Wu Ruimeng: Two Months of Vibe Coding, $20K in Revenue

Intro

Wu Ruimeng's day job is CTO of a startup. In his spare time, he raised no funding, hired no one, and wrote no grand narrative. He just leaned on a way of working called "vibe coding" — say your idea out loud and let AI write the code — and pulled in over ¥150,000 (about $20K) in two months from client work and small apps.

The product that surprised even him was the one that sounded most worthless: a baby-naming app. Each order cost 9.9 yuan, roughly $1.40. "This app sounds like it has no value," he told a reporter. "But it closed a few hundred orders, just like that."

Story

On March 3, 2026, Southern Metropolis Daily's Southern Plus app published a feature titled "The Rise of One-Person Companies: How Can Individuals Ride AI into Market Innovation?" Wu Ruimeng was its protagonist — a self-described "super individual": CTO in the large-model industry by day, AI-tool freelancer by night and weekend. His core platform was Baidu's "Miaoda" (秒哒), a no-code app-generation tool first unveiled at Baidu World in November 2024, built to take someone "from idea to app prototype" fast.

What he shipped in those two months sounds deeply unsexy — all long-tail micro-needs.

First: an AI comic-drama app for a client, generating illustrated short dramas with AI. Second: a corporate website for an enterprise. Third: an "itasha" simulation and customization tool for a Taobao shop owner — itasha being the anime-wrapped cars of otaku culture; buyers want to preview the wrap before committing, and this shopkeeper needed a try-it-yourself tool for customers.

Fourth was the 9.9-yuan baby-naming app. The trigger was personal twice over: naming his own newborn had nearly broken his brain, and then he saw someone online looking for a custom naming app. He thought: "There has to be an opportunity here." He built it, listed it, and it closed roughly a few hundred orders at 9.9 yuan each.

Put the naming app's numbers in perspective: a few hundred orders at $1.40 each is only a few hundred dollars — pocket change next to the ¥150,000 total. The real money came from client work: the AI comic-drama app, the enterprise site, the itasha tool. The naming app mattered for a different reason: it proved that a product "with no value" in the eyes of any investor could still find hundreds of paying users, because it scratched a real, specific itch. Nobody pitches 9.9 yuan on a demo day. Nobody copies it either.

All four together: more than ¥150,000 in two months. In his own words, it took "not much extra time or energy, and no additional funding or staff."

What makes this interesting is where it sits — exactly at the intersection of two trends. On one side, the one-person company (OPC) is becoming a recognized form, even by policy: Apple acquired invrs.io, an AI startup with a single employee, and cities like Hangzhou, Suzhou and Shenzhen have rolled out OPC cultivation programs. On the other, the tooling has been completely democratized: Wu notes there are now no-coding products where "even people who know zero code can program." And the economics of the platform layer make the math work. Miaoda's product lead, Zhu Guangxiang, described a three-tier model: pay per large-model call ("second-points," a virtual currency) during the app-building stage, pay-as-you-go cloud hosting once it runs, and future paid plug-ins and template marketplaces as value-added services. In other words, the cost structure of shipping software has collapsed from salaries and servers to pocket change per API call — the exact inversion that makes a 9.9-yuan product viable. Wu also flagged where the demand hides: the long tail of small organizational needs, like school course-selection systems or gym booking and membership tools — boring, specific, and numerous enough that no agency bothers to serve them.

He distilled the approach into two principles, both counterintuitive.

The first is about picking your arena: only work inside your professional domain. His day job is the large-model industry. AI iterates so fast and time is so precious, he argues, that developers lose money doing unfamiliar things. It sounds like conservatism; it's actually a speed strategy. In a domain you know, AI amplifies your judgment. In a domain you don't, it just helps you make mistakes faster.

The second is about breaking out: "Speed itself is the moat." When everyone can call the same models, copycat competition is inevitable. His answer has two parts — execute so fast nobody catches up, and accumulate personal brand and personal resources: network, cash flow. "A good product plus good influence makes success much more likely."

He offered an example: "I noticed a 'digital nomad' who makes over a million a year alone... his profit already exceeds most ordinary companies." And he pointed at companies like Midjourney and MiniMax — just dozens or a couple hundred people, with outsized influence. A small team's leverage, he says, was never headcount.

Lessons

Three mechanisms worth copying from Wu Ruimeng's story: 1. Only hunt niches "too small to pitch": a 9.9-yuan naming app, an itasha preview tool, a school course-selection system — needs too small for any VC to glance at, but real, and people pay. Baidu Miaoda's product lead confirmed the pattern: long-tail scenarios hide huge real development demand. Make it executable: keep a list of problems "people around you would pay $1.50 to solve," ship one MVP per week, and only scale after 50 paid orders. Skip the branding and the moat talk — a tiny niche's moat is that "nobody bothers to copy it." 2. Compress "idea to prototype" into a single weekend: Wu's weapon is vibe coding — speak the idea, let AI write the code, no-code tools push build cost toward zero. He shipped four products in two months not through effort but by driving each idea's validation cost down to days. Make it executable: one hard rule — any new idea must have a prototype that can take money (or take users) within 7 days; miss the deadline and it goes to the archive. Use Miaoda-style no-code tools or any AI coding assistant — the deadline is the point. 3. Professional domain × personal brand = the one-person engine: he only does large-model work, his professional home turf (the expertise moat), while building personal brand and resource reserves (the distribution moat). Either one alone fails: a great product with no influence means nobody finds the 9.9-yuan app; influence with no expertise means the next AI wave wipes you out. Make it executable: pick a domain where you already have 3+ years of experience and go deep; publish a public post-mortem for every product you ship (even 500 words). Product and influence multiply.

Closing

Wu Ruimeng didn't quit his job, didn't go all-in, didn't write a manifesto. He just used AI as leverage, treated "long-tail micro-needs" as prey, and spent two months proving a formula: judgment in your professional domain, times the speed of vibe coding, times the amplifier of personal brand — equals a company with no employees. The ¥150,000 isn't the point; it's the receipt. It proves that in the AI era, one person doing small things seriously can collect a big company's profits.