AI 把讲授的边际成本打到零，教育史上第一次，瓶颈从'谁来教'搬到了'人为什么想学'。顺着一个 AI 社群里关于'孩子为什么不肯学'的讨论，我把动机科学、学习科学和 alignment 的证据串了一遍：需求不是找出来的，是点着的；'以后用得上'输给的是一条折现曲线；动机满格的学生加一个直接给答案的 AI，考试成绩反而比不用的还低 17%。教育要解的是双重对齐——先让人想学，再保住那份不可外包的认知劳动。

风格库再大，也装不下所有人的偏好。有人想写成半佛仙人，有人想写成和菜头，还有人想照着昨天看到的一篇没听过名字的作者的文章来。怎么办？让 AI 现场提取风格 DNA——给一篇范文，改成那个味。

A style library, no matter how large, can't cover everyone's taste. Some users want to write like an author who isn't in the library. Some want the voice of one specific article, not a generalized style. The solution: real-time style extraction. Give the AI any reference article, and it clones the writing DNA on the fly.

一套写作指南搞定了翻译腔，但科技评测文章写成半佛那味儿就离谱了。不同内容需要不同声音。我研究了怎么从任何一个作者身上'偷'写作风格——收集文章、拆解模式、提炼成结构化的 style profile。这篇用卡兹克做例子，走一遍完整流程。

One style guide fixed translation-ese, but a tech review shouldn't sound like social commentary. Different content needs different voices. I developed a systematic method for 'stealing' any author's writing style — collecting articles, dissecting patterns, and extracting a structured Style Profile. Here's the full process, demonstrated with a real example.

我博客上的中文文章一直是 claude 帮我翻译加改写的，攒下来读着全像英文直译。后来我去拆了半佛仙人、九边这些人的写法，把规律整理成一份写作指南喂回给 claude，让它把文章全部重写了一遍。这篇讲整个过程是怎么做的。

My blog had 83 Chinese posts — all of them read like translated English. Stiff, formal, academic. So I studied China's best WeChat writers, distilled their craft into a 435-line writing system, and fed it to Claude as a system prompt. Then I rewrote every single post. This is the story of that guide.