<?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>Program-Synthesis on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/program-synthesis/</link><description>Recent content in Program-Synthesis 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>Mon, 14 Sep 2026 08:00:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/program-synthesis/index.xml" rel="self" type="application/rss+xml"/><item><title>AI Native Digest 2026-09-14</title><link>https://douyixuan.github.io/posts/ai-native/2026-09-14/</link><pubDate>Mon, 14 Sep 2026 08:00:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-native/2026-09-14/</guid><description>今天关注 AI-native 软件工程的两个底层设计问题：Yad Konrad 用 program synthesis / CEGIS 重新解释 prompt、context、harness、eval 与 software factory，指出它们其实是在重新发现 specification、search space、search technique 和 verifier；另一篇新论文则直接比较 Agent Skills 与 Subagents，证明可复用知识该放进主 context 还是隔离到独立 subagent，关键取决于技能是否具有清晰的输入输出契约。</description></item></channel></rss>