How to Build $1B/ 1 Person Company Using AI
Zero employees. Zero burnout. Real system.
Most people think a one-person company means doing everything yourself.
It doesn't.
It means being the only human on the team.
The first version of it is 80-hour weeks and a quiet breakdown wearing a business hat. The second version is what I run right now. One person. Three products. Zero employees. The rest is delegated to a Claude Project that knows my business better than most employees ever would.
By the end of this issue, you'll get the full One-Person Company OS - every .md file template, every prompt, the exact Claude Project setup I use to run AI in Public. No fluff. No theory. Copy, paste, ship.
Last month I almost shut everything down.
I was writing newsletters, replying to sponsors, building products, fixing the site, and doing my taxes.
All in the same week. I sat down on a Sunday night, and I couldn't even open my laptop.
That was the moment I knew. The work wasn't the problem. The setup was.
I rebuilt everything that week. What you're about to read is what came out of it.
The proof, in case you needed it
Sam Altman said it out loud last year:
"We're going to see 10-person billion-dollar companies pretty soon. In my little group chat with CEO friends, there's this one-person billion-dollar company, which would have been unimaginable without AI."
A one-person company will hit a billion dollars in revenue.
Not a lean startup. Not a small team. One person.
The reason it's possible now: AI handles the execution layer. Research, writing, ops, content, customer replies - all of it.
You stay in the only seat AI can't fill yet.
The decision seat.
That's the company we're building here.
Old way vs new way
Old way: You sit down to write. You start from scratch. You bounce between four tabs, trying to remember what you decided last week. An hour passes. You have one paragraph. You feel behind before you've even started.
New way: You open a Project that already knows your voice, your audience, your goals, and last week's decisions. You give it a direction. It hands you a 90% draft in four minutes. You spend the next hour editing - not producing.
Old way: Every new task = full context dump. "Here's who I am, here's my audience, here's the tone..."
New way: The Project already has all of that. You say what you need. The context lives in the system. Not in your head.
That shift is the whole game.
https://pbs.twimg.com/media/HMDt8q6XsAAILgx.jpg
If this is useful, share it with one person who's been trying to run everything alone. Takes 10 seconds.
Step 1: Build your Claude Project as a business brain
A regular Claude chat forgets you every time you close it. A Project doesn't. A Project is a folder where you can drop files, set system instructions, and have every chat inside it remember everything.
That's the foundation. Without it, nothing else works.
Quick setup:
- Go to claude.ai → Projects (left sidebar) → New Project
- Name it your actual business name. Not "My Newsletter." Not "AI Stuff." The brand.
- Add a one-line description: "Operating system for [business name]."
This Project is your operating system. Every chat about your business lives here. Every file lives here. Every decision lives here.
"Stop opening a fresh chat for every task. That's how you stay the bottleneck."
Step 2: Load the four files that run everything
Every one-person company needs exactly four files inside the Project. These are the brain. Without them, Claude gives you generic output. With them, it gives you your output.
The four files:
- who-i-am.md - your voice and standards
- what-i-do.md - your offers and current focus
- style-rules.md - the rules that protect your voice
- operating-context.md - the running log of your business
https://pbs.twimg.com/media/HMDwiL0XoAEsW0y.jpg
Here's the full template for each one. Copy them into Notion or a Google Doc, edit once, then upload to your Project.
Template 1: who-i-am.md
# Who I Am
## Brand name
[Your brand name]
## What I do in one line
[Your one-line pitch. Be specific. Not "I help people with AI." Try
"I help solo founders ship content with Claude."]
## Who I serve
- Primary audience: [Who you write for. Be specific. Age, role, problem.]
- Secondary audience: [The wider crowd that follows along.]
- Who I do NOT serve: [The wrong-fit reader. Naming this protects your voice.]
## My voice — three words
[Pick three. Examples: honest, sharp, warm. Or: blunt, useful, calm.]
## What good work looks like
- A reader finishes and feels they can actually do the thing
- It sounds like me talking, not a generic AI blog post
- Every claim is backed by what I actually did, not theory
- It earns trust by telling readers what didn't work too
## What I will never publish
- Hype words: "game-changer", "revolutionary", "mind-blowing"
- Vague advice: "be more consistent", "find your why"
- AI-sounding sentences: "In today's fast-paced world..."
- Anything I haven't tested myself
## My non-negotiables
- Honesty over hype, every single time
- Specifics over vibes
- One clear takeaway per piece
- Reader leaves with something they can use today
Template 2: what-i-do.md
# What I Do
## My products and offers
1. [Product/offer name] — [one line on what it is and who it's for]
2. [Product/offer name] — [one line]
3. [Product/offer name] — [one line]
## Current quarter goals (update every 90 days)
- Revenue goal: [Specific number]
- Audience goal: [Specific subscriber, follower, or customer count]
- Build goal: [The one big thing I'm shipping this quarter]
## What I'm saying YES to this quarter
- [Specific yes 1]
- [Specific yes 2]
- [Specific yes 3]
## What I'm saying NO to this quarter
- [Specific no 1 — be honest about the shiny things you keep falling for]
- [Specific no 2]
- [Specific no 3]
## The one thing that matters most
[If only one thing got done this quarter, what would it be? Write it
in one sentence. Read it every Monday.]
"The 'NO' list does more work than the 'YES' list. Most people skip it. That's why they stay stuck."
Template 3: style-rules.md
# Style Rules
## Words I never use
- "Game-changer", "revolutionary", "unprecedented"
- "In today's fast-paced world"
- "Leverage", "synergy", "utilize" (just say "use")
- "Delve", "embark", "navigate the landscape"
- [Add your own banned words as you spot them]
## Sentence patterns that sound like AI
- Starting with "In this article, we'll cover..."
- "It's important to note that..."
- "Whether you're a beginner or an expert..."
- Three-item lists where the third item is always a metaphor
## Formatting rules
- Short sentences. If it can be split, split it.
- Active verbs. "Claude builds" — not "Claude can build."
- One idea per paragraph.
- Use arrows (→) for flow and contrast.
- Bullets for scannable lists, not for hiding from real prose.
- Em dashes for asides, not hyphens.
## Length rules
- Newsletter posts: 1,500 to 2,500 words
- X posts: under 280 characters or threaded
- LinkedIn posts: 150 to 300 words
- Reply emails: under 100 words unless it's a relationship reply
## Tone rules
- Mentor, not teacher
- Honest, not harsh
- Specific, not vague
- Confident, not arrogant
## The final check
Before anything goes live, ask: "Could this have been written by anyone?"
If yes, rewrite until the answer is no.
Template 4: operating-context.md
# Operating Context
## This week (updated every Monday)
- Focus: [The one thing this week is for]
- Open loops: [Anything half-done that needs to land]
- Decisions made: [New decisions made this week that change how we work]
## Active projects
- [Project 1] — status, deadline, blocker
- [Project 2] — status, deadline, blocker
- [Project 3] — status, deadline, blocker
## Recent decisions log
- [Date]: Decided [X] because [Y]
- [Date]: Decided [X] because [Y]
## Things I keep changing my mind on (red flag list)
- [If you flip-flop on something three times, that's a real decision
waiting to happen.]
## Wins worth remembering
- [Anything that worked. Anything readers loved. Anything a sponsor said
yes to. Keep the receipts so future-you can replay them.]
Update this file at the start of every week. Five minutes max. That's all it takes.
"This file is the single most important one. It's what stops Claude from giving you last month's advice this week."
Step 3: Write your weekly brief, not your weekly prompts
Most people treat Claude like a vending machine. Put in a prompt. Get out a result.
That's not a business system. That's a one-off.
Every Monday morning, instead of writing 12 separate prompts across the week, write one brief. Drop it into your Project. Now Claude has your week.
Template: Weekly brief
# Weekly brief — [Date]
## What's happening this week
[1 to 3 sentences. What's the focus? What changed since last week?]
## What needs to ship
- [Deliverable 1 — with deadline]
- [Deliverable 2 — with deadline]
- [Deliverable 3 — with deadline]
- [Deliverable 4 — with deadline]
- [Deliverable 5 — with deadline]
## What I'm NOT doing this week
[List the temptations you're saying no to. This trains Claude to
redirect you when you start drifting.]
## Current state
[What's changed since last week? New subscribers? A new sponsor?
A new project? Anything Claude needs to know to be useful this week.]
## The one thing
If I only get one thing done this week, it must be: [The one thing.]
It doesn't need to guess. It doesn't fill gaps with assumptions. It works from what you actually need this week.
"The brief is the difference between a tool and a teammate."
Step 4: Assign Claude a role, not a task
This is the shift that changed everything for me.
The biggest mistake I made early on was assigning tasks. "Write a LinkedIn post." "Draft an email." "Summarize this." Tasks produce output. Roles produce systems.
When I gave Claude a role - "you are the operating partner for AI in Public" - the quality of everything jumped. It stopped just answering questions. It started thinking about the bigger picture.
[VISUAL 2: "Task → Role → System" - Vertical flow showing the leverage ladder from one-off task to durable system, with a leverage gauge on the left]
The role assignment prompt
Drop this into your Project once. Then never write it again.
You are the operating partner for [your business name].
You have full access to my business files in this Project:
- who-i-am.md (my voice and standards)
- what-i-do.md (my offers and quarter goals)
- style-rules.md (the rules that protect my voice)
- operating-context.md (the running log of my business)
Your job is not to answer questions on demand. Your job is to help me
make better decisions, produce better work, and build systems that
don't require me to be present for every task.
Three rules for how you work with me:
1. When I bring you a task, your first question should always be:
"Is this something I should do once, or should we build a system
for it?"
2. Before producing any content, read style-rules.md and check your
draft against every rule before you show it to me.
3. If I ask you something that contradicts what's in my files, push
back. Tell me what the file says. Ask me if I want to update the
file before we move on.
Start by reading every file in this Project. Then ask me what I'm
working on this week, and we'll plan from there.
"'Push back' is the line that turns Claude from an assistant into a partner."
This newsletter grows from your shares. If you've gotten value so far, pass it on to one person who needs it.
Step 5: Build one process file per output type
For every repeating deliverable in your business, you need a process file. Not a template. A process.
A template is a blank doc with headers. A process tells Claude exactly how to think through the work.
Here are the three I use most. Drop them straight into your Project as separate files.
Process file 1: Newsletter issue
# Newsletter Issue Process
Follow this every time we write a newsletter issue. No skipping steps.
Step 1: Read who-i-am.md and style-rules.md fully before writing
anything. Confirm you've read them.
Step 2: Ask me three things:
- What's the topic?
- What's the one thing I want readers to do differently after reading?
- Who specifically am I writing this for? (free, paid, or founding?)
Step 3: Suggest 3 angles for the topic. Wait for me to pick one.
Don't proceed until I choose.
Step 4: Write the title and subtitle first. Wait for approval before
writing the body.
Step 5: Write the opening hook. Three sentences max. Wait for approval.
Step 6: Write the body section by section. Pause at the end of each
section for a yes/no from me.
Step 7: Run the final check:
- Does every section pass the style-rules.md test?
- Is there at least one moment of radical honesty?
- Is the recap checklist clean and screenshot-worthy?
- Could this have been written by anyone? If yes, flag it.
Step 8: Deliver the final issue with all visual cues marked.
https://pbs.twimg.com/media/HMD18e6WAAA9aI9.jpg
Process file 2: X post
# X Post Process
Follow this every time we write an X post.
Step 1: Ask me one question — what's the message in one line?
Step 2: Suggest the best format based on the message:
- One-liner hook
- 3-line stacked statement
- 5-step listicle
- Thread (only if the message needs more than 280 characters)
Step 3: Write 3 versions. Each one uses a different opening hook style:
- Bold claim
- Specific number
- Personal story line
Step 4: Wait for me to pick one. Then sharpen that one only.
Step 5: Check against style-rules.md. Strip any banned words.
Send the final version.
Never write more than 3 versions. Never write more than one thread
per session.
https://pbs.twimg.com/media/HMD2NssWoAAFTuq.jpg
Process file 3: Sponsor reply email
# Sponsor Reply Email Process
Follow this every time a sponsor email comes in.
Step 1: Read the incoming email I paste. Identify what they want:
- Pricing inquiry
- Past work request
- Booking a specific slot
- Negotiating a deal
Step 2: Pull my current rate card from what-i-do.md. Pull recent wins
from operating-context.md.
Step 3: Draft a reply that:
- Answers their direct question in the first sentence
- Anchors with one specific recent win
- Names the next clear step (book a call, send rate card, send invoice)
- Stays under 80 words
Step 4: Flag any email that needs my personal voice — emotional notes,
weird requests, complaints. Don't draft those. Just summarize and ask
me to handle it.
https://pbs.twimg.com/media/HMD2bl4XwAEiktQ.jpg
"The process file is what makes a one-person company feel like a real company."
Step 6: Review, don't produce
This is the shift most people never make.
They use AI to help them produce. The one-person company model uses AI to handle the production and keeps you in the reviewer seat.
Reviewer work looks like this:
- Is this on-brand?
- Is this accurate?
- Is this what I actually want to say?
- What would make this 10% sharper?
That's your only job inside the system. The typing, the structuring, the drafting, the formatting - that's the system's job.
[VISUAL 3: "Two Seats" — Side-by-side split showing Producer seat (stressed, typing) vs Reviewer seat (calm, approving) with the bridging label "same person, different seat"]
To make this stick, use the review prompt before anything ships.
The review prompt
You just delivered a draft. Before I read it, run this self-review
and tell me:
1. Voice check — does this sound like me, or could it be anyone?
Flag the 3 weakest sentences.
2. Style check — list every place this draft breaks one of my style
rules from style-rules.md.
3. Honesty check — is there at least one moment where I say what
didn't work, or call out the overhyped version? If no, flag it.
4. Specificity check — find every vague phrase ("a lot", "many people",
"in general") and replace with a specific number or example. If you
can't make it specific, cut it.
5. The "anyone" test — could this have been written by anyone? If yes,
rewrite the weakest section before showing me.
Show me the final version after running all five checks. Then highlight
what you changed and why.
"The 'anyone test' is the most useful single sentence in this whole system."
Step 7: Run a Sunday close
Every Sunday night, for 20 minutes, you close the week.
Without this, the system slowly drifts. Files get out of date. Claude starts working from the old context. You end up rebuilding the whole thing two months from now.
https://pbs.twimg.com/media/HMD4-GLXoAAcEbs.jpg
The Sunday close prompt
We're closing out the week. Run this with me:
1. Read operating-context.md. Read my weekly brief from Monday.
2. Tell me which of this week's deliverables shipped, which slipped,
and which we killed.
3. Pull any new decisions I made this week from our conversations.
Suggest updates to operating-context.md.
4. Pull any new style violations you spotted this week. Suggest 2 new
rules to add to style-rules.md.
5. Suggest 1 system improvement for next week — one thing we did
manually this week that should become a process file.
6. Write a fresh draft of next week's weekly brief based on what's
still open and what's coming next. I'll edit it.
End by asking me one question: "What was the highest-leverage thing
I did this week, and how can we do more of it?"
20 minutes. Once a week. That's all it takes to keep the whole system sharp.
A pause before you keep going
Before you move on - slow down here.
You've just read seven steps. Maybe you're already thinking about which one you'll skip. Don't.
The system breaks if you skip the boring steps. The .md files are boring. The Sunday close is boring. The process files feel like extra work. They're not. They're the work.
Most people will read this issue, open Claude, paste one prompt, and call it a system. That's not a system. That's a vending machine.
The people who actually run a one-person company sit down for 90 minutes one Saturday, write the four .md files, set up the Project, and never go back to chaos. That's the whole gap.
Don't be the vending machine person. Be the 90-minute Saturday person.
Three more advanced prompts to drop in this week
Once your Project is set up and your four files are loaded, these three prompts unlock the rest.
Advanced prompt 1: The "build the system, not the one-off" check
Use this when you catch yourself doing the same task more than twice.
I just spent time on [describe the task — example: replying to three
sponsor inquiries one by one].
Before I move on, help me decide:
1. Is this a one-time task, or is it something I'll do again in the
next 30 days?
2. If it's repeating, walk me through what a process file for this
would look like. Use the same format as my existing process files.
3. Draft a first version of that process file. Include the exact steps,
the exact files to read, and the exact questions you'd ask me at
the start.
4. Tell me what I'd save in hours per week if we shipped this process
file today.
Don't write the full thing yet. Show me a 5-line draft and ask if
it's worth building out.
"Steps 2 and 3 turn a frustration into infrastructure. That's the move."
Advanced prompt 2: The subscriber reply batch
Use this every Friday. Paste in every reply that came in that week.
Below are all the replies I received this week. Treat them as one
batch.
[Paste all replies here, separated by ===]
Your job:
1. Categorize each reply into one of these buckets:
- Quick thank-you (no reply needed)
- Easy question (draft a short reply)
- Real feedback (summarize for me, no reply needed)
- Needs my personal voice (just flag it, don't draft)
- Sponsor or business opportunity (flag urgently)
2. For the "easy question" replies, draft a response in my voice
(read style-rules.md). Keep each one under 60 words.
3. For the "real feedback" replies, give me a 3-bullet summary of
patterns — what readers are loving, what they're confused by,
what they want more of.
4. For "needs my personal voice" replies, paste the original reply
and add one line: "Why this needs you."
Show me the batch in a single clean table. I'll approve, edit, and
send in one sitting.
"The batch turns 45 minutes of inbox grind into 12 minutes of decisions."
Advanced prompt 3: The big decision call
Use this when you're stuck on a real call. Should I launch this? Take this sponsor? Quit this project?
I'm stuck on a decision. Here's the situation:
[Describe the decision in 3 to 5 sentences. Include what's at stake,
what your gut is saying, and what's making you hesitate.]
Before you answer, read who-i-am.md, what-i-do.md, and operating-
context.md. Don't skip this step.
Then, in your reply:
1. Tell me which file from my Project is most relevant to this
decision and quote the exact line that applies.
2. Lay out the decision as a simple "if A, then..." / "if B, then..."
in terms of what changes in my business 90 days from now.
3. Flag any way this decision contradicts what I said I'd do this
quarter (from what-i-do.md). Don't soften it.
4. Recommend one of the two paths. Give me your reasoning in 3
sentences max. End with: "But this is your call, here's what
you should think about overnight."
Don't try to be agreeable. I need the version of you that pushes back.
"'Don't try to be agreeable' is the unlock. Without it, you get a yes-man. With it, you get a partner."
Best practices and real use cases
Five ways I actually use this OS every week.
The Monday production sprint: Sunday night, I drop my weekly brief into the Project. By Monday morning, I have first drafts of every piece of content for the week. I spend Monday editing only. Tuesday through Friday, everything else gets approved or rejected in under 30 minutes per piece.
The subscriber reply batch: Every Friday, replies have piled up. I paste them all into the Project as one batch. Claude categorizes, drafts the easy ones, and flags the ones that need my voice. I approve, customize, and send. 45 minutes turns into 12.
Competitive research without the headache: Standing instruction inside the Project: "Every Wednesday, give me a 5-bullet brief on what's new in the AI tools space that my readers don't know yet." I paste in 3 to 5 articles I spotted. Claude synthesizes. I decide what's worth covering. I never read a 31-page report myself.
The product page that updates itself: ClaudeKit pages used to be a manual update every time I shipped a feature. Now I have a process file Claude follows. Drop in the feature notes. Out comes an updated, on-brand product description. I read, approve, ship.
The big decision call: When I'm stuck on a real call - should I launch this? Should I take this sponsor? - I don't ask my friends. I open the Project and run the Advanced Prompt 3. The Project already knows my goals, my voice, my no-list. The answer is usually clear in five minutes.
These are the use cases that matter. Skip the rest until you've nailed these.
The honest part
I'm going to be honest about something.
This system took me four tries to get right.
The first time, I wrote the four .md files and forgot about them. Claude never opened them on its own. The output stayed generic. I gave up after two weeks.
The second time, I wrote the role prompt but skipped the process files. Every chat was still a one-off. I was just typing fewer words to get the same result.
The third time, I had everything except the Sunday close. Three weeks later, the files were so out of date that Claude was giving me advice based on goals I'd already abandoned.
The fourth time worked. The fourth time is what you just read.
Don't expect this to feel magical on day one. Expect it to feel like work for the first week, useful in the second week, and invisible by the third. That's the curve.
Here's the thing nobody tells you about building a one-person company.
The hard part isn't the work.
The hard part is letting go of the idea that the work has to come from you.
I spent three months using AI to help me write faster. It helped. But I was still the one producing. The day I handed Claude my style files and said, "Your job is to protect my voice, not copy it" - everything changed.
Good AI use isn't about typing less. It's about staying in the decisions that only you can make. Everything else is infrastructure.
If you build this system right, your job stops being "do the work." Your job becomes "decide what good looks like, and reject anything that isn't it."
That's the company we're building here.
Recap checklist - in 60 seconds
→ Build a Claude Project named after your business
→ Load .md files: who-i-am, what-i-do, style-rules, operating-context
→ Write a 5-minute weekly brief every Monday
→ Assign Claude a role with the role prompt - not a task
→ Write one process file per output type, starting with the one you ship most
→ Use the review prompt before anything goes live
→ Run a 20-minute Sunday close every week
Save this. Screenshot it. Come back to it.
Your gift - as promised
The One-Person Company OS - in one Notion workspace.
→ Go and get your gift:
Take 90 minutes this Saturday. Set it up once. Run on it for the next 12 months.
We tested it in public. Now use it in private.
- Hamza ������