermod-cover.png</link></image><generator>Hugo -- 0.146.0</generator><language>en-us</language><lastBuildDate>Tue, 22 Sep 2026 08:00:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/tool-design/index.xml" rel="self" type="application/rss+xml"/><item><title>AI Native Digest 2026-09-22</title><link>https://douyixuan.github.io/posts/ai-native/2026-09-22/</link><pubDate>Tue, 22 Sep 2026 08:00:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-native/2026-09-22/</guid><description>今天关注 AI-native 工程里两个越来越关键的接口层问题：Prasenjit Paul 用 54 次受控实验说明，Agent tool 的命名、schema、返回值与成本提示会直接改变上下文消耗和行动路径；NVIDIA 则从安全工程角度强调，生产 Agent 的身份、权限、网络边界、审计和验证必须由模型之外的 runtime 与平台层强制执行。</description></item></channel></rss>