<?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>Token-Economics on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/token-economics/</link><description>Recent content in Token-Economics 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>Fri, 18 Sep 2026 08:00:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/token-economics/index.xml" rel="self" type="application/rss+xml"/><item><title>AI Native Digest 2026-09-18</title><link>https://douyixuan.github.io/posts/ai-native/2026-09-18/</link><pubDate>Fri, 18 Sep 2026 08:00:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-native/2026-09-18/</guid><description>今天关注 AI-native 工程里的两个实际瓶颈：Nous Research 用 1,393 个子 Agent 完成百万行 Python 仓库重构，展示了大规模并行 coding agent 的任务切片、worktree 隔离、基线验证与人工 review；Orbyt Labs 则用 35 个真实工作 session 的 token 数据说明，Agent 成本和行为主要由反复读取的 context 决定，因此 context 文件本身应像代码一样被预算、版本化和测试。</description></item></channel></rss>