New · 2026新书 · 2026

Taming the Intelligence驾驭智能

Chat, Prompt, Workflow, Constitution — A History of Learning to Work with AI对话、提示、工作流、宪法——一部学会与 AI 协作的历史

A timeline-driven history of learning to work with AI — from the night ChatGPT launched to autonomous workflows that run a business while their creator sleeps. Four eras, four rungs on one ladder: Chat, Prompt, Workflow, Constitution.一部以时间线展开的人机协作史——从 ChatGPT 上线的那一夜，到能在创造者熟睡时独自运营业务的自主工作流。四个时代，同一架阶梯的四级刻度：对话、提示、工作流、宪法。

COMING SOON即将上市

1313

Chapters章

20262026

First Edition首版

2022–262022–26

A History of AI WorkAI 协作史

Overview概述

About the Book关于本书

On a Tuesday evening in February 2026, a machine wrote a video script, generated thirty-five images to illustrate it, narrated it in a cloned voice, edited a fifteen-minute video, and published it to five platforms — all while its creator was asleep. The first thing the pipeline ever made entirely on its own became the most successful thing he had ever published.2026 年 2 月的一个普通周二夜晚，一台机器独立写好了视频脚本，生成三十五张配图，用克隆声线念出旁白，剪成十五分钟视频，分发到五个平台——而这一切发生时，它的创造者早已入梦。流水线第一次完全自主跑出的成品，恰好成了他发布过最成功的作品。

The road to that night was nearly four years long — and it is not only one person's story. Between November 2022 and mid-2026, millions climbed the same ladder: from typing a question into ChatGPT and marveling at the answer, to designing autonomous systems that run businesses while their owners dream. The rungs have names: Chat, Prompt, Workflow, Constitution. Each felt like a summit. Each turned out to be a basecamp.通往那一夜的路走了将近四年——而这并非一个人的故事。从 2022 年 11 月到 2026 年中，数百万人沿着同一架阶梯向上攀爬：起点是在 ChatGPT 对话框里敲下问题、对回答惊叹不已，终点则是设计出能在主人熟睡时自主运转的商业系统。这架阶梯有四级刻度：对话、提示、工作流、宪法。每登上一级都以为抵达峰顶，事后才发现不过是下一段攀登的大本营。

This book traces that climb. It is a history, a technical manual, and a practitioner's confession rolled into one. The history is real — built from the funding rounds, the firings, the open-source leaks, the lawsuits, and the tweets that shaped the AI era. The technical material is concrete — built from codebases you can inspect on GitHub. And the confession is the author's own: an angel investor with zero engineering background who vibe-coded his way into running multiple autonomous AI pipelines.本书记录了这场攀登。它兼具断代史、技术手册与实战自白三重属性。历史部分绝非虚构——由融资、裁员、开源泄露、诉讼与推文交织而成，勾勒出 AI 时代的真实轮廓。技术部分拒绝空谈——所有代码均可在 GitHub 查验。至于自白，则是作者本人的坦诚交代：一个毫无工程背景的天使投资人，凭直觉与 AI 协作编程，硬是跑通了多条自主 AI 流水线。

Argument核心论点

What This Book Argues这本书主张什么

Working with AI evolved through four eras — Chat, Prompt, Workflow, Constitution — and every one of them felt like a summit before it became a basecamp.人机协作经历了四个时代——对话、提示、工作流、宪法——每一个时代在当时都像是峰顶，事后才知不过是下一程的大本营。

You do not need to be a programmer or an AI researcher. You need curiosity and a willingness to think in systems.你不必是程序员，也无需身为 AI 研究员。你需要的，是好奇心，以及愿意建立系统思维。

Prompt engineering, RAG, and multi-agent frameworks were each a rung, not a destination — context, not clever wording, is what makes AI reliable.提示词工程、RAG、多智能体框架，都是阶梯上的一级，而非终点——让 AI 真正可靠的，是上下文，而非措辞的花样。

The frontier is no longer a better prompt. It is architecture — a constitution that lets AI design, run, and improve its own workflow.前沿早已不是更好的提示词，而是架构——一部让 AI 自行设计、运行并迭代自身工作流的宪法。

The real leverage is becoming a meta-designer: set the direction, define the boundaries, and let the machine run.真正的杠杆，是成为一名元设计者：设定方向，划定边界，然后让机器自己跑起来。

This history is real and inspectable — funding rounds, firings, lawsuits, and open-source code you can read on GitHub yourself.这段历史真实可查——融资、裁员、诉讼，以及你能亲自在 GitHub 上读到的开源代码。

Structure结构

Inside the Book书中内容

Four eras, each a phase transition in how humans work with AI.四个时代，每一个都是人机协作模式的一次相变。

Part I · Chapters 1–2第一部 · 第 1–2 章

Chat: The Machine Talks Back对话：机器的回应

The ChatGPT explosion, the Google panic, the OpenAI boardroom coup, and the co-founder exodus that seeded the whole industry with competing visions. The most story-driven section — the people and the politics behind AI.ChatGPT 爆红、谷歌恐慌、OpenAI 宫斗，以及为整个行业播下百家争鸣种子的联合创始人出走。全书叙事性最强的一段——AI 背后的人与权力。

Part II · Chapters 3–5第二部 · 第 3–5 章

Prompt: Learning to Speak Machine提示词：学说机器语言

The rise and fall of prompt engineering, the RAG revolution and its limits, and the emergence of context engineering as a discipline. Where the book shifts from narrative to craft — why AI sometimes brilliantly succeeds and sometimes confidently fails.提示词工程的兴衰、RAG 的革命与局限，以及上下文工程作为一门学科的崛起。全书由叙事转向技艺之处——AI 为何时而惊艳，时而自信地犯错。

Part III · Chapters 6–8第三部 · 第 6–8 章

Workflow: Orchestrating the Swarm工作流：编排智能集群

The multi-agent framework wars, the vibe coding movement, the $50-billion editor wars, and the pivot to agentic engineering. A founder is acqui-hired in eight weeks, Musk sues Altman, and Karpathy buries his own coinage.多智能体框架之争、氛围编程运动、五百亿美元编辑器大战，以及向智能体工程的转型。创始人八周内被人才收购，马斯克起诉奥特曼，Karpathy 亲手埋葬了自己创造的术语。

Part IV · Chapters 9–13第四部 · 第 9–13 章

Constitution: Designing the Designer宪法：为设计者立法

The heart of the book. Why AI agents forget, the nine-philosophy architecture that solves it, a real production system dissected to its CLI commands, and the open-source tool handed to you to build your own.全书的核心。AI 智能体为何遗忘、解决它的九条哲学架构、一套被拆解到命令行级别的真实生产系统，以及交到你手中、助你自建系统的开源工具。

Readers读者

Who This Book Is For这本书写给谁

The Operator操盘者 (Founder · Exec · PM) （创始人 · 高管 · 产品）

Wondering how AI agents will reshape your workflows在琢磨 AI 智能体将如何重塑你的业务流程
Wants the architecture that makes autonomy practical想要那套让自主系统真正落地的架构
Start with Part I, then jump to Part IV先读第一部把握全局，再跳读第四部
No code required — systems thinking is enough无需写代码——有系统思维就够了

The Builder构建者 (Developer · Engineer) （开发者 · 工程师）

Already shipping with LLMs every day已经每天用大语言模型交付东西
Wants the path from prompts to multi-agent orchestration想看清从提示词到多智能体编排的演进路线
Chapter 12 dissects a production pipeline to its CLI第十二章把一套生产流水线拆解到命令行级别
Every claim is checkable against public code每条论断都能对照公开代码逐一查验

The Creator创作者 (Writer · Educator · Solo) （写作者 · 教育者 · 独立从业）

Wants AI to handle the mechanical work希望 AI 代劳那些机械重复的活
Curious about vibe coding — Chapter 7对氛围编程心生好奇——第七章
The constitution framework as an on-ramp (Ch 10–13)把宪法框架当作入门坡道（第 10–13 章）
Make AI a genuine partner — even a source of income让 AI 成为真正的伙伴——甚至一份收入来源

The Author作者

About the Author关于作者

Leo Wang is an angel investor, author, and cross-disciplinary thinker who has spent two decades working at the intersection of technology, capital, and human potential.王利杰在科技、资本与人的交汇处，已行走了二十年。他是天使投资人，是作者，也是一位跨越学科边界的思考者。

As the founder of PreAngel Fund, Leo has made over 300 early-stage investments in companies pushing the boundaries of what machines and humans can do — including EHang (autonomous aerial vehicles), Hesai Technology (LiDAR for self-driving systems), DOBOT (collaborative robotics), and Xunce Technology (AI-powered prediction). These weren’t safe bets. They were wagers on first principles: that intelligence — biological and artificial — follows structural patterns, and that understanding those patterns is the closest thing to an edge that an investor can have.创办 PreAngel 基金以来，他投出了三百多个早期项目，都在探索人与机器能力的边界。亿航的自动驾驶飞行器、禾赛的激光雷达、越疆的协作机器人、迅策科技的 AI 预测模型，西井科技的无人港口、无人集卡都在其列。这些从来不是稳妥的买卖，而是对第一性原理的押注：智能，无论生于碳基还是硅基，底层都有迹可循。看透这套结构，便是投资人最可靠的依仗。

That same instinct for structural pattern-recognition led Leo in an unexpected direction. Years of studying artificial intelligence architectures — neural networks, predictive models, objective functions, training loops — made him notice something he couldn’t unsee: the human mind runs on the same architecture. Not metaphorically. Structurally. The prediction engine that powers a large language model is, in its essential logic, the same engine that generates your thoughts, your emotions, and your sense of self. Once he saw the parallel, he couldn’t stop pulling the thread.正是这种对结构的敏感，将他引向了一条意想不到的路。他并非 AI 科研出身，对智能的理解全从应用一端摸来——最早押注无人车与 AI 的投资人，自 ChatGPT 问世便重度使用，又是个不知疲倦的 vibe coder。从这个视角，他渐渐看清一件事，便再也无法视而不见：人心的运作机制，与机器并无二致。这并非比喻，而是结构上的同构。驱动大语言模型的预测引擎，其核心逻辑，与生发你我念头、情绪乃至自我意识的引擎，本是同源。一旦看见这条线索，便只能顺着它一直走下去。

He is the author of The Alternative Investor (originally published in Chinese), a book on venture capital and unconventional thinking, and the creator of MacroAlpha, a daily YouTube channel exploring economics, AI, and the nature of consciousness. He splits his time between Tokyo and California, which he considers less a lifestyle choice and more a permanent experiment in running two cultural operating systems simultaneously.他还写过《投资异类》，谈风险投资与非常规思维。他还主理着 YouTube 频道「宏观阿尔法」，日日更新，探讨经济、AI 与意识的本质。

He can be reached at me@leowang.net, and his work lives at leowang.net.联络他可致信 me@leowang.net，更多作品见于 leowang.net。

Preface序言

Read the Preface先读序言

How to read this book — and why you might not be able to stop. The full opening, in the author's own words.如何阅读本书，以及为何你可能爱不释手。全书开篇原文——作者的亲笔。

On a Tuesday evening in February 2026, a machine wrote a video script about predictive coding and Buddhist consciousness theory, generated thirty-five images to illustrate it, synthesized my cloned voice narrating the script in Mandarin, composed everything into a fifteen-minute video, and published it to five platforms. I was already asleep. The machine had set its own alarm. That video — "The Neuroscience Truth of Meditation" — went on to become my most-watched episode on YouTube, with over two hundred thousand views, and my first to break one million plays on Douyin. The first product my pipeline ever made entirely on its own turned out to be the most successful thing I have ever published. Not bad for something produced while its creator dreamed.

The road to that night was nearly four years long. In March 2022, Shanghai locked down for COVID. Twenty-five million people were sealed inside their apartments. I was one of them. For three months I could not leave my front door, so I decided to use the time to teach myself programming — no AI assistants, no code completion, just tutorials and determination. I posted a handful of clumsy learning videos on my personal YouTube, WeChat, and Douyin accounts. The clips were so crude that I later deleted most of them.

On October 24, 2022, I moved to California. Weeks later — as if by some cosmic arrangement — ChatGPT launched and the world changed overnight. I was right there, in the epicenter. By early 2023, I was learning English and coding simultaneously, and I became one of the first users of GitHub Copilot. The term "vibe coding" would not be coined for another two years, but the experience already felt like magic: an AI completing my half-formed thoughts, filling in syntax I had never memorized, turning a beginner's intent into running code. For someone who could barely read a stack trace, the leap was staggering. Every day brought a new moment of astonishment at what AI could do.

Over the next three years, I built video production workflows that grew steadily more automated — from copy-paste scripts to semi-automated pipelines where I still had to babysit every step. The February 2026 episode was the moment the system finally ran end to end without me. Full autonomy: the first time I woke up to a finished product I had not touched.

Four months later, in June 2026, I made a deliberate trade. I abandoned the Python workflow that had taken years to build and switched to the constitution-based LLM workflow this book describes. On raw efficiency the old pipeline was faster — it still works, and I could fire it up any time. But the old system was brittle in a specific way: every upgrade was painful, every change rippled through tightly coupled code. The new workflow is slower per video, but the moment I have an idea I can hand it to the AI to build. Better yet, the system iterates on itself.

This is not a story about me. Or rather, it is not only a story about me. Between November 2022 and mid-2026, millions of people climbed the same ladder I did — from typing a question into ChatGPT and marveling at the answer, to designing autonomous systems that run businesses while their creators dream. The rungs of that ladder have names: Chat, Prompt, Workflow, Constitution. Each one felt like a destination at the time. Each turned out to be a basecamp.

Who This Book Is For

You do not need to be a programmer. You do not need to be an AI researcher. You need curiosity and a willingness to think in systems.

If you are a founder, executive, or product manager wondering how AI agents will reshape your workflows, start with Part I for the landscape and skip to Part IV for the architecture that makes autonomy practical.

If you are a developer or engineer already building with LLMs, you will find the most value in Parts II and III — the technical evolution from prompts to multi-agent orchestration — and in Chapter 12, which dissects a production pipeline down to its CLI commands.

If you are a creator, writer, educator, or solo operator who wants AI to handle the mechanical work so you can focus on the creative work, Chapter 7 (vibe coding) and Chapters 10–13 (the constitution framework) are your on-ramp.

If you are simply curious about the biggest technological shift since the internet, read it straight through. The story has a narrative arc that rewards sequential reading — but it does not require it.

How This Book Is Organized

The book follows four eras, each corresponding to a phase transition in how humans work with AI:

Part I — Chat: The Machine Talks Back (Chapters 1–2) covers the ChatGPT explosion, the Google panic, the OpenAI boardroom coup, and the co-founder exodus that seeded the entire AI industry with competing visions. This is the most story-driven section. If you care about the people and politics behind AI, start here.

Part II — Prompt: Learning to Speak Machine (Chapters 3–5) traces the rise and fall of prompt engineering, the RAG revolution and its limits, and the emergence of context engineering as a discipline. This is where the book shifts from narrative to craft. If you are already past the "wow, it can chat" phase and want to understand why your AI tools sometimes brilliantly succeed and sometimes confidently fail, this section explains the mechanics.

Part III — Workflow: Orchestrating the Swarm (Chapters 6–8) covers the multi-agent framework wars, the vibe coding movement, the $50-billion editor wars, and the pivot to agentic engineering. This section is where the story and the technology collide most dramatically — a founder gets acqui-hired in eight weeks, Elon Musk sues Sam Altman, and Andrej Karpathy buries his own coinage. If you want to understand where the AI coding revolution stands right now, this is the section.

Part IV — Constitution: Designing the Designer (Chapters 9–13) is the heart of the book. It diagnoses why AI agents forget, presents the nine-philosophy architecture that solves the problem, walks through a real production system in full technical detail, and hands you the open-source tool to build your own. This section is written in first person throughout — I am no longer narrating history but sharing the workshop. If you read nothing else, read Chapter 9 (the problem) and Chapter 13 (the solution). They are designed to stand alone.

How to Skip Around

Every chapter opens with a time period or thematic label and closes with natural bridges to related chapters. You can follow these bridges or ignore them. Here are some suggested paths:

The Story Path — for readers who want the drama: Ch1 → Ch2 → Ch7 → Ch8 (the OpenClaw and Musk v. Altman sections). Four chapters, pure narrative.

The Builder Path — for readers who want to build something: Ch5 → Ch9 → Ch10 → Ch13. Four chapters from context engineering to a working autonomous project.

The Deep-Dive Path — for readers who want the full technical picture: Ch4 → Ch5 → Ch6 → Ch9 → Ch10 → Ch11 → Ch12. Seven chapters tracing the evolution from RAG to a constitution-based production system.

The "Just Show Me" Path — for readers who want the punchline: Ch12 → Ch13. Two chapters: a real system dissected, then the tool to build your own.

A Note on Sources

The factual claims in this book — dates, funding amounts, valuations, quotes, GitHub star counts — have been individually verified through web search against primary sources. Where numbers conflict across sources (and in the fast-moving AI world, they often do), I have used the most authoritative available and noted uncertainty where it matters. Where I describe my own projects, the code is public and inspectable. The Workflow Design Bible is MIT-licensed and lives at Github. The MacroAlpha pipeline itself is not open-source — it contains too much personal configuration and credential wiring to share safely — but that is precisely the point. The Bible is the incubator that generated that pipeline. You do not need my pipeline; you need the tool that builds one tailored to your own channel, your own content philosophy, your own platform mix. Describe what you want, and the Bible scaffolds it for you.

A small caveat: this book was itself produced by the system it describes. The first drafts were written by AI sub-agents working in parallel, then fact-checked, polished, and refined through the same four-step pipeline I teach in Part IV. I wrote the architecture. The machine wrote the words. I edited the words. The machine checked the facts. We went back and forth until neither of us could find anything wrong. If that process sounds strange, it will feel natural by the time you finish Chapter 11.

The Ladder Has No Top

When I started this project, I thought I was writing a history book — a record of what happened between 2022 and 2026. Somewhere around Chapter 10, I realized I was writing a manual for a way of working that did not exist three years ago and will be commonplace three years from now. The ladder I climbed is still being built. Every rung we add reveals the next one.

The question is no longer can AI do this? It is what should I ask it to become?

Let's find out.

— Leo Wang
Tokyo, 2026

2026 年 2 月，一个普通的周二夜晚。一台机器独立完成了以下全套操作：撰写一期关于预测编码与佛教唯识学的视频脚本，生成三十五张配图，用我的克隆声线合成中文旁白，剪辑成十五分钟视频，分发至五个平台。彼时我早已入梦，机器连闹钟都替自己定好了。那期视频——《打坐的神经科学真相》——后来成了我 YouTube 频道播放量最高的内容，超过二十一万次观看，也是我在抖音上第一个破百万播放的作品。我的流水线第一次全自动跑出来的成品，恰好也是我发布过的最成功的作品。

通往那个夜晚的路，走了将近四年。2022 年 3 月，上海因疫情封城。两千五百万人被关在家里，我也是其中之一。三个月出不了门，我决定把这段时间用来自学编程——没有 AI 辅助，没有代码补全，只有教程和死磕。我在个人 YouTube 频道、微信视频号和抖音上发过几条学编程的视频，现在回看确实可笑，后来大部分都删了。

2022 年 10 月 24 日，我搬到了美国加州。几周后——仿佛冥冥之中的安排——ChatGPT 横空出世，世界一夜之间变了。我恰好身处震中。2023 年初，我一边学英语一边继续编程，成了 GitHub Copilot 的第一批用户。「氛围编程」这个词要到两年后才会被发明，但那种体验已经像魔法一样：AI 替你补全半成型的想法，填上你从未背过的语法，把初学者的意图变成可运行的代码。对于一个连报错信息都读不太懂的人来说，这个跃迁是惊人的。每天都在惊叹 AI 又能做到什么新事情了。

此后三年，我的视频生产工作流一步步走向自动化——从最初的手动复制粘贴，到半自动化的流水线，但每一步仍需要我亲自盯着。2026 年 2 月那期视频，是系统第一次真正端到端地独立运行。完全自主：我第一次醒来，发现一个自己完全没有碰过的成品已经上线了。

四个月后，2026 年 6 月，我做了一个刻意的取舍。我放弃了花费数年搭建的 Python 工作流，切换到本书所描述的基于宪法的大语言模型工作流。论单纯的效率，老流水线更快——它至今仍然在线，随时可以启动。但老系统有一个致命的脆弱性：每一次升级都很痛苦，每一处改动都会牵一发而动全身。新工作流单期视频慢一些，但我有了想法就能马上安排 AI 开发。更妙的是，它还会自我迭代。

这不仅仅是我的故事，甚至可以说，主角并非我一人。从 2022 年 11 月到 2026 年中，数百万人沿着同一条阶梯向上攀爬。起点是在 ChatGPT 对话框里敲下问题、对回答惊叹不已；终点则是设计出能在创造者熟睡时自主运转的商业系统。这条阶梯有四个刻度：对话、提示词、工作流、宪法。每登上一级，当时都以为抵达峰顶，事后才发现不过是下一段攀登的大本营。

本书记录了这场攀登。它兼具断代史、技术手册与实战自白三重属性。历史部分绝非虚构，由融资、裁员、开源泄露、诉讼与推文交织而成，勾勒出 AI 时代的真实轮廓。技术部分拒绝空谈，所有代码均可在 GitHub 查验。至于自白，则是我个人的坦诚交代：一个毫无工程背景的天使投资人，凭直觉与 AI 协作编程，硬是跑通了多条自主 AI 流水线。我会和盘托出事情经过、踩过的坑，以及换来的教训。

谁适合读这本书

你不必是程序员，也无需身为 AI 研究员，只要有好奇心，愿意建立系统思维即可。

若你是创始人、高管或产品经理，关心 AI 智能体如何重塑业务流程，建议先读第一部分把握全局，再跳至第四部分，了解让自主系统落地的架构设计。

若你是开发者或工程师，已在使用大语言模型，第二、三部分对你价值最高。这部分梳理了从提示词工程到多智能体编排的技术演进。第十二章更是将一套生产级流水线拆解到了命令行级别。

若你是创作者、写作者、教育者或独立从业者，希望 AI 代劳机械性工作，以便专注创作，第七章的「氛围编程」与第十至十三章的「宪法框架」便是最佳切入点。

若你只是对这场互联网以来最大的技术变革感到好奇，不妨从头读到尾。全书叙事脉络连贯，顺序阅读体验更佳，但跳读也无妨。

全书结构

本书按四个时代划分，对应人机协作模式的四次相变。

第一部分「对话：机器的回应」（第 1—2 章）聚焦 ChatGPT 爆红、谷歌恐慌、OpenAI 宫斗及联合创始人出走。这些事件播下了 AI 行业百家争鸣的种子。本章叙事性最强，若对 AI 背后的人与政治博弈感兴趣，请由此开始。

第二部分「提示词：学说机器语言」（第 3—5 章）回顾了提示词工程的兴衰、RAG 技术的革命与局限，以及上下文工程作为独立学科的兴起。全书至此由叙事转向技艺。若你已过了「AI 居然能聊天」的新奇阶段，想弄明白 AI 工具为何时而惊艳、时而自信地犯错，这部分揭示了底层机理。

第三部分「工作流：编排智能集群」（第 6—8 章）涵盖了多智能体框架之争、「氛围编程」运动、五百亿美元编辑器大战，以及向智能体工程的转型。此处故事与技术激烈碰撞：创始人八周内被人才收购，马斯克起诉奥特曼，Karpathy 亲手埋葬了自己创造的术语。想了解 AI 编程革命现状，此部分不容错过。

第四部分「宪法：为设计者立法」（第 9—13 章）是全书核心。这部分诊断了 AI 智能体遗忘问题的根源，提出九条哲学架构予以解决，详解一套真实生产系统的技术细节，并提供开源工具助你自建系统。此部分全程采用第一人称，我不再是历史讲述者，而是带你走进工坊的实操者。若时间有限，仅读第九章（问题）与第十三章（解法）亦可，这两章彼此独立。

跳读指南

每章开头标有时间段或主题标签，结尾设有通往相关章节的衔接线索。你可以顺着线索走，也可以另辟蹊径。以下推荐几条路径：

故事线：适合想看风云变幻的读者。第 1 章 → 第 2 章 → 第 7 章 → 第 8 章（含 OpenClaw 与马斯克诉奥特曼案）。四章纯叙事，跌宕起伏。

构建线：适合想动手造物的读者。第 5 章 → 第 9 章 → 第 10 章 → 第 13 章。四章带你从上下文工程走到可运行的自主项目。

深潜线：适合想掌握完整技术图景的读者。第 4 章 → 第 5 章 → 第 6 章 → 第 9 章 → 第 10 章 → 第 11 章 → 第 12 章。七章篇幅，追踪从 RAG 到基于宪法的生产系统的完整演化。

速成线：适合只想看结论的读者。第 12 章 → 第 13 章。两章内容：拆解真实系统，交付自建工具。

关于资料来源

书中事实陈述，包括日期、融资额、估值、引语、GitHub 星标数等，均经网络检索与一手信源交叉核实。AI 领域瞬息万变，数据冲突在所难免，遇此情况，我采信最权威来源，并在关键处标注不确定性。凡涉及我个人项目，代码全部公开可查。《工作流设计圣经》采用 MIT 协议，托管于 Github。宏观阿尔法的流水线本身并未开源——里面涉及太多个人配置和隐私，直接公开并不合适——但这恰恰说明了问题。圣经是生成那条流水线的底层孵化器。你不需要复制我的流水线；你需要的是能根据你自己的频道定位、内容理念和平台组合，定制化生成专属流水线的工具。把你的想法告诉它，它会帮你搭好骨架。

需特别说明：本书本身即由书中所述的系统生成。初稿由多个 AI 子智能体并行撰写，随后经事实核查、润色打磨，流程与第四部分教授的四步法完全一致。我设计架构，机器生成文字；我编辑文字，机器复核事实。双方反复迭代，直至无可挑剔。若你觉得这过程离奇，读到第十一章时，便会习以为常。