The one-person company
A book about the AI-first company that comes with one installed.
Every role becomes a repo. Every procedure becomes a playbook. Every decision gets its reasoning written down. AI operates the whole thing on your machine, and you make the calls that matter. That is the entire idea, and this is the manual for building it.
7 parts. 31 chapters. One payment of $17.
Get the book, $17Instant access after payment. 14 days to change your mind. No subscription behind it, now or later.
The operator's manual
HeadcountZero
How one person runs an entire company on AI
7 parts · 31 chapters
Get it now for $17
Secure checkout · Payments by Stripe · USD
14-day refund. No subscription, ever.
Nobody gets replaced. Nobody was hired.
01 / The trap
Not because people are the problem. Because of what happens to everything you teach them.
Here is the pattern behind this book, and it is ours, not a case study borrowed from someone else. Years of hiring. Dozens of people, taught at real cost, one procedure at a time. Every one of them got good. Every one of them eventually left. And on the day they left, the company forgot everything they had learned in it.
Not the files. The files were fine. What walked out was the part nobody writes down: why we stopped doing it the other way, which supplier fails in August, what the number has to be before you kill it. The training was real and the retention of it was zero.
The obvious fix is better hiring. Better onboarding. A wiki nobody reads. We tried all of it. The fix that actually worked was structural: stop keeping the company in people's heads. Put every role in a repo. Put every procedure in a playbook a stranger could execute. Write down the reasoning behind decisions, not just the decisions. Then hand the whole thing to an AI that reads it and runs it, and keep the judgment for yourself.
An employee quits, forgets, sleeps, and takes the reasoning with them. The repo does none of those things. That is not a slogan. It is the difference between a company that resets every eighteen months and one that compounds.
A plan is not a company.
Most AI advice hands you a diagnosis and a to-do list. This hands you the architecture, and then hands you the thing itself.
02 / A test you can run right now
If the answer is yes, you are not AI-first. You bolted AI onto work that is still employee-shaped.
This is the diagnostic the book opens Part II with, because it separates the two things people call the same name. Retrofitting means a person still owns the procedure and asks an AI for help inside it. AI-first means the procedure itself lives in a file, the AI executes it end to end, and the person shows up at the decision.
Retrofitting feels productive and produces almost nothing durable. When the person leaves, the procedure leaves with them, same as always. You have bought a faster typist for a company that still cannot remember anything.
The book does not pretend this is comfortable. Chapter 7 walks the test through four common setups and tells you honestly which parts of yours survive it.
03 / The mechanism
Four moving parts. That is the whole architecture, and the book builds it one part at a time.
One orchestrator that knows what the company is, who does what, and where everything lives. It routes work and verifies results. It stays deliberately thin, because a front door that hoards craft becomes the bottleneck it replaced.
Operations, marketing, finance. Each one is a folder holding its craft, its memory, its playbooks, and its own log of why it does things the way it does. A fresh session opens the folder and acts correctly. If it cannot, the folder is broken, not the AI.
Every procedure written in the same four blocks: the command, how to verify it worked, how to recover when it did not, and the gotcha that bit us last time. Written this way, a procedure runs the same on a bad day as a good one.
Nothing gets full autonomy on day one. A role climbs four rungs: advise only, propose and wait, apply within a cap, apply. You grant a rung, you can revoke it, and the ladder is written down so the AI knows exactly which one it is standing on.
The AI does the two hundred hours. You make the twenty decisions. AI-first, not AI-only.
04 / Where this is going
Four things people actually said and did, with their sources. Nothing here is rounded up.
In a conversation with Alexis Ohanian first circulated in early 2024, and written up by Fortune on 4 February 2024, Sam Altman said:
"In my little group chat with my tech CEO friends there's this betting pool for the first year that there is a one-person billion-dollar company. Which would have been unimaginable without AI and now will happen."
At Code with Claude in May 2025, Anthropic's Dario Amodei put the first billion-dollar solo company at 70 to 80 percent likely by the end of 2026.
Two of the people closest to the models think this is close. Neither of them said it has happened.
In 2015, WhatsApp served roughly 900 million users with about 50 engineers. That is 18 million users per engineer, and it happened before any of this.
Sahil Lavingia reported Gumroad doing $20.7M in revenue and $8.9M in net profit in 2023 with zero full-time employees, running on part-time contractors. Pieter Levels publishes his own numbers monthly and said in 2025, plainly, "I don't write code anymore."
The case everyone cites is a telehealth company the New York Times profiled in April 2026: $401M in its first full year with two employees, financials the Times said it verified directly.
Then the other half surfaced. An FDA warning letter in February 2026, a class action, reporting on fabricated doctor personas and altered patient photos, and an editor's note from the Times saying the profile should have disclosed the scrutiny.
Chapter 3 tells both halves, because the second half is the entire argument for keeping a human at the gate.
No one-person billion-dollar company exists yet. Nobody has done it. Anyone telling you otherwise is selling from a headline they did not read to the end. What has changed is the ceiling, and it is rising faster than anyone planned for. This is the manual for the people racing at it.
05 / Who is behind this
Most books in this category sell you on one person's biography. We cannot do that, and we would rather not. The operating business stays unnamed on purpose. What is shown here is the architecture itself, working.
The OS is the optional upgrade offered after purchase, not part of the $17. It is listed here so you know what the architecture looks like when it is built out, and so nothing about the offer is a surprise later.
A note on where this came from. The architecture was built running a real company, and everything sold here was rebuilt clean for you. No client facts, no borrowed internals, nothing that belongs to somebody else.
06 / What's inside
A chapter proves one thing and then tells you what to go do about it today. No chapter ends by teasing the next one.
Part I
Part II
Part III
Part IV
Part V
Part VI
Part VII
The install step
Not a worksheet, not a reflection prompt. A concrete action in your own folder, doable the day you read it, with nothing else to buy. By the end of Part VI you have a front door, one working role, and one procedure that runs itself.
07 / The line, drawn honestly
A page that says everyone is lying to somebody. Column three is not softened, because the credibility of the first two rests on admitting it.
08 / What else you could do
Including doing nothing, which is a real option and sometimes the right one.
| Option | What it costs | What you end up with |
|---|---|---|
| Do nothing | $0 | The same company, one year older. Fine if it is working. |
| Hire a person | $60,000+ / year | Real capacity, and a copy of your procedures stored in a head that can leave. |
| Agency retainer | $3,000+ / month | Output while you pay. The knowledge lives at their office, not yours. |
| An AI course or prompt pack | $50 to $2,000 | Techniques. Nothing that keeps running after you close the tab. |
| This book | $17, once | The architecture, written down, plus one install step per chapter and a company that remembers what it learned. |
09 / Why $17
The standard move in this category is a cheap book that exists to find people for a subscription. We are telling you now that there is not one.
You have probably bought the $7 book that was a doorway. It works because the book was never the product. The product was the monthly community, the workshop, the mastermind, and the book was the cost of finding out you were the kind of person who buys those.
There is no continuity here. There will not be one. Not a community, not a monthly fee, not a renewal. If we ever build something else, it will be a thing you buy once, and this page will say so.
So $17 has to make sense on its own, and it does. It is priced where a real book is priced, low enough that a wrong guess costs you almost nothing, and high enough that we can keep writing it. The people who want the runnable system will be offered it once, after they have already got what they paid for. Everyone else keeps the book and never hears from us about it again.
One more disclosure, before you spend anything: the book is written for Claude Code, which is Anthropic's product and needs its own subscription. That is not our revenue, and we do not get anything if you subscribe. Entry plans start around $20 a month at the time of writing. You can read the entire book first and decide afterward.
10 / The offer
HeadcountZero, the book
$17
One payment. Not a trial, not a first month.
The one-payment pledge, in plain words.
$17 is charged once. So is everything else on this page. There is no subscription attached to any of it, no trial that converts, and no renewal. Optional add-ons and upgrades are offered, each a single payment, each permanently available at the same price. You can decline all of them and still have the whole book.
11 / If you are not buying today
Roughly weekly. What we changed in the architecture and why, what broke, and which parts of the AI-first story turned out to be nonsense when we checked the source.
No sequence, no countdown, no pitch disguised as a lesson. If a piece of it is worth paying for we will say so once and get back to the note. One click unsubscribes and we do not sell the list.
Free. Nothing is charged, and subscribing does not buy the book.
You are on the list.
The next note goes out with everyone else's. Nothing else happens in your inbox in the meantime.
12 / Questions
No, and the book says so in chapter 23 rather than in a sales bullet. You need to be willing to open a terminal and type instructions in English. The architecture is files and folders with words in them. The parts that look like engineering, the gates and the checks, are shown as complete examples you copy and adjust.
What you do need is the patience to write a procedure down before handing it over. That is the actual skill, and it is not a technical one.
It is Anthropic's tool that lets an AI work directly in your files and run things on your machine. It is not ours and it is not included. It needs its own subscription, entry plans start around $20 a month at the time of writing, and we earn nothing from it either way.
You can read the whole book before subscribing to anything. If you decide the architecture is not for you, you never spend that money.
No. There is no prompt library in it, on purpose. Prompts are the layer that goes stale fastest and the layer that teaches you the least. The book is about structure: how a role is written down, how a procedure is written so it survives a bad day, how autonomy is granted in rungs, and how decisions get their reasoning attached.
That structure has outlived several model generations at this point. The prompts you would have bought last year have not.
That is not what this is, and the book argues against using it that way. The whole framing is for the operator who never scales payroll in the first place. Nobody gets replaced because nobody was hired.
Chapter 29 goes through the public cases where companies cut staff on an AI thesis and then reversed it, and chapter 30 is about the honest modal outcome, which is one operator plus a few specialists, not a solo empire.
It is arguably more useful. A team of five with everything in five heads has the same fragility as a team of fifty, just cheaper. Writing the roles down means people work on top of the company instead of being the company.
Chapter 30 is written specifically for this, including how brains get co-owned by a person and an AI rather than handed over.
The OS is the architecture already built: the front door, three seeded role brains plus the template, the playbooks, the registry files, the mandate ladder, 35 or more files that run in your own terminal. The book teaches you to build that. The OS hands it to you built.
You do not have to buy it. It is offered once after you buy the book, it stays available at the same price forever, and the book is complete without it. We are telling you it exists here so it is not a surprise on the next page.
Checkout charges $17 for the book. Optional add-ons may appear there as checkboxes, never pre-ticked. After you buy, and only after the book is already in your hands, we offer upgrades: the Company OS at $97, which is the architecture already built and running in your own terminal, and a wider edition of it.
Every one of them is a single payment. None of them is required, none of them holds back anything that is in the book, and all of them stay available later at the same price, so there is no reason to decide under pressure.
What you will never be offered is a subscription. There is no monthly fee behind any of this.
Fourteen days, full refund, unconditional. Write to us and it is done, usually within 48 hours, with no form and no hoops.
One honest note: this is early, so we may write back and ask what did not work. The refund is already yours either way and you are free to ignore the question. We ask because the answer is how the next revision gets better.
No. Not a community, not a monthly fee, not a renewal, not now and not later. Every part of this offer is a single payment. If that ever changes for something new, it will be a separate thing you decide to buy, and it will never convert your purchase into a recurring charge.
Because the operating business it came from is not the product and naming it would drag its customers into a sales page. What is sold here is the architecture, rebuilt clean, with none of that company's facts inside it.
You do not have to take that on faith. The book contains the reasoning, the OS contains the files, and both are yours to judge on their own contents. There is no biography here doing work that the material should be doing.
It will, and the architecture assumes it. That is what the mandate ladder is for, what the verify step in every playbook is for, and why anything that commits money or your name stops and waits for you.
Chapter 8 states the failure modes plainly: it forgets between sessions, it acts confidently on a stale instruction, and it fails badly when the written material is thin. The book's answer to all three is structural rather than hopeful.
Part VI is a 90-day path, and it is paced for someone with a company to run rather than a free week. Day zero is an evening. The first working role is the first couple of weeks. After that it is a weekly cadence of run, capture, and write down.
The book is explicit that the second role should not be started until the first one is genuinely reliable. Getting that wrong is the most common way this fails.
Someone is going to be the proof that the prediction was right. They will not have done it with prompts. They will have done it with a company written down well enough that a machine could run it, and the judgment to know which twenty calls stayed theirs.
One payment. Instant access. 14 days to change your mind. No subscription, ever.