How One Chinese Man Used AI to Build a One-Person Company and Made $28,000 in Two Weeks

In April 2026, a man in southern China registered a one-person company, built a quiz app for WeChat, spent about $14 promoting it, and pulled in roughly $28,000 (200,000 RMB) in advertising revenue before the month was over.

His name is Wang — Mr. Wang to the handful of people who've written about his story on Chinese social media. He's not a famous entrepreneur. He doesn't have a following. He wasn't backed by venture capital. He was, until recently, a regular office worker who decided to try the one-person company (OPC) model that's been gaining traction in China over the past year.

What makes his story worth examining isn't just the money — though $28,000 from a single mini-program in two weeks is impressive by any standard. It's the process. Every step he took was deliberate, documented in retrospect by people who studied his trajectory, and remarkably reproducible. He used AI tools not as a gimmick but as the actual operational backbone of his business — from choosing what to build, to building it, to figuring out how to get people to use it.

Let's walk through the whole thing, step by step.

Week One, Days 1-3: Picking the Right Tools

When Mr. Wang registered his one-person company in early April, he didn't have a product idea yet. What he did have was a clear constraint: he was doing this alone, with minimal capital, and AI would need to handle most of the heavy lifting.

His first three days were spent entirely on tool selection. Not building anything. Not brainstorming product ideas. Just testing.

He tried multiple AI coding assistants and API services. According to accounts of his process that circulated on Chinese tech forums, he evaluated tools on three criteria: could it write working code for WeChat Mini Programs, could it iterate quickly based on conversational feedback, and was the API cost low enough that he could afford to experiment without burning through his budget.

He settled on two tools:

OpenAI Codex (the AI coding agent) for front-end development and UI generation. Codex was strong at translating his descriptions into working WeChat Mini Program code — handling the layout, interactions, and WeChat-specific API calls.
DeepSeek V4 for backend logic and content generation. DeepSeek's API pricing ($0.14 per million input tokens, $0.28 per million output tokens) meant he could generate thousands of quiz questions, scoring algorithms, and result descriptions for pennies.

The choice was pragmatic, not ideological. He didn't pick these because they were the "best" models in some abstract benchmark sense. He picked them because together they covered his needs — Codex for the visual and interactive layer, DeepSeek for the content and logic layer — at a combined cost that wouldn't blow his budget during the experimental phase.

Three days. No code written for any actual product. Just tool evaluation. This turned out to be one of the smartest things he did.

Week One, Days 4-10: Choosing What to Build

With his tools selected, Mr. Wang spent the next seven days on what he later described (in paraphrased accounts) as the hardest part: deciding what to make.

He knew three things going in:

The platform would be WeChat Mini Program. In China, mini-programs are lightweight apps that run inside WeChat — no download required, no app store approval, instant access to 1.3 billion WeChat users. They're the default distribution channel for consumer-facing apps in China.
The business model would be pure advertising revenue. WeChat mini-programs can integrate Tencent's ad network (similar to Google AdMob). No in-app purchases, no subscriptions, no payment processing headaches. Users use it for free; ads pay the bills.
The product category would need to be inherently shareable. WeChat mini-programs have a built-in sharing mechanism — users can share results to their chat groups and Moments feed. If the product made people want to share, distribution would be free.

He used AI to help narrow down the product direction. According to reports, he fed his AI tools constraints like "must be completable in under 60 seconds," "must generate a shareable result that makes the user look good or feel entertained," and "must work with simple input — no account creation, no complex forms."

After a week of brainstorming, prototyping rough versions, and testing them on friends, he landed on his product: a fun personality assessment quiz. Not a serious psychological evaluation — a lighthearted, slightly irreverent quiz that tells you something amusing about yourself based on how you answer a series of quick questions. Think "What type of office plant are you?" or "Which historical figure matches your work style?" — the kind of thing people take on a lunch break and share with coworkers because the result is funny.

It sounds trivial. That's the point. Trivial products with viral mechanics are a proven category in the Chinese mini-program ecosystem. The hard part isn't the concept — it's the execution.

Week Two, Days 1-3: Building the Thing

This is where the AI tools really earned their keep. Mr. Wang built the entire mini-program in three days.

Here's how the development worked:

Frontend (Codex): He described the quiz flow in plain language — welcome screen, question cards with tap-to-select answers, animated transitions between questions, a loading screen while "calculating" results (deliberately padded to feel like the quiz was doing something complex), and a results screen with a shareable card. Codex generated the WeChat Mini Program code, which uses a framework similar to React but with WeChat-specific components. He iterated on the UI through conversation, asking Codex to adjust colors, spacing, and animations until it looked polished.

Content and backend (DeepSeek V4): He used DeepSeek to generate the quiz questions, answer options, scoring logic, and — most importantly — the result descriptions. The personality quiz had 12 possible outcomes, each with a funny, slightly flattering description designed to make people want to share it. DeepSeek also handled the backend logic: calculating scores, mapping answers to personality types, and generating dynamic result text.

The total API cost for development? Based on DeepSeek's pricing and the volume of content generation involved, likely under $3 for the entire development phase. Codex was accessed through a subscription plan.

By the end of day three, the mini-program was functional, tested, and ready to submit to WeChat for review.

Week Two, Days 4-10: Launch and Viral Explosion

WeChat's mini-program review process typically takes 1-3 business days. Mr. Wang submitted his quiz app and it was approved quickly — the app was simple, didn't access sensitive data, and met WeChat's content guidelines.

Then came the critical step: initial distribution.

He spent 100 RMB (about $14) on a small Douyin (the Chinese version of TikTok) ad campaign. Not a big promotional push — just a minimal test to see if the quiz had any traction. The ad showed a quick screen recording of someone taking the quiz and getting a funny result, with a prompt to "try it yourself."

The 100 RMB bought him a few hundred initial users. And that's when the sharing mechanics kicked in.

The quiz was designed so that the result screen was inherently shareable. It showed your personality type with a funny description and a colorful card that looked good in a WeChat chat. People took the quiz, got their result, laughed, and forwarded it to their group chats. Their friends took it. Their friends shared it. The cycle repeated.

Within seven days, the mini-program had accumulated 2.8 million page views. Not 2.8 million unique users — page views, which in the mini-program context includes repeat visits and shares. But even accounting for that, the traffic was enormous for a product that cost $14 to promote.

The advertising revenue followed immediately. WeChat's ad network pays on a CPM (cost per thousand impressions) basis. For entertainment content, CPM rates in China typically range from $5 to $15 depending on user demographics and ad placement. With 2.8 million page views and multiple ad placements per session (banner ads between questions, interstitial ads before results), the math works out:

2.8 million page views × estimated 3-4 ad impressions per session = roughly 8-11 million ad impressions
At an average CPM of $2-3 (conservative estimate for this content type) = approximately 200,000 RMB ($28,000)

That number — 200,000 RMB — is what circulated on Chinese tech forums and caught people's attention. A one-person company, built in two weeks with AI tools, generating a month's worth of a senior developer's salary in ad revenue from a single quiz app.

Why It Worked: The Anatomy of the Process

After Mr. Wang's story started getting attention, several Chinese tech bloggers and OPC enthusiasts dug into what he'd done. The consensus was that his success wasn't lucky — it was methodical. Here's what stood out:

1. He spent more time choosing than building. Ten days of tool selection and product direction research, followed by three days of actual development. Most people do the opposite — they start building immediately and then spend weeks fixing problems they could have avoided with a week of upfront thinking. Mr. Wang's ratio was roughly 75% planning, 25% execution.

2. He chose a platform with built-in distribution. WeChat mini-programs have a structural advantage: sharing is frictionless. You don't need to convince people to download an app, create an account, or add a bookmark. You just need them to tap a shared link. This meant his $14 in Douyin ads wasn't buying users — it was buying the initial spark for a chain reaction.

3. He chose a product that rewards sharing. Personality quizzes are one of the most reliably shareable content formats in any market. The result says something about you. You want your friends to see it. They want to take it too. The product's core mechanic — answering questions and getting a result — naturally leads to sharing. There's no additional "share to unlock" gimmick needed.

4. He used AI for what AI is actually good at. He didn't use AI to "run his company" in some vague, hand-wavy sense. He used it for specific, well-defined tasks: generating quiz content, writing result descriptions, building front-end code, designing the scoring algorithm. Each AI tool had a clear job. The tools were workers, not magic wands.

5. His cost structure was absurdly low. Total investment to launch: WeChat mini-program registration (about $45 annually for a business account), DeepSeek API usage during development (~$3), Douyin ad spend ($14), and his own time (roughly two weeks). Total cash outlay: under $65. Even if the product had flopped completely, his financial risk was essentially zero.

The OPC Model in Context

Mr. Wang's story landed in the middle of a broader trend in China. The one-person company (一人公司, or OPC) movement has been growing since 2025, driven by two factors: AI tools that make solo development practical, and a regulatory environment that makes it easy to register a business as a single individual.

In China, registering a one-person limited liability company takes about a day and costs very little. You get a business license, which lets you open a business bank account, sign contracts, and — crucially — register for WeChat's business-tier services including mini-program publishing and ad monetization.

The combination is powerful: cheap business registration + free distribution platforms (WeChat, Douyin) + AI coding tools + near-zero API costs = a viable startup ecosystem for individuals.

Mr. Wang wasn't the first person to do this. But the speed and efficiency of his execution — from registration to $28,000 in revenue in under a month — made him a reference point. Chinese social media posts about his story were shared widely, and several OPC-focused WeChat groups discussed his process in detail.

What Would This Look Like in the US?

Here's where things get interesting for an American reader. Mr. Wang's playbook relied on infrastructure that is specific to China — WeChat mini-programs, Douyin ads, Tencent's ad network, and a regulatory environment that makes solo business registration trivial. You can't copy-paste his exact approach to the US market.

But the underlying principles translate. And some of them translate very well.

The tool selection phase applies everywhere. Spending three days evaluating AI tools before writing a line of code is just as smart in the US as it is in China. The tools are different — you might choose Claude for content generation, GPT-4.1 for reasoning tasks, and a coding assistant like Cursor or Codex for development — but the principle of matching tools to tasks before building is universal.

The "shareable result" product pattern works in English too. BuzzFeed built a media empire on personality quizzes. Sporcle does millions in revenue from trivia. The format is proven in the English-speaking market. The difference is distribution — in the US, you don't have WeChat's seamless sharing. You'd need to think about TikTok, Instagram Reels, or even a simple web app with social sharing buttons.

The ad-supported model works at scale. Google AdSense, Mediavine, and other ad networks pay significantly higher CPMs for US traffic than Chinese ad networks pay for Chinese traffic. A quiz app that generates 2.8 million page views from US users would earn considerably more than $28,000 — potentially $50,000 to $100,000+ depending on the niche and ad placement.

The low-risk experimentation model is even more accessible in the US. You don't need to register a company to launch a web app. You can build something on Vercel or Netlify for free, use a cheap AI API like DeepSeek ($0.14/MTok input) for the intelligence layer, and test it with a $50 TikTok or Instagram ad budget. If it doesn't work, you've lost an afternoon and fifty bucks. If it does work, you've found a revenue stream.

The real lesson from Mr. Wang's story isn't "go build a WeChat quiz app." It's the process itself:

Pick your tools carefully. Don't start building until you know exactly which AI services you'll use and why.
Choose a product direction with built-in distribution. The product itself should make people want to share it. If you have to bolt on sharing as an afterthought, the product is wrong.
Build fast. Once you know what you're building and which tools you're using, execution should take days, not weeks. AI coding tools make this possible.
Test with minimal spend. Mr. Wang spent $14 on Douyin ads. You could spend $50 on TikTok ads. The point isn't to run a marketing campaign — it's to light the match and see if the sharing mechanics catch fire.
Let the product distribute itself. The $14 in ads didn't generate 2.8 million views. It generated a few hundred views. The product's shareability generated the rest.

The Math That Matters

Let's put Mr. Wang's economics in perspective for an American considering the same approach:

Time invested: ~2 weeks (10 days of planning + 3 days of building + launch week)
Cash invested: ~$65 (business registration, API costs, initial ads)
Revenue: ~$28,000 (200,000 RMB)
Return on investment: ~430x

No venture capital. No employees. No office. No inventory. Just one person, two AI tools, and a clear process.

Now, not every quiz app will hit 2.8 million views. Mr. Wang's product clearly struck a nerve — the right quiz, the right tone, the right sharing mechanics, at the right time. There was skill involved, and probably some luck.

But the structural advantages of the approach remain regardless of whether your first product hits or misses. The cost of experimentation is near zero. The tools are available to anyone with an internet connection. The distribution platforms exist.

The only real question is whether you're willing to spend two weeks trying.

Note: Details of Mr. Wang's story are based on accounts shared on Chinese tech forums and social media in April-May 2026. Currency conversions use an approximate rate of 7.2 CNY/USD. Individual results will vary. This article is for informational purposes and does not constitute financial or business advice.