Overestimated Individuals: China's "One-Person Company" Survival Guide

Byline: Ada, Deep Tide TechFlow

At the end of 2024, Israeli developer Maor Shlomo returned from military reserve duty, opened his laptop, and started writing a project. No funding, no team, no Slack channel. Six months later, Wix bought his company Base44 for $80 million in cash. At the time, the product already had 250k users and monthly profit of $189k. Three months before the acquisition, he hadn’t written a single line of front-end code.

Pieter Levels, a Dutchman, is even more extreme. One person, zero employees, running three products—Nomad List, Remote OK, and Photo AI—using the most basic PHP and jQuery. Total revenue in 2022 alone already reached $2.7 million. He had never been to work a day, had never raised funding, and lived as a digital nomad across more than 40 countries and over 150 cities.

These stories are too good—so good that they create a kind of illusion: with one person and one computer, you can build skyscrapers right on flat ground.

One-person-company incubators in Shenzhen, Shanghai, Suzhou, and Hangzhou have been springing up everywhere. In Nanshan, Moli Park covers 100k square meters, with 700 companies applying for it. In Lingang, the Zero-Inefficiency Magic Cube workstations are free; the first batch of 300 desks was snapped up, and the second batch of 8,000 square meters is already on the way. In Chengdu, Dan Shao runs a one-person-company community; after operating for less than a month, every event was packed.

The boom really did explode. According to the “China One-Person Company Development Trend Report (2025-2030)”, as of June 2025, the number of one-person limited liability companies across the country had already exceeded 16 million. In the first half of 2025 alone, the number of newly registered OPCs reached 2.86 million, a year-over-year surge of 47%, accounting for 23.8% of all newly registered enterprises.

But what’s under the boom?

We talked with three people who are already on this road. One is an observer who has run a one-person-company community for nearly two years and holds more than 2,500 real samples. One is a Gen Z founder who built two companies across the U.S. and China from Silicon Valley. And one is an independent developer who switched from an FA role in the primary market to building an AI Agent product.

The stories they told don’t quite match the vibe on social media.

The underlying logic of success

Dai Wenqian runs the one-person-company community SoloNest in Shanghai. The origin was very simple. In June 2024, she had just left the online education industry and wanted to know what a single person could really do. She searched books and videos but couldn’t find an answer, so she did her own field research instead. Set up meetings, review samples, conduct interviews.

Nearly two years later, she had accumulated more than 2,500 samples. Among them, 20% managed to run the business closed loop and got results. She also wrote her observations of the samples into a book: 《One-Person Company》.

A joint study by Qichacha and Xiao Report shows that by 2021, the three-year survival rate of enterprises established in China had already dropped to 71%, and nearly one quarter of them died outright in the first three years. Clearly, SoloNest’s 20% success rate is already ahead of the overall market.

But Dai Wenqian cares even more about the remaining 80%.

“People who don’t make it fall into two types. One type tries hands-on but doesn’t get it to work—the traffic cuts off, or the model isn’t sustainable. The other type never starts at all.” she analyzed.

People who never start are far more numerous than you’d imagine.

“Because they’re not hurting enough, there’s an escape route. The brain tricks them into thinking they want to start, but in essence they’re just afraid of missing out. Most of them worry about their current job, and they think OPC entrepreneurship might be a solution, but fear-driven behavior is a terrible starting point.” Dai Wenqian said.

Leon, who used to work as an FA in the primary market, has seen more people like this. He went to an offline event for one-person companies and found that many people had no direction: attending events everywhere, buying courses everywhere. “No one can help you figure out how you should make money. The correct path is to just do it, step on the potholes, and take the losses,” Leo said.

Among those who started, who actually survived?

The answer is very counterintuitive. In effective samples, Dai Wenqian saw a stable commonality: almost everyone who managed to run the loop was hardly doing anything in their original industry.

They don’t pick a track based on “what I’m good at.” Instead, they start from “where there are unmet needs.”

Dai Wenqian uses herself as an example. She used to work on branding at Himalaya, but she hadn’t done offline events or built a community. However, she had curiosity about people, an aesthetic sense for building products, and the ability to express things in structured form—these underlying capabilities aren’t limited by the industry. Once you find the market demand point, the scenarios, and the industry pain points, you can transfer those underlying capabilities over.

In the SoloNest community, there’s also a guy who makes tennis bags. He wasn’t making bags before either. Because he enjoys playing tennis, he discovered a demand point that wasn’t satisfied by existing products on the market, so he prepared 100k yuan in startup funds and built an original tennis bag. After two years, he can sell more than 300 units every month on a stable basis. This has an exact match in Pieter Levels’ methodology. Levels set himself a challenge in 2014: make 12 products in 12 months, then throw them into the market to see which gets a response. Nomad List was the 7th one—and the only one that actually worked.

The key isn’t only choosing the right track; you also need to validate enough assumptions quickly.

Dai Wenqian breaks the process into three watersheds. The first: do you dare to manually make something and throw it into the market for validation? Unfortunately, many people don’t even have the mindset for validation—they just think without doing. The second: after people become interested, can you sell it? There’s a chasm between “some people think it’s nice” and “people keep paying.” The third: can you free yourself from delivery work.

The first two weed out the people who don’t move, and the people who don’t move.

The third one is the real hard battle.

A trap of 1.2 million

Those who get through the first two gates will find that they’re alive, yes—but above their heads there’s a ceiling, firmly welded in place.

Dai Wenqian gave a precise number: if a one-person entity relies purely on personal delivery, the annual revenue ceiling is roughly 1 to 1.2 million RMB.

“No matter how diligent you are, selling time has a cap.” she said.

This is the most realistic dilemma for one-person companies in China. Social media talks about Maor Shlomo’s $80 million exit and Pieter Levels’ $2.7 million annual revenue, but that’s the story of Silicon Valley SaaS and globalized digital products. In China’s soil, one-person companies grow more on the C-end, in the tertiary industry, and in the experience economy—where delivery chains are heavy and people are tied together.

Barry, who operates across China and the U.S., saw a more essential divide. What American kids want to start is B2B SaaS and AI Agents. What Chinese kids want is tangible industries like pets, elder care, food, and so on. This isn’t about who is smarter—it’s about differences in industry structure and willingness to pay. U.S. companies have strong paying awareness; you can make it work with a small SaaS tool. But China’s B-to-B ecosystem is completely different.

So how do you actually break through the 1.2 million ceiling?

The most intuitive path is automation—using AI to pull yourself out of the delivery chain.

But this path is far harder than how the narrative makes it sound.

In the SoloNest community, there’s a typical case. Jason’s business is job-hunting “run-with-you” coaching—helping interns and new graduates for internet operations roles revise their resumes, do mock interviews, and assist with job searches. It starts with pure time sales: taking a dozen-plus clients in a month.

Many peers get stuck and die right here. Jason’s approach was to find a batch of peers who struggled to sustain customer acquisition, train them, and route clients to them. One deal, one deal of money—no employment relationship. Dai Wenqian calls it “multiple one-person companies cooperating, not one multi-person company.” Later, he also grew a To B operation-and-management business, moving from pure C-end to C plus B.

Now Jason is working on the third step: using the consulting know-how and interview transcripts from the past two years to build a knowledge base and develop a semi-automated delivery product. But in two months, he only completed 60%.

Why so slow? Dai Wenqian provided a math model: “Assume your delivery chain has 5 key nodes, and at each node, automation can only reach 80% of what you could do by hand. Then is the qualified rate after automating the entire chain 80%? Not at all. It could be 0.8×0.8×0.8×0.8×0.8, which is only 33%. The longer the chain, the harder it is to automate. It’s not additive—it’s multiplicative.”

That’s why it can feel like “you can automate quickly” with lobster stories, but once you build it for real, you know that if any step in the middle isn’t done well, what comes out is garbage. And the prerequisite for using AI well is that when you do it by hand, you’ve already done it very well—otherwise you won’t know where the problem is.

Leon has the strongest technical foundation among the three interviewees. He now builds AI Agent products himself; he doesn’t write a single line of code, and all development is handled by AI. AI’s penetration rate in his workflow is close to 100%.

But his judgment about automation is very restrained: “To evaluate whether a task can be handed to AI, look at three points: is the cost low, is the risk big, and is the outcome good. Services for high-net-worth people can’t use AI. The way AI works is that you allow it to make mistakes, and it optimizes strategy through those mistakes. But for services that are forced by high-net-worth customers, making mistakes is not allowed. If you make a wrong turn due to one communication, the whole business is over.”

Some business steps simply can’t be replaced by AI.

Dai Wenqian also admits that her AI penetration rate is only 30%. Because her core delivery is offline human-to-human interaction, and that part cannot be automated. What she can do is partial automation, including content-driven customer acquisition, knowledge base accumulation, and so on—but she can’t extract herself completely from the business.

She works more than 14 hours a day. Attracting new users with content, chatting with people to filter them, maintaining partners, designing products, breaking down samples—every weekend there are also two fixed offline events.

“Many one-person company founders won’t post this kind of thing online. Nobody watches. Everyone prefers to see glamorous pictures: having coffee here, visiting an exhibition there, making a million a year, and being a big female lead. But the reality is that entrepreneurship is lots of dirty and exhausting work—repeated again and again, iterating again and again.” she said.

A one-person company is not the end state

Automation is one path, but not the only path.

Dai Wenqian observed another way to break through: instead of replacing yourself, you splice yourself together.

Jason’s case follows this logic. He doesn’t hire people; he collaborates with other one-person companies. Each “Lego piece” is an independent entity: each has its own capabilities and customers, and when they are put together, they create incremental value.

And if every one-person company can be enhanced by AI, then putting them together is enhancing the Lego. Dai Wenqian thinks this is the biggest source of imagination for one-person companies: “It’s like Lego. Not every Lego piece has to be 100% AI-enabled, but every piece gets enhanced by AI. Three enhanced Lego pieces put together isn’t 1+1+1; it’s 3×3×3.”

Another path is to copy your experience and methodology to more people. Barry validated this model through practice. He is the founder of two companies that are structured as one-person entities. From 0 to 1, he explores everything himself. After the business closed loop runs, he withdraws and hands it over to the team for management—letting the team take over like a relay baton—then he goes and runs other businesses.

Maor Shlomo also had a similar choice. Base44 grew to 250k users within six months, and monthly profit was close to $200k—but he still chose to sell to Wix. His explanation was that although the growth is astonishing, the scale and size we need can’t be organically grown from one person alone. One person can build a product from 0 to 1, but from 1 to 100 requires organizational, resource, and distribution capabilities—something one person can’t do.

Three different paths—AI productization, cooperative splicing, partner expansion—but the underlying logic is the same: a one-person company is not a final state. It’s a launchpad. Verify something with the lowest cost. After you’ve run it to success, you must find a way so you’re no longer the bottleneck. Otherwise you’ll be welded forever onto that 1.2 million line.

Before the door closes

The 2026 data is looking bright. Shenzhen issued an OPC entrepreneurship ecosystem action plan, aiming to build more than 10 OPC communities on the scale of 10,000 square meters by 2027. In Shanghai’s Pudong, newly registered one-person companies get up to 300k yuan in free computing power. In Suzhou, 300k college students were attracted in 2025, and the talent pool is rapidly expanding.

But Dai Wenqian said something that keeps people grounded.

“Thresholds have been greatly lowered. Before, you had to find money, find people, and find venues—the startup cost was extremely high. Now you can basically validate anything with near-zero cost. But it’s indiscriminate access. You make it easier for yourself, and it also makes it easier for others. With more players, traffic becomes more expensive. This is a military arms race.”

Pieter Levels can make $2.7 million a year as a one-person operator because he started in 2014 and built a ten-year SEO moat and community trust. Maor Shlomo can sell Base44 within six months because he had already built a data company that raised $125 million before; his network, judgment, and speed weren’t from zero.

These people aren’t the “ordinary people can do it too” of the one-person company narrative. They are the brightest few points in survivor bias.

The real one-person company world—inside SoloNest’s community, with those 2,500+ samples—is this: 20% keep earning money and are moving into the next stage; 40% get stuck in various ways but are still iterating, still thinking about breaking through; and the other 40% are still迷茫 and looking for direction. Among the 20% that are alive, most don’t earn more than 1.2 million a year. They work until past midnight every day, with weekdays fully booked, and no weekends.

In fact, in this one-person company business, what you’re really earning is the time gap between “a niche demand has been discovered” and “yet to be occupied by organized capital.” That time gap has a name: the shelf life.

Shelf life depends on two things: the timing when you discover the demand, and the speed at which you run it.

Lowering thresholds won’t make shelf life longer. Quite the opposite—it makes shelf life shorter. Because if something can be validated with near-zero cost, others can validate it with near-zero cost too. The demand you see, others see as well. Today you manually build an MVP and throw it out for three months without dying; tomorrow, ten identical products will appear on the same users’ phones.

That’s also why most people get “stuck.” The essence of getting stuck isn’t an ability problem—it’s that the rate at which they cash out time on their work can’t keep up with the market’s crowded speed.

The reason Maor Shlomo and Pieter Levels are advertised instead of being treated as samples is precisely because they solved the shelf-life problem in two opposite ways. Levels extended shelf life to ten years through first-mover advantage and compounding returns; Shlomo compressed shelf life to six months through speed and exit.

For most Chinese one-person company founders, that middle path is the most dangerous: you neither have ten years to slowly build your flywheel, nor does Wix come writing checks. You work 14 hours a day just to maintain the 1.2 million ceiling, thinking that if you just endure a bit more you’ll break through. But the market won’t wait for you to hold on. Another peer who validates at zero cost will show up at any time and bulldoze your little moat flat.

A one-person company has never been a condition you can stay in long-term. It’s a window with a shelf life.

When the window opens, thresholds are low, tools are cheap, and demand is clear—so it looks like the best era for ordinary people. But the window won’t stay open forever. It will be filled by people coming in later, crushed by more efficient tools, and finally shut completely—either by some startup that secures funding, or by a suddenly downsized business line from a big company.

Whether you can move yourself out of that bottleneck position before the door closes is the only real question in this business.