The One-Person Billion-Dollar Company: What AI Actually Replaced (and What It Didn't)

In 2024, Sam Altman told a story about his group chat with other tech CEOs. They had a running bet: what year would the first one-person company hit a billion dollars in value. "Which would've been unimaginable without AI," he said. "And now it will happen."

Source: the conversation, as reported by Every

It was a casual line in a conversation, not a roadmap or a research paper. But two years later, it's worth checking against what solo founders are actually doing, not what the quote implies.

The clearest real example

Maor Shlomo built Base44 alone. He'd left his previous company, spent a year in Israel's army reserves, and came back with an idea: an AI tool that lets anyone build software by describing what they want.

He launched in February 2025. Three weeks later he'd crossed $1 million in annual revenue. One month in, Base44 had done nearly $1.5 million in subscription revenue. Four months after launch, Wix bought it for $80 million in cash.

Source: TechCrunch

Here's the part that matters for the "AI replaces a team" version of this story: Shlomo didn't hire a QA team, a support team, or a marketing team. He built AI agents to do that work instead. They ran quality-assurance tests, flagged UX problems, triaged product feedback, and generated marketing content straight from shipping data. If you want to see what a real stack behind that kind of setup looks like, I mapped 10 real Claude Code repos that do this sort of work, not theory, actual tools people run today.

That's real compression. Work that used to need three or four hires got handled by agents Shlomo set up and pointed at the problem.

What the same story says about the founder

Now the other half, from the same reporting.

Shlomo set alarms every two to three hours, overnight, for months, so he could personally check on his servers. Once, that habit caught an outage in 10 minutes instead of six hours. No agent did that. He did.

NYU professor J.P. Eggers, quoted in the same piece, put it plainly: AI "couldn't substitute for the judgment that comes from having specialists in the room."

Source: Fortune

And here's the detail I keep coming back to: Shlomo sold. Not because the product stopped working, but because scaling further needed something he didn't have: real consumer marketing expertise. No AI agent supplied that. He made a call about his own limits, and he sold the company because of it.

AI ran his QA. AI wrote his marketing copy. AI didn't decide whether to sell the company.

That was Shlomo, thinking like a founder.

A second data point, with an honest asterisk

Pieter Levels has been running solo internet businesses for over a decade. In February 2025, he built a browser flight simulator, fly.pieter.com, in three hours using Cursor, Claude, and Grok 3. Seventeen days later it was making $87,000 a month, on pace for $1 million a year.

Source: Pieter Levels' own writeup

AI coding tools clearly helped compress the build from months of traditional development into days. That much is a fair claim. What's less fair is treating the tools as the whole story.

Here's the asterisk. That 17-day number didn't come from nowhere. Levels had already spent 10 years building an audience: over 600,000 followers, 40-plus previous products, a habit of sharing everything he builds in public. When he shipped the flight sim, an audience was already there to try it. A year later, revenue settled to somewhere around $70,000 to $90,000 a month, not the launch-week peak.

AI compressed the build. It didn't build the audience, and it didn't hold the number at its highest point forever. Both of those needed a founder who'd been doing the work for a decade.

So what actually changed

Put the two stories together and a pattern shows up, and it's more useful than either "AI replaces founders" or "nothing's really changed."

Where does this leave someone starting solo

Not "you don't need anyone," and not "none of this is real." Both of those are wrong in the way that makes for bad advice.

The honest version: AI has genuinely lowered the cost of the execution work that used to require hires. That part of Altman's bet is already happening, not someday, right now, in public, with real revenue numbers attached to it. But nobody in either of these stories got to skip being a founder. Shlomo still had to know when to sell. Levels still had to spend a decade building the audience that made 17 days possible.

If you're weighing whether to start something on your own, that's actually good news: the execution bar is lower than it's ever been. The founder bar isn't. If that's the position you're actually in, the practical starting point isn't "learn everything about AI first," it's selling one automation to one type of business, the same lesson both stories above teach a different way.