"I logged in one morning in my pajamas and saw that someone had paid me $100. It was shocking. I actually danced with joy."
Maor Shlomo, 31, of Israel. What he built was a tool that produces apps from natural-language description instead of code. First revenue: about $100. Six months later, the company sold for $80 million, plus performance-linked earnouts.
Six months is the number that grabs attention. But the more useful question is what he did and didn't do during those six months — the actual structure of one-person company automation.
In February 2025 he released Base44, a tool for building web apps. Ten thousand users in three weeks, 250,000 in six months. Ad budget: zero. On June 18, 2025, Israeli website builder Wix acquired the company for $80 million in cash.
In May, the month before the sale, net profit was $189,000 — after paying for all the high-end language model usage. Not the familiar pattern of burning cash to grow, but profitable from the start.
That profit came with a condition, though: every app a user builds costs AI usage fees, so cost scales with users. What that condition eventually produces comes later in this story.
Early in 2025 he was traveling in Thailand with his partner, a tattoo artist who needed a site to showcase her work. He tried to build it, and existing tools frustrated him — getting the shape he wanted always ended in touching code.
The thought that struck him: "A language model could build this."
Not a grand market analysis. The starting point of a company sold for $80 million was one girlfriend's website. Most of us work the other way: size the market, map the competition, write the plan, then build. That sequence filters out a signal like "I couldn't make my partner's website and it annoyed me" — it looks too small. But frustration you experienced yourself is more accurate than market research. At minimum, there's one confirmed user.
Asked how he found early users: "I begged my friends to try it. I literally sat next to them, watched them use it, saw what broke, and fixed it right there." That's how he got his first ten.
The origin of an AI automation success story is the most primitive kind of manual labor. Sitting beside a user can't be automated. Automation comes after. Invert the order and you build something nobody uses, very efficiently.
He's also stopped believing in the MVP orthodoxy. "I don't really believe in MVPs anymore. People's attention is too short. Build something proper for the small number who will actually use it." A user who is disappointed once doesn't come back.
Marketing was LinkedIn. He published his development process daily — and not just the wins. "I shared failures, lessons and numbers without hiding them."
The clever part came next: users who shared what they built got additional usage credit. When a user showed off their work, that was the advertising. It wasn't free — the credits cost him AI usage fees. Precisely, he spent his own cost of goods and the daily labor of publishing instead of media budget.
Strangers started promoting him unprompted around then. "People genuinely wanted me to succeed. Someone at MIT wrote to say they were spreading the word there." That was the return on publishing failures and numbers every day.
This is the most transferable part. He didn't stop at having AI write code.
"I spend 20-30% of my time not writing code but automating the business." And: "I automated everything that could be automated. I structured the codebase so a language model could write most of the code."
That sentence is worth reading twice. Not structured for human readability — structured for the AI to write into. He deliberately skipped a type-checking layer, reasoning that what protects a developer obstructs a language model.
Splitting models by task was practical too: different models for interface work and logic work rather than one model for everything. It's notable that the AI-agent travel company build reached the same conclusion independently.
He solved focus with tools as well, running distraction blockers and building his internal content management app with his own product.
His advice to founders compresses to one line: "Spend at least 50% of your time on what you're good at and enjoy."
His product principle was to "let people use as many features as possible without API keys." In plain terms, he hid the whole process of signing up for this service, connecting that one, and issuing keys.
His reasoning: "Even a very technical person wants to spend their time building, not signing up for services and wiring them together."
That translates directly to marketing organizations. What actually blocks staff adopting a new tool is usually setup, not features. They exhaust themselves creating accounts, requesting permissions and connecting integrations, and never use the thing.
His most practical advice for operators: "Is there a field you know better than anyone? Can you reach a specific group or customer base? If I've worked with doctors, I can now build a platform for doctors. That's an edge over the average person attempting it."
As tools get cheaper, ideas get cheaper with them. What remains is domain knowledge and customer access. For an operator who has spent years working with one industry, that experience is now the asset — because building costs almost nothing.
The headline was $80 million upfront, with performance-linked payments through 2029 and a $25 million retention pool for the team, which numbered eight at closing. It was one person at the start; the first hire came about six weeks before the sale.
Then the twist. Base44's annual recurring revenue reportedly passed $100 million nine months after acquisition and reached $150 million at eleven months. Based on Wix quarterly filings, reporting suggested Shlomo's total consideration climbed toward $150 million.
Over the same period, Wix stock fell roughly 50% in 2026. Market capitalization hit $2.2 billion, a four-year low. On the day Q1 results were published, the stock dropped 27% in a session.
The cause wasn't failure — it was success. The condition flagged earlier detonated here. Profitable at small scale, the structure inverted as it grew. By early 2026 Base44's gross margin was effectively 0%: the AI usage cost of each app a user built roughly equaled the fee charged.
Wix poured money into pushing the growth: around $90 million on customer acquisition in Q1 2026 and about $24 million producing and buying Super Bowl advertising. Quarterly revenue rose 14.3% to $541.2 million while the company swung to a loss. The company also acknowledged Base44 was cannibalizing its existing products.
Selling a company that hit $150 million ARR eleven months later for $80 million looks like a discount, and competitors in the same space carry valuations an order of magnitude higher.
The counterargument is heavier. First, the margin was 0% — growth consumed cash, and surviving that needed serious capital. Second, $80 million wasn't the end; earnouts moved total consideration toward $150 million, transferring risk while keeping upside. Third, the buyer bled. Fourth, paper valuations are paper: a funding round price is not a price anyone actually paid.
He didn't maximize price. He timed the exit.
He wasn't an unknown starting from nothing. He had already run a 100-person company for seven years, with a network and a name from elite military intelligence service. Public building on LinkedIn worked partly because he was already a known founder, and the acquirer was another Israeli company.
Strip those conditions out and "$80 million in six months" is a misreading. What does transfer at any scale is the operating structure: start from a problem you had, sit next to your first users, design your systems for AI to work in, split models by role, and remove setup from the product.
$80 million in cash to Wix in June 2025, with performance-linked payments through 2029 that reporting suggests moved total consideration toward $150 million.
Daily public building on LinkedIn — including failures and numbers — plus extra usage credits for users who shared what they built, turning user showcases into the advertising.
Beyond having AI write code, he structured the codebase itself so a language model could write most of it, spending 20-30% of his time on that business automation.
Base44's gross margin was effectively 0% — the AI usage cost per app built roughly matched the fee charged, so growing revenue consumed cash rather than generating it.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
Waiting at the booth rarely works at B2B trade shows. How to control what you reveal, book meetings in advance, stop passersby cheaply and follow up the same night.
Can employees take prompt libraries built at work when they leave? How U.S. work-for-hire and trade secret law apply to AI prompt ownership, and what firms should do.
Four teams named four different core customers, all based on gut feel. A step-by-step way to define your core customer from repeat behavior and map your beachhead.