Preface: The Last Scarcity

1. A group diving #

Let’s look at a number first. The cost of AI inference to reach the same level of capability was US$20 per million tokens in November 2022, and US$0.07 in October 2024. Within two years, it has dropped to about one two-hundred-and-eightieth of its original level (Stanford HAI’s measurement basis, every number cited in this book will be cited like this).

This 280-fold drop is not a piece of technical news, but a price list of the times. Writing code, doing design, producing reports, responding to emails, translating, and editing were all execution tasks that used to rely on human labor-hour pricing, but the price curve is collectively diving. Knowledge acquisition has also plummeted: introductory knowledge in any field is now only a conversation away from anyone. The "competence" that your company is proud of, and the "proficiency" on your resume are depreciating at a rate of a hundred times in two years.

But one line hangs high on the price list: The ability to judge what to do and what not to do in the face of unprecedented situations. AI has lowered the price of answers to the floor, but raised the price of asking the right questions and selecting the right criteria to the ceiling.

This book is about this unbroken line. The name I gave it is written on the title of this preface: the last scarcity, which is also the name of the second part of this book. And Rewrite the DNA on the cover names the action you want to do to it: rewrite the organization to be able to produce this scarcity by itself.

2. Why is it "last" #

Scarcity determines the price, and the last scarcity determines the final winner. The history of business is a history of relays of scarce resources: landlords win in the era of scarcity of land, bankers win in the era of scarcity of capital, and those who know first win in the era of scarcity of information; in the era of scarcity of execution, which is the era that has just ended, companies with strong organizational and mobilization capabilities win. The pattern for each is the same: once a resource becomes available to everyone, it drops out of competition and the baton is passed to the next scarcity.

The baton of execution is being handed over, and the next question is to whom it goes.

Let’s first eliminate the strongest candidate answer: computing power and energy. They are indeed tight, and they are indeed expensive, but they have a fatal attribute: they can be bought with money. Things that everyone can buy will not constitute a competitive gap; the gap will always come from things that money cannot buy (the scarcity of money itself will be dealt with separately in Chapter 5: ). You can buy computing power, but not judgment; you can buy models, but you cannot buy the ability to know what to ask the model.

So the baton is handed over to the ability to: face a problem that has never been done before, quickly identify what standards you lack, borrow a set of proven standards from where, judge whether it can be transferred to your own scenario, and then verify and correct it at the minimum cost. I call this ability adaptive insight. Calling it the "final" scarcity is not a rhetorical exaggeration, but a testable inference: execution, knowledge, content, and code are all on the zero list, and every time adaptive insight is approached by AI, the unit price of the part that has not been replaced increases by one point. It’s the only thing on this list that gets more expensive the more you consume it.

Three or four words, talking about the same thing #

The most common mistake a book about concepts makes is to have the same thing appear under different names. Therefore, I will clarify the relationship between the four most important words in the book at once in the preface, and they will not change again in the rest of the text:

**"The last scarcity" is its name; adaptive insight is its form in the individual (a capability); judgment is its form in the organization (an asset); standards are its written form (an executable document). **

Four words, talking about different states of the same thing, like water, ice and steam. In individuals, it manifests itself as adaptive insight, which is the four actions of discovering gaps, borrowing standards, transferring judgment, and verifying and correcting (expanded in Chapter 3). After settling into the organization, it becomes a judgment asset that can be reused, passed on, and will not evaporate when employees leave (Chapter 4 to expand). And the only way it can settle is to be written into a standard, that is, an executable document of "conditions plus actions plus acceptance" (expanded in Chapter 6). Whenever you see these four words in the following text, please automatically convert them into the same thing.

4. How to read this book #

There are thirteen chapters in the book, divided into four parts. The structure is the order of argument:

Diagnosis (Chapter 1-5): Why does the execution price approach zero, why most companies have not become stronger after using AI (paradox of abundance), what is competition in the era of adaptive insight, what is organizational DNA, and how to find limiting resources. After reading this, you should be able to answer: Where is my organization stuck?

Production (Chapter 6-9): How to create scarce things, including how to establish standards engine, how to expand one's own options with Clarity Method, how to narrow other people's options with Interference Method, and how to assemble the three into consensus engine with shared context. This is the methodological core of the book.Three battlefields (Chapter 10-12): How to implement new genes in talent structure, personal time, and product path.

Manual (Chapter 13): Assemble the previous twelve chapters into a 90-day execution checklist, with eight acceptance questions.

I abide by the same contract in each chapter: the conclusion is condensed into a repeatable sentence; the skeleton diagram can be drawn on the whiteboard; the cases are given in pairs, with a loser for each winner, because the methodology of only looking at survivors is divination; and there must be a "What to Do Monday Morning" at the end. In addition, there are two honesty clauses throughout the book: all unaudited data from the company will be marked with quality; all evidence that I have not verified or found yet will be marked directly as "to be added". I would rather let you see the gap than cover it up with adjectives.

There is one more thing that should be explained before you start reading, rather than letting you find out after reading Chapter 8: I wrote this book and implemented the methods in the book, and the target is you. The title of the book is narrowing the vocabulary you use to describe this problem, the four words in the third section are unifying the language between you and me, and the "Monday morning" in each chapter is forcing the conclusion into action. What you hold in your hand is both an instruction manual and a running prototype. Every standard in the book is implemented every day in my own company, bioby.ai; I do not promise that they are applicable to all organizations, but I promise that there is no one that I dare not implement (the author's introduction page has the full version of this promise).

Paper books will be frozen on the day of submission, but the organization will not. So I put the running version of this method on Regenic.ai: the entire book is free and open to the public, and cases, tools, version revisions, and new enterprise practices continue to be verified there. There is only one sentence between the three: Rewrite the DNA is the frozen version of this methodology at the moment of publication, Regenic.ai is the continuously updated running version, and bioby.ai is the first experimental site.

5. Two questions before starting #

No need to wait until you finish reading, you can do your first self-test now:

1. **If tomorrow the cost of execution and knowledge acquisition dropped dramatically at the same time (which is actually happening), what value would my company have left? **
2. **When encountering a problem that has never been done before, can we quickly find, migrate and verify a set of standards instead of starting from scratch or waiting for a capable person? **

If the first question silences you, it means that your value is based on the resource that is approaching zero; if the answer to the second question is no, it means that you do not yet own the resource that will not approach zero. If neither answer is ideal, this book is for you.

It’s best to start with skepticism. Chapter 1: First deal with the most reasonable doubt you have at the moment: Will the execution price really approach zero?

Chapter Acceptance Self-Check (special contract in the preface) #

1. The conclusion can be restated in one sentence ✓.
2. Framework diagram ✓ (Multiple cost curves dive collectively, judgment hangs high, and adaptive insight is marked with four actions).
3. Case contract ✓: The preface does not include the case, only a set of verified data (, Stanford HAI, 280 times); the strongest rhetorical question (computing power/energy) has been actively responded to.
4. The relationship between the four core words is explained clearly at once ✓ (name/adaptive insight/judgment/standard, no changes will be made later).
5. Reading map of the whole book + contract for each chapter + two honesty clauses + explanation that the book itself is a prototype ✓; two questions at the end + doubtful entrance handover Chapter 1 ✓.