An engineering notebook on AI, systems, and everyday experiments.

用好AI的第二步不是更会写 prompt，而是先外化、再复用。本文讲清 Skill 如何承载工作知识、好 Skill 的三要素，以及如何组织 Skill 文件夹让 Agent 自动找到。

Step two isn't better prompting. It's externalize first, reuse second. This post explains how Skills carry work knowledge, the three parts of a good Skill, and how to organize them so agents find the right one.

作为一个重度AI用户，我在经历长期严重失眠后没有走常规的"排除变量"路线，而是用AI写了一个iOS app导出HealthKit数据，做多变量回归分析找到了真正的原因——晚上使用AI高强度思考。这篇文章分享了AI如何在全链条上提供执行力支持，也反思了人的judgment和认知上的成本结构，在AI时代如何重塑我们的决策路径。

After weeks of severe insomnia, I used AI to build an iOS app that exported HealthKit data and ran multivariate regression to find the root cause—late-night AI-assisted intense multitasking. This post explores how AI provided end-to-end execution support and why certain things still require human judgment.

在iOS上查询排版结果只需一行代码，Web上需要触发整个页面的重新布局。这不是因为浏览器工程师蠢，而是CSS在1994年做了一个声明式的架构选择。这个选择的天花板更高，但代价是中间状态不可查询。Facebook在2012年因为不理解这个trade-off付出了数亿美元的代价。SwiftUI和Jetpack Compose证明了声明式和可观测可以共存，关键在于分层。这个教训适用于所有系统设计：好的抽象让你选择在哪一层工作，坏的抽象把所有层粘在一起让你没得选。

Querying layout results takes one line of code on iOS, Android, Qt, and Flutter. On the web, it requires triggering a full-page reflow. This isn't because browser engineers are incompetent. CSS made a deliberate architectural choice in 1994 toward declarative layout, which has a higher ceiling but hides intermediate state. Facebook paid hundreds of millions of dollars in 2012 for not understanding this trade-off. SwiftUI and Jetpack Compose prove that declarative and observable can coexist through proper layering. The lesson applies to all system design: good abstractions let you choose which layer to work at; bad abstractions glue all layers together and leave you no choice.

LLM的默认输出是consensus：正确但平庸。Deep Research其实是Wide Research。我们找到了一种系统性方法，用个人认知上下文把LLM从consensus里强行扯出来。一年实验，有控制变量证据。

An LLM's default output is consensus: correct but mediocre. Deep Research is really Wide Research. We found a systematic way to pull LLMs out of consensus using personal cognitive context. One year of experimentation, with controlled evidence.

会用AI和用好AI之间差的是10倍。这个差距的根源在于工作方式，而非模型。本文通过一个完整的工作流例子和上中下三策的框架，解释为什么应该从ChatGPT切换到Cursor这类Agentic工具。

The gap between using AI and using AI well is 10x. That gap comes from how you work, not which model you use. This post walks through a complete workflow example and a Three Tiers framework to explain why you should switch from ChatGPT to agentic tools like Cursor.