The one-person company: a solo founder's AI playbook

TLDR

• One person can now cover every function. Build, design, marketing, support, and ops each have an AI tool that does the job a hire used to do.
• The proof is public. Pieter Levels runs Photo AI at ~$132K/month with zero employees and an 87% margin (source). Marc Lou made $1,032,000 in 2025 across small products, solo (source).
• The starter stack is cheap. Cursor at $20/mo (source), Lovable's free tier, and Intercom Fin at $0.99 per resolved ticket (source) replace roles that used to cost thousands a month.
• AI is the multiplier, distribution is the moat. Every one of these founders had an audience before the tools. The stack does not sell for you.
• Solo has real limits. Support quality, burnout, and unmaintainable code are the three ways it breaks. Plan for them.

The CEOs of the biggest AI labs keep making the same bet. Sam Altman told a group of founders there is a pool going on the first one-person billion-dollar company, something he called unimaginable before AI (TechCrunch). Anthropic's Dario Amodei put a date on it, predicting a billion-dollar business run by a single person could appear in 2026 (Inc.). Ignore the billion-dollar headline for a second. The useful part is smaller and already real: a single founder can now run every function of a business that used to need a team of ten.

What actually changed

For most of software history, a company was a way to assemble labor. You needed engineers to build, designers to make it look right, marketers to get attention, support reps to answer tickets, and an ops person to hold it together. Each function was a headcount, and headcount is slow to hire, expensive to keep, and hard to manage.

AI did not remove those functions. The work still has to happen. What changed is that each function now has a tool a single non-expert can operate. You are not learning to code, design, and write ads. You are learning to direct five tools that do those jobs, and to spot when the output is wrong. That is a much smaller skill set, and it is why one person can now hold the whole thing.

The stack, function by function

Here is the concrete version. For each core function, the tool a solo founder reaches for and the hire it stands in for.

A few notes so you use this well. On the build row, Cursor and Claude Code assume you can read a little code and describe what you want. If you cannot, start on the no-code row with Lovable, which has a genuinely usable free tier (30 credits a month) before its $25/mo Pro plan (pricing). On support, Intercom's Fin only charges when it actually resolves a ticket, so a hundred solved tickets costs you $99, not a salary. That per-outcome model is the clearest sign of where this is going: you rent the result, not the seat.

The trap on every row is the same. The tool gets you to 80% and stops. Your job is the last 20%: the taste to reject a bad design, the judgment to catch a support bot about to give a wrong refund, the read on whether the ad copy sounds like a human. That last 20% is the entire job now, and it does not come from the tool.

What one person actually earns

A handful of solo founders have already run the playbook in public, with real and sourced numbers.

Pieter Levels is the cleanest case. His Photo AI, an app that turns a few selfies into professional headshots, does roughly $132K to $138K per month, about a $1.6M annual run rate, with zero full-time employees (Indie Hackers deep dive). The economics are the interesting part. His server costs about $40 a month, his AI compute runs around $12K to $13K a month on the Replicate API, and the net margin sits above 87% (breakdown). He rents the hard parts (GPU training, payments, moderation) instead of building them, which is what lets one person run something that looks like a small Adobe. Across his whole portfolio (Photo AI, Nomad List, Remote OK, Interior AI) he clears roughly $3M a year, all solo.

The honest footnote: Levels once hired an AI developer temporarily for model setup. Solo does not mean he never paid anyone. It means no payroll, no HR, no management.

Marc Lou shows the other shape: not one hit, but many small products compounding. He publishes his revenue every month, and reported making $1,032,000 in 2025 (his newsletter). That total came from a spread of products (ShipFast and CodeFast at roughly $20K/month each, plus DataFast, TrustMRR, and others), not a single breakout. In July 2025 alone he broke it down publicly as $48,921 across CodeFast, ShipFast, DataFast, and smaller lines (LinkedIn). His 2025 was actually 20% down from 2024, a choice he made to work less and diversify. He uses occasional contractors for video edits and translations but runs the businesses as a one-person studio.

Danny Postma built HeadshotPro, another AI headshot product, and cleared over $100,000 in its first two weeks, later reaching about $300K/month at peak, alone from a laptop in Bali (Startup Founder Stories). His story carries the most useful warning in this whole piece. He eventually stopped running it solo and built a team, because past a certain scale one person could not hold the support, ops, and product load at once (interview). Solo is a starting configuration, not a law.

Notice the common thread. All three had a real audience before they had the product. Levels had spent a decade building in public with hundreds of thousands of followers. Marc Lou built his following posting revenue screenshots. The AI stack let them ship in weeks instead of months, but the distribution is what turned a launch into revenue. The tools are portable. The audience is not, and it is the part you cannot rent.

Where solo plus AI breaks

Three failure modes, all predictable.

Support quality. An AI support bot pointed at your whole product will confidently give wrong answers, and a frustrated customer with no human to reach will churn and tell others. The founders who get this right keep the bot's scope narrow (FAQ and known issues), and set a hard rule to hand off to a human after a couple of failed turns. Measure whether tickets reopen, not just whether they closed.

Maintainable code. If you ship an app you cannot read, every bug becomes an emergency and every new feature fights the AI's earlier decisions. Vibe-coded products often die right after the first viral spike for exactly this reason. Use AI for the boring scaffolding, but keep the core logic simple enough that you, or a developer you call, can actually reason about it.

Burnout. The most repeated confession from solo founders is not about tools, it is about load. One put it plainly in a widely-shared post: sitting at $3K MRR and feeling tired instead of proud, because the list of features, bugs, content, and customer emails never ends. AI removes the labor of individual tasks. It does not remove the fact that you are the single point of failure for all of them.

The tech stack doesn't matter. Getting customers, getting paid, and iterating fast is all that matters. (Pieter Levels)

What to do

If you are starting from zero, do this in order.

• Pick one narrow problem with obvious money attached. Photo AI works because a $200 studio headshot becomes a $39 upload. The math should be obvious to the buyer.
• Build the audience in parallel, not after. Spend real time on one channel (X, LinkedIn, a niche subreddit, a newsletter) before launch. This is the part the stack cannot do for you.
• Ship on the cheapest stack that works. Lovable's free tier or Cursor at $20/mo, deployed on Vercel. Do not spend month one choosing a framework.
• Add functions with tools only as they hurt. Wire up Intercom Fin when support eats your day, Zapier when the same task repeats. Do not pre-build a team of bots you do not need.
• Write down the handoff rules before you are drowning. When does the support bot escalate, when do you pause automation. Decide it calm, not during a bad week.

The realistic target is not $138K/month in 18 months. That took a rare audience and a first-mover window. A first-time solo founder running this pattern into a narrow niche can plausibly reach $10K to $30K a month within six to twelve months. That is a real income from a business one person controls end to end, which a few years ago simply was not on the table.

The billion-dollar solo company may or may not arrive on Amodei's timeline. It does not need to. The version that matters already exists: one person, a laptop, a stack of AI tools standing in for a whole org chart, and public proof it pays. Pick your problem, build your audience, and direct the tools.