The Three-Layer Dividend of the One-Person Company—an OPC Roundtable Talk

I sat on a roundtable today, topic: the dividends and opportunities of the one-person company (OPC) in the AI era. Here’s the gist of what I said.

I’ve run my own one-person company for eight years, so I have some standing here—and the host opened with two questions: where exactly is the OPC dividend? Does an OPC actually make money?

Laying my cards on the table

Twenty years in, the through-line is digital media marketing—Huawei’s ad agency, plus two stints in and out of Tencent, four years total. Plenty of side lines too—restaurants, guesthouses. I kept my own company the whole time I was at the big firms, disclosed it to my employer, no overlapping business. The one-person company is about eight years old.

Two people are on payroll today. But count everyone I can call on to get work done, and it still fits inside Dunbar’s 150—flexible contractors to sales at roughly 2:1.

The business boils down to enterprise services for AI-generated professional content:

• Medical education—built on years of managing health for corporate executives, now producing medical-education content backed by real papers, plus digital avatars of doctors
• E-commerce / social media photo returns—developed with top photography specialists
• Publishing / comic drama—all professional content production done with specialists in each industry, and the platform gets reused inside a small circle of those industries

The four-part OPC dividend

One: cost

The most obvious piece is no rent and low tax—small companies get real breaks. Development cost is the starker one. Back when I ran a product company, hiring a technical co-founder—no vague equity promises, real salary—cost at least 30-40k a month.

Now, with two 20× Claude Code plans, I alone can cover a technical co-founder’s workload.

Two: talent

Getting a specialist to work with me used to be close to impossible. They sat inside an organization with KPIs and OKRs, so there was no way they’d put serious time into exploration or joint R&D with me.

Now that a lot of specialists have been cut loose from those organizations, I can work with content producers and salespeople on flexible terms and go deep with them. A flexible network that big wasn’t possible before.

Three: competition

This one is specific to AI-era OPCs—everybody is a new company.

On the 29th I’m going to Shanghai for a pitch, and the company pitching against me has serious capital behind them, several backers with hundreds of millions. I get to compete at the same table—because we’re all new companies. You’ve been around two years, I’ve been around eight. There are more of these chances, and they’re fairer, than before.

Four: productization and scale

I did restaurants before—exhausting. Guesthouses—no way to scale. One of my old products went through planning, research, validation, layer after layer of development, and half a year later PMF still didn’t clear. A flat failure.

Now I just run all three directions at once, they’re all built the next day, I ship, and there’s a real chance of revenue right away. That was unimaginable before.

The three layers for choosing what to work on

The host pushed me on how I actually pick projects. As I see it now, there are roughly three kinds.

Layer one—genius tier (getting harder)

You have a brilliant idea, build the product, and get acquired by big capital. But the time it takes to clone a product keeps shrinking—first-mover advantage alone isn’t enough to make capital write the check anymore.

The ways a one-person company gets rich fast are all long shots—either it’s in the criminal code, or it’s a licensing monopoly, or it’s fraud. Forget this tier.

Layer two—certain money (the realistic advice)

Is the company supposed to feed you at this stage, and earn enough to cover daily life? That was my thinking when I started out.

The advice: make something standardized and productized, but reason about it backwards from certainty:

On day one of this product, either I collect a big share upfront, or it makes money from the day it launches.

Then find a niche must-have—one the giants won’t want even after you’ve worked at it for years. Two concrete cases:

• Embroidery pattern software—embroidery runs on machines now, like 3D printing, but you need software to make the pattern. Worldwide there are only a handful of companies at Photoshop’s level, all traditional software firms. AI and Claude Code can already do this—with some domain knowledge and engineering, you can build a full AI pattern-making stack. Users pay a few yuan, maybe a dozen, per pattern generated. Tiny niche, steady cash.
• Authenticating used and high-end guitars—the whole Chinese market trades only a few hundred million a year. What exactly would capital want to fight you over? If you have deep resources in that industry and can do the work at a professional level, that’s exactly the kind of market I want to find.

But the direction is the same: certain money, steady cash that covers your own living expenses.

Layer three—interest tier

What genuinely excites me—especially over that Spring Festival, and it wasn’t the crayfish, it was Claude Code: I went five days and five nights without sleep. At the time it felt like a dream coming true.

Concretely:

• Wrote an input method in Rust, only eight megabytes. I know nothing at all about input methods, and it still pulled it off
• Hardware—schematics done entirely by AI, then a factory that can turn it around fast
• Content production—no matter how the models iterate, professional methodology, thinking, and engineering experience ride along with them—the more the models iterate, the happier I am, because my output gets better

The key at the interest tier: something you’re genuinely into, plus a direction that compounds.

The single most important line

The most important thing is still that you have to go do it. I’m making myself post to WeChat Channels every day, about half an hour to get a long video out—you only get feedback after you do it, and only then can you iterate.

Source: EP0015_audio.mp3 · ASR model gemini-2.5-pro (chunked parallel) · full text of the 9:53 original recording · roundtable panelist remarks

[00:00] [Host] So today we’ve invited four people from very different fields—some of them are OPCs themselves, and some are founders serving OPCs. Let’s talk together about, if we want to become an OPC, where the opportunity is right now, and whether OPCs actually make money. Okay, so let’s start by having each of the four of you introduce yourself, and tell us how many people are on your team right now.

[00:23] [Zhang Pinpin] I was just running through it in my head, and my background is probably a bit on the complicated side.

[00:27] [Host] Keep it short—your résumé is too long.

[00:29] [Zhang Pinpin] Right. In about 20 years of working, the main thing, mostly, has been digital media marketing—digital media marketing at Huawei’s ad agency, and then on the media side, at places like Tencent, I went in and out twice, four years total.

[00:43] Outside of that there’s a whole bunch of other stuff—I’ve run restaurants, and I’ve run homestays. And even when I was at the big tech companies, I always had my own company on the side—you declare it to your employer, and as long as there’s no overlapping business it’s fine. So if we’re talking about a one-person company, I’ve been at it for about eight years now, so there’s quite a bit I can share.

[00:58] Then the business, right. On the business side—back at Tencent I also worked as a producer on some well-known TV dramas, and I incubated short-video accounts for celebrities. So the core business I do now looks pretty scattered, but if I had to put one label on it, it’d be enterprise services for AI-generated professional content.

[01:19] Let me give a concrete example. I spent two years in healthcare—doing health management specifically for corporate executives and well-known big names. Because of that stretch, one of my mature businesses now is generating medical explainer content and producing the videos—think medical education content backed by real expertise and real papers, delivered by a doctor’s digital avatar.

[01:47] And another one, for example, is content we developed together with top photography experts—consumer-facing photo turnarounds for e-commerce and for social media. The point is that expert knowledge is baked into that content, and what we produce is stuff the generic tools on the market simply can’t do.

[02:06] Same with publishing, and with the comic-drama platform I built a while back—all of it is producing professional content in partnership with experts inside the industry. And once that content is produced, the platform can be reused within a small circle in that industry. That’s roughly the model.

[02:20] As for the company right now, two people are on the payroll. I’ve actually never answered this question before, but I did the math—the number of people the company can tap to actually get stuff done roughly fits Dunbar’s number, so under 150. And a quick count: the people who do the work, the flexible contractors, versus the sales and channel side, it’s about a 2:1 ratio—so maybe close to 100 people who can execute, people I can hand tasks to, and another 50 or so who sell. That’s the shape of it.

[02:56] [Host] So I want to, I want to ask all of you—everyone’s talking about the OPC dividend. From an OPC’s point of view, what does that dividend actually refer to? Whether you’re doing product development, or operations, or growth—what’s the difference from the old model? What’s the single biggest thing it’s made easier for you?

[03:18] [Zhang Pinpin] I thought about it carefully just now, and I see four layers.

[03:20] The first layer is cost. It’s really obvious—I’ve been running my own company all along, and I used to pay rent. Starting this year, rent’s gone—right, and that’s a very direct break for a small company. Second, taxes went down too. I think those are the most immediate ones.

[03:37] Then, still on cost—development cost. Back when I ran a product company, finding a technical partner, if you’re not selling them a dream—no equity talk, just straight up, they work and I pay salary—that’s at least 30,000 or 40,000 a month. Now, with two 20x Claude Code plans of my own, I can cover a technical co-founder’s workload, right? That’s the first and most concrete layer.

[04:01] The second layer, I think, is talent. In the past it was hard to get an expert to work with me, because they were inside an organization with the organization’s KPIs or OKRs, and there was no way they’d put much time into exploring or co-developing something with me.

[04:16] But now, with a lot of experts freed from those organizations, I can work very flexibly and very deeply with them, whether they’re professional content producers or professional salespeople. That wasn’t really imaginable before—like I said, having that big a network of flexible collaborators just wasn’t possible.

[04:39] The third one, I’d say, is competition. I think this is a dividend specific to AI OPCs—everybody’s a new company. Example: on the 29th I’m going to Shanghai for a competitive pitch. The firm pitching against me has raised serious money—several hundred million from several funds. And I now have a shot at competing with them at the same table, because we’re all new companies, right? You’ve been around two years, I’ve been around eight—that kind of shot is more common and fairer than it used to be, because the client’s willing to believe a real pro can put a genuinely good product or solution on the table. Yeah, that’s the third point.

[05:15] And the fourth point, which I think is the biggest dividend—the two before me already touched on it—is the productization and scale dividend. I used to do restaurants, which was exhausting, and homestays, which just can’t scale. But now it can go fast. Li Biao and I were talking about this backstage a minute ago: with my old products, from planning to research to validation to building it out layer by layer, half a year would go by, and then PMF just—didn’t—clear, so it’s a failed product. Now I just build all three directions at once, they’re done the next day, I ship them, and there might be revenue right away. That was unimaginable before—the possibility of productizing and scaling. Those are the four layers.

[05:54] [Host] A lot of the projects you do are really cross-disciplinary, even if they all touch the content ecosystem somehow. So now, when you take on a new project, or when you greenlight a product of your own, how do you decide it’s worth doing? Can you walk everybody through that?

[06:09] [Zhang Pinpin] Since I’m sharing this publicly, I want to be responsible about it. Coming at this purely from my own angle, I probably can’t cover every situation. As I see it right now, there are roughly three kinds.

[06:22] The first kind is getting harder and harder—the genius tier. You have a brilliant idea, you build a product, and some big player buys you out. Honestly, cloning a product keeps getting faster, so being first isn’t enough to get investors to write the check anymore. So for a one-person company hoping to get rich fast, the possibilities left all look thin—either it’s the kind of thing written up in the criminal code, right, or it’s a licensing monopoly, or it’s the folks who were talking a while back about how great they were doing, which is just fraud. So I don’t think this category is even in scope for most people.

[06:56] Then the second kind is more realistic: at whatever stage you’re at, do you need this company to feed you, to earn the money that covers daily life? That’s how I was thinking about it when I first started out too.

[07:10] My advice for that direction is—make it standardized and productized, but think about it in terms of certainty. On day one, this product either collects a large percentage upfront, or it makes money the day it goes live. That’s what I mean by certainty.

[07:27] Second is a niche must-have in the field I’m in—something where even if I work at it for a long time, the giants still won’t bother with it. Let me give a few small examples I’ve come across recently.

[07:38] For instance—do you all know about embroidery? Embroidery is now fully machine-driven, like 3D printing, but it needs software to digitize the pattern. Internationally that software comes from only a handful of companies, Photoshop-type companies, and it’s traditional software. AI, Claude Code, can already do it—if you know some of the relevant domain knowledge and engineering, you can build a whole AI pattern-digitizing setup. And for every pattern you generate, overseas users are paying a few bucks, even ten-plus bucks. That’s an extremely niche market.

[08:13] Another one I heard about a while ago—authenticating secondhand guitars, good guitars. That whole market across China does only a few hundred million a year in transactions. So what’s big money going to fight you for? If you have a lot of resources in that industry and you can do the work really professionally, that’s the kind of market I’d love to find and go after.

[08:30] But the big direction here is money with certainty—I want this thing to keep generating cash, enough to cover my personal living expenses.

[08:38] Then below that, the third layer is interest. What genuinely excites me—especially over Chinese New Year, and it wasn’t because of OpenClaw, it was because of Claude Code, I didn’t sleep for five days and five nights. What it felt like at the time was a dream coming true.

[08:51] Since hardware came up earlier—during that same stretch I also built an input method myself, an input method written in Rust, only eight megabytes. I don’t understand the first thing about input methods, and it still pulled it off.

[09:01] And now I can do hardware too—AI does all my drawings, and there are factories that’ll turn them around quick, so I can move fast. That’s where we are right now: a dream can actually get built.

[09:12] So given that, wherever my interest is, I put myself into it—after the first two stages, of course, once that first stage is behind you. Really strong interest, and, and where the skills stack up over time.

[09:25] Where the skills stack up—for example, I’ve been doing content production all along, and in there, no matter how the models iterate, the domain methodology, the thinking, and the engineering experience carry across model iterations—the more the models iterate, the happier I am, my output gets better. So that’s a direction that’s both interesting and compounds.

[09:42] Right. And I think the most important thing is still, still that you have to, have to do it. Right? I’m forcing myself to post on WeChat Channels every day now, spending about half an hour putting out a long video—only after you do it do you get feedback, and only then can you iterate.