What happened
Vibe coding — building software by talking to an AI model instead of typing the code yourself — has a hype problem. You've heard the claims: a million dollars in 17 days, a million from a single prompt, a million by October. So somebody finally did the unglamorous part and checked the receipts.
The bar was set to match the first real win for a startup: a project makes the list only if there's a public report of $100,000 in monthly revenue, or $1,000,000 in total earnings once you count the sale of the company. Five projects cleared it. Four projects that get passed around constantly did not.
Revenue and ARR are not the same thing
Worth clearing up before the numbers, because this is where almost all of the confusion lives. Revenue is money that actually came in. ARR — annual recurring revenue — is this month's revenue multiplied by twelve. It's a projection, not a deposit. Wherever a project reports ARR below, the monthly figure is shown separately and labeled as such.
"Made with vibe coding" is also being read narrowly here: the code was written in a back-and-forth with a model rather than by hand. That one distinction decides two of the cases below — and it sends them in opposite directions.
Medvi: $401M in year one, two people on payroll
Medvi is a telehealth company selling GLP-1 prescriptions, the class of weight-loss drugs you've seen everywhere. Founder Matthew Gallagher launched it in September 2024 with about $20,000. By press accounts, the first year produced $401 million in revenue, roughly 250,000 customers and $65 million in profit at a margin near 16%. The staff is two people: Gallagher and his brother Elliot. Everyone else is a contractor.
The code, the site and the agents came out of Claude, Grok and ChatGPT. Ad visuals came from Midjourney and Runway, and customer calls ran through ElevenLabs. But the part that actually makes the money isn't the code. Doctors, the pharmacy, prescriptions and the legal paperwork were all handled outside, through CareValidate, OpenLoop Health and LegalZoom. Gallagher did the assembling and the traffic.
Base44: side project to an $80M sale in six months
Base44 is an app builder. You describe what you want in plain text and it produces the app. Maor Shlomo, a developer in Israel, launched it in early 2025. In May it posted $189,000 in monthly profit. In June, Wix announced it was buying the company for $80 million in cash, with 250,000 users on the books.
Shlomo gets called a solo founder, and that's accurate about ownership — he was the only shareholder. But eight people worked there when it sold, and $25 million of the purchase price went to them as retention. "One person against the world" describes the start, not the whole ride. On tools, Shlomo said he deliberately structured the codebase so that a large share of it could be generated by models.
Cal AI, Peec AI and Wave AI round out the five
Cal AI counts calories and reportedly runs at about $2 million a month, backed by coverage from CNBC Make It and founder interviews. Peec AI sat at $4 million ARR when it raised its Series A, per Tech.eu. Wave AI, covered by Mixergy, is at $7 million ARR. That's the whole list above the line — five projects, four of them modest by startup-press standards and one that's genuinely enormous.
The four that didn't clear the bar
fly.pieter.com reached $1 million ARR with a peak around $87,000 a month, according to the author's own posts. Stanley was at $1 million ARR by October 2025, per the founders. Lumoo hit €700,000 ARR in a case study published by the Lovable platform, and Plinq reached $456,000 ARR in another Lovable case, plus an interview with Exame.
None of those are failures. They're one zero short of the threshold — or, in a couple of cases, they're missing nothing but the label.
What it means for you
At home
The takeaway isn't "quit your job." It's that the bottleneck moved. Medvi's founder didn't write code; he wired up doctors, suppliers and traffic. Think of it like cooking: the recipe app didn't make dinner, it just removed the guessing. The same shift is now available on your laptop for free.
At work
If you're the person who always has to beg engineering for a small internal tool, you can now draft one yourself and hand over something real instead of a wish list. That changes how fast your ideas get judged — and how often they get greenlit.
In business
Base44 sold for $80 million with eight employees. The pattern worth copying isn't the exit, it's the thin payroll plus specialists for the parts that aren't code — legal, fulfillment, compliance. Medvi farmed all of that out rather than hiring for it.
Studying
A flashcard app, a revision tracker, a quiz generator built from your own notes: all of these are afternoon projects now. The skill you're actually practicing is describing a problem precisely, which is useful far beyond school.
Creative work
Medvi's ad visuals came from Midjourney and Runway. The job isn't pushing pixels anymore; it's direction, taste and knowing what the ad is supposed to do. That's the part no model has taken over yet.
Side income
The four near-misses are the realistic target, not the $401 million headline. $456,000 ARR is a business. $100,000 a month is a milestone. Aim at one specific irritation for one specific group of people, and the numbers have somewhere to grow from.
If you want to see what's already out there before you pay for anything, mykreatool.com rounds up free AI tools in one place — a sensible first stop before you start stacking subscriptions.
How to try it right now
1. Start on a free tier. Open ChatGPT or Claude and use the free plan. No card, no trial clock running against you.
2. Pick one small irritation. Not a startup idea — something you personally do every week and dislike. Tracking expenses, formatting reports, sorting leads.
3. Describe it like you'd describe it to a friend. Who's it for, what goes in, what should come out. Skip the technical vocabulary; the model doesn't need it.
4. Test the dumb version first. Ask for the simplest possible thing that works. Ugly is fine at this stage. You're checking whether the idea holds, not whether it's pretty.
5. Only pay when a free tool hits its ceiling. At that point, an app builder like Lovable turns a written description into a working product — it's the same platform behind the Lumoo and Plinq case studies above, both of which passed €456,000 in ARR without ever crossing the $100,000-a-month line.
6. Keep humans on the high-stakes parts. Medvi outsourced prescriptions, pharmacy and legal. Whatever your version of that is — contracts, payments, safety — keep a professional in the loop.
Upsides and what changes
The distance between an idea and a working prototype has collapsed. Medvi started with roughly $20,000 and reached nine figures in year one. Base44 went from launch to an $80 million acquisition in about six months. A solo owner with eight employees can now compete with a funded team.
What changes for ordinary people is the cost of trying. You no longer need to hire a developer to find out whether your idea is any good — you need a free account and an hour. That moves the hard part from "can I build it?" to "is this actually worth building?" which was always the more important question.
Limitations
The receipts are thinner than the headlines suggest. Medvi's numbers come from press coverage of the company, not audited filings, and several others come from founder posts. ARR flatters everything: multiply a good month by twelve and you have a story, not a bank balance. Exit money is a one-time payment, not recurring income, and mixing the two into a single "threshold" hides that difference. It's also worth noticing who's on the list — Base44's founder was an experienced developer who could read and steer the generated code, which is not the same as a total beginner shipping an app. And this is a survivorship list: you're seeing the handful that cleared the bar, not the many that quietly stalled. Vibe coding lowers the cost of starting, but it doesn't lower the cost of being wrong.
Conclusion
Five projects cleared $100,000 a month or $1 million in total earnings after being built by talking to AI models. Four heavily hyped ones didn't, and the gap between the two groups is rarely about the code — it's about distribution, operations and picking a problem someone will pay to have solved.
One thing to do today: open a free ChatGPT or Claude chat and describe the most annoying recurring task in your week in three sentences. Ask for the simplest tool that would fix it. You'll know within twenty minutes whether this is worth more of your time.
Comments 0