Back to Details

ljg-book

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

ljg-book: 拆书

ljg-book: Book Teardown

输入一本书，输出它的骨架。
Input a book, output its core framework.

拆书是什么

What is Book Teardown

不是读后感，不是章节摘要，不是夸——是把作者的骨架抽出来摆桌上看。
读一本书容易迷在作者的笔触里。拆一本书是从笔触里跳出来，看作者的取景框。取景框看清，整本书才看清。
It’s not a book review, not a chapter summary, not praise—it’s pulling out the author’s core framework and laying it out on the table.
It’s easy to get lost in an author’s writing style while reading. Book teardown means stepping out of that style to see the author’s "viewfinder." Only when you understand the viewfinder can you truly understand the whole book.

五件事

Five Key Points

每一件单独问一遍。问完串起来。
Ask each question separately, then string the answers together.

1. 核心问题

1. Core Question

作者真正在答什么？这是一个问题，不是一个题目。
题目可以是"自由意志"——问题是"自由意志这种东西到底存不存在，还是我们以为有但其实没有？"
题目是 topic。问题是 question。
问题要锋利——它得是作者写整本书前就在挠他的那个东西。书是答案，问题是挠痒处。
怎么找：读作者的导言/前言，找"我写这本书是为了……"那一句。读不到，看作者反复回到的那一处——他绕了几百页都甩不掉的东西，就是问题本身。
What is the author really answering? This is a question, not a topic.
A topic could be "free will"—the question would be "Does free will actually exist, or do we just think it does?"
A topic is a topic. A question is a question.
The question must be sharp—it’s the itch that bothered the author before they even started writing the book. The book is the answer; the question is the itch.
How to find it: Read the author’s introduction/preface, look for the sentence that starts with "I wrote this book to..." If you can’t find it, look for the point the author keeps returning to—the thing they can’t shake off even after hundreds of pages, that’s the core question.

2. 基础假设

2. Fundamental Assumptions

作者立在什么不证之物上？
每本书都有几条天花板——作者不去证它们，他靠它们站着。动摇任何一条，整本书就塌。
操作问句两条，对着书问：
  • 这书里什么是不许问的？（一问就要被作者归为外行）
  • 什么是作者要相信才能往下写的？（少了它，后面所有论证失支点）
答出来的几条，就是这本书的基础假设。
亮出来不展开。这几条本身不证——摆桌上就行，证不证不在拆书人手里。
What unproven foundations does the author stand on?
Every book has a few "ceilings"—the author doesn’t prove them, they rely on them to build their argument. Shake any one of them, and the whole book collapses.
Use two operational questions to analyze the book:
  • What is off-limits to question in this book? (Asking it would get you labeled an outsider by the author)
  • What must the author believe to keep writing? (Without it, all subsequent arguments lose their foundation)
The answers to these are the book’s fundamental assumptions.
Present them without elaboration. These assumptions don’t need proof—just lay them out; proving them isn’t the teardown’s job.

3. 分析框架

3. Analytical Framework

作者用什么看世界？
每个作者都有一台取景框。Darwin 的取景框是"变异 + 选择 + 时间"——他看蝴蝶看人看部落都是这台框。Marx 的取景框是"生产关系决定上层建筑"——他看历史看社会看艺术都是这台框。Freud 的取景框是"意识冰山一角，水下才是真主角"——他看梦看口误看政治都是这台框。
找作者的框：他反复用什么工具？他从什么角度切入？他看到 X 第一反应想到的是 Y——这个 Y 就是他的框。
框不一定是名词。可以是一个动作（"还原到 root cause"）、一个区分（"自然态 vs 文化态"）、一个对照（"显性逻辑 vs 隐性结构"）。
找框要够狠——一句话讲清楚，不是一个章节。
What lens does the author use to see the world?
Every author has a viewfinder. Darwin’s viewfinder is "variation + selection + time"—he looks at butterflies, humans, and tribes through this lens. Marx’s viewfinder is "relations of production determine superstructure"—he looks at history, society, and art through this lens. Freud’s viewfinder is "consciousness is just the tip of the iceberg; the unconscious is the real protagonist"—he looks at dreams, slips of the tongue, and politics through this lens.
How to find the author’s lens: What tool do they repeatedly use? What angle do they always take? When they see X, their first thought is Y—that Y is their lens.
The lens doesn’t have to be a noun. It can be an action ("reduce to root cause"), a distinction ("natural state vs. cultural state"), or a contrast ("explicit logic vs. implicit structure").
Be ruthless when defining the lens—explain it in one sentence, not an entire chapter.

4. 核心观点 / 结论

4. Core Argument / Conclusion

作者最后得出了什么？
不是"他举了多少例子"，不是"他论证了多少环节"。是：合上书，作者要你带走的那个东西。
往往就一两句话。
检验：把这句话单独抽出来，脱离作者上下文，它还有没有力量？如果只是复述书名，那不是核心观点——是 label。如果只是套话（"事物都是辩证的""万物相联"），那是空话。
核心观点要锋利——能立得住，能反驳，能被引用。
不能软。"作者认为人性是复杂的"是软话——什么不是复杂的？
What did the author ultimately conclude?
It’s not "how many examples they cited" or "how many arguments they made." It’s: when you close the book, what is the one thing the author wants you to take away?
It’s usually just one or two sentences.
Test: If you extract this sentence out of the author’s context, does it still hold weight? If it’s just repeating the book’s title, that’s not a core argument—it’s a label. If it’s just a cliché ("things are dialectical" "everything is connected"), that’s empty talk.
The core argument must be sharp—it must stand on its own, be debatable, and quotable.
No weak statements. "The author believes human nature is complex" is weak—what isn’t complex?

5. 上帝之眼压缩

5. God’s-Eye Compression

整本书几百页，从上帝之眼看，压成几句话是什么？
不是 abstract——abstract 是作者自己写的产品介绍。上帝之眼压缩是你站在作者头上看：作者其实在 say 什么。
三五句话即可。每一句必须挑一根承重梁——基础假设一句、分析框架一句、核心观点一句、再加一两句承接现实的落点。
写完默问：一个没读过这本书的人，看完这几句，能不能 say 他大概知道这本书在干嘛？能 → 过。
The entire book is hundreds of pages; from a god’s-eye view, what does it compress into a few sentences?
It’s not an abstract—abstracts are the author’s own product introduction. God’s-eye compression is you standing above the author and seeing: what is the author really saying?
Three to five sentences are enough. Each sentence must pick a load-bearing beam—one on fundamental assumptions, one on analytical framework, one on core argument, plus one or two sentences connecting to real-world implications.
After writing, ask yourself: Can someone who hasn’t read the book say they roughly know what the book is about after reading these sentences? If yes → pass.

怎么写

How to Write

冷静、简洁、犀利、直白。
不夸作者，不贬作者，不替作者辩护，不替作者道歉。
五段，标号 1-5。段内一气呵成，不绕弯。
每段开头一句立住——这一段要 say 的核心是什么。然后展开两三句把它讲透。不要列点列到天荒地老——一段散文胜过十个 bullet。
短句多。介词砍。"通过 / 借 / 凭 / 在 X 上"——能不用就不用。
中文白话——汪曾祺、王小波、阿城那一路。每写完一段默念出声，卡的地方整句重写（不是抠字）。
完整反翻译腔规则：
~/.claude/PAI/USER/AI_WRITING_PATTERNS.md
（必扫 Layer A）。
不要学术腔。"作者论证了""作者认为"少用——直接说作者干了什么。"X 在第 N 章指出……" 全部杀掉。
Calm, concise, sharp, straightforward.
Don’t praise the author, don’t criticize them, don’t defend them, don’t apologize for them.
Five numbered sections (1-5). Each section should flow smoothly, no detours.
Start each section with a topic sentence stating the core point. Then elaborate in two or three sentences. Don’t overuse bullet points—a single paragraph of prose is better than ten bullets.
Use short sentences. Cut prepositions. Avoid "through / by / relying on / based on X" whenever possible.
Use plain Chinese—similar to the style of Wang Zengqi, Wang Xiaobo, or A Cheng. After writing each section, read it aloud; if it feels awkward, rewrite the whole sentence (don’t just tweak words).
Complete anti-translation rules: 
~/.claude/PAI/USER/AI_WRITING_PATTERNS.md
(must review Layer A).
Avoid academic jargon. Use "the author argued" "the author believes" sparingly—directly state what the author did. Eliminate all phrases like "X pointed out in Chapter N..."

怎么找材料

How to Find Materials

如果用户只给书名——
  • 先 WebSearch / WebFetch 找作者、导言、目录、核心论点的二手解读
  • 拿到核心 argument 后再答
  • 不要凭印象瞎拆——印象拆出来是 wikipedia 风的空话
如果用户给了 PDF / arxiv / 内容——直接读，不用查。
如果是经典书——可以靠 LLM 内部知识，但仍要在心里默问：我说的这条是书里实际有的，还是我编的？编的就删。
If the user only provides a book title—
  • First use WebSearch / WebFetch to find the author, introduction, table of contents, and secondary analyses of the core argument
  • Answer only after obtaining the core argument
  • Don’t teardown based on memory—memory-based teardowns result in Wikipedia-style empty talk
If the user provides a PDF / arxiv / content—read it directly, no need to search.
If it’s a classic book—you can use the LLM’s internal knowledge, but still ask yourself: Is this actually in the book, or did I make it up? Delete anything you made up.

输出

Output

  1. 获取时间戳：
    date +%Y%m%dT%H%M%S
    和 
    date "+%Y-%m-%d %a %H:%M"
  2. 写入 
    ~/Documents/notes/{时间戳}--拆书-{书名}__book.org
  3. org-mode 格式（标题用 
    *
    ，加粗用 
    *bold*
    单星号，禁止 markdown 双星号）
  4. 文件头：
#+TITLE: 拆书：《{书名}》
#+SUBTITLE: {作者} | {一句话核心观点}
#+DATE: [{YYYY-MM-DD Day HH:MM}]
#+FILETAGS: :book:{书的领域 e.g. philosophy/biology/economics}:
#+IDENTIFIER: {YYYYMMDDTHHMMSS}
  1. 正文五段，标 1-5，每段散文。
  2. 报告文件路径给用户。
  1. Get timestamps: 
    date +%Y%m%dT%H%M%S
    and 
    date "+%Y-%m-%d %a %H:%M"
  2. Write to 
    ~/Documents/notes/{timestamp}--book-teardown-{book-title}__book.org
  3. Use org-mode format (titles use 
    *
    , bold uses 
    *bold*
    single asterisks,禁止 markdown 双星号)
  4. File header:
#+TITLE: Book Teardown: 《{book-title}》
#+SUBTITLE: {author} | {one-sentence core argument}
#+DATE: [{YYYY-MM-DD Day HH:MM}]
#+FILETAGS: :book:{book-field e.g. philosophy/biology/economics}:
#+IDENTIFIER: {YYYYMMDDTHHMMSS}
  1. Five prose sections in the body, numbered 1-5.
  2. Report the file path to the user.

红线（每条必过）

Red Lines (must follow every one)

  1. 不夸不贬不替辩护 — 拆书人站作者头上看，不是粉丝也不是反对者
  2. 冷静、简洁、犀利、直白 — 每段读出声不卡
  3. 核心观点不能软 — 软话（"X 是复杂的"）杀掉
  4. 基础假设不展开论证 — 亮出来就行
  5. 上帝之眼压缩 ≤5 句 — 一句一根承重梁
  6. 零学术腔 — "作者论证了 / 认为 / 指出"少用
  7. 反翻译腔 — Layer A 标志词全清；写完默念汪曾祺会这么写吗
  1. No praise, no criticism, no defense — The teardown stands above the author, not as a fan or opponent
  2. Calm, concise, sharp, straightforward — Read each section aloud without awkwardness
  3. Core argument cannot be weak — Eliminate weak statements ("X is complex")
  4. No elaboration on fundamental assumptions — Just present them
  5. God’s-eye compression ≤5 sentences — Each sentence is a load-bearing beam
  6. Zero academic jargon — Use "the author argued / believes / pointed out" sparingly
  7. Anti-translation style — Remove all Layer A marker words; after writing, ask: Would Wang Zengqi write this way?