一个人、六个月、八千万美元：Base44 如何卖给 Wix

Base44 is an AI app-building platform where users describe an app, game or website in natural language and have AI code it. Shlomo says the category is crowded, but Base44 takes an opinionated, "batteries included" approach: every app comes with a built-in database, integrations, user management and analytics, with no third-party services or API keys. Competitors often rely on integrations such as Supabase, which he says can be slightly less strong than building everything full stack.

The idea started when his girlfriend, an artist with a small business, needed a website to capture leads, and the drag-and-drop tools he tried were painful. He had worked with LLMs in his previous company and knew models could write that code if given the right infrastructure. The second trigger was volunteering with a huge Scouts organization in Israel that had no internal developers and was being quoted around a million dollars by agencies for software he felt could be built much more cheaply.

He had previously founded Explorium, an enterprise data company that raised $130 million and where he was CEO for seven years. After the October 7 war broke out he served in the reserves for almost a year, then traveled and realized he missed coding. Starting Base44 with little to lose, he says, gave him energy, and he never expected it to take off so fast.

Shlomo says going solo and bootstrapped does not suit every company: a B2B or enterprise business needs salespeople and marketing, and buyers hesitate when one person owns the product. But if a product can be viral and target the masses, and you can reach product-market fit, he thinks solo bootstrapping beats the alternative, even purely on financial outcome. He also finds it less stressful to wake up profitable, using the idea of being "default alive", which helps him keep energy up for what he hopes is a marathon.

The downside is that nobody shares the burden. With no DevOps team or on-call rotation, he once had to open a laptop during his brother's wedding photo shoot after a friend reported a crypto-scam hack of Base44. It turned out the LLM had used a Node package called `cryptography`, and a non-technical user saw the error and assumed the worst. Shlomo says he also had to prioritize brutally and switch constantly between what he wanted to work on, usually coding, and what the business needed, often marketing.

Shlomo says he has severe ADHD and treats it as a potential superpower, but he structures his day for deep work. He uses RescueTime to block Twitter and LinkedIn, which was hard while building in public. He spent a lot of time agonizing over how to structure his code repository so he could use Cursor for the backend and Base44 for the front end to write code quickly.

He also vibe coded his own internal tools on Base44, including a content app. His process used to involve fighting with ChatGPT over tone, then making an image elsewhere, then reworking a LinkedIn post into a Twitter thread. The app took his weekly ideas, wrote in his own voice using previous posts he liked, and generated the variations, and he calls this kind of malleable software adaptive: after the Wix deal, two prompts changed the process.

On whether he could compete with funded rivals such as Lovable, Bolt, Replit and Vercel, Shlomo says Base44 was far more profitable than he expected and came close to $200K in profit in May. He says he has not written a line of HTML or JavaScript in three months, because even a solo founder is effectively managing teams of AIs. He expects 10x engineers to become 100x engineers, so that team size and funding matter less in winning a category.

Shlomo says he could have stayed independent and made good money, because users who came to Base44 stuck even though they knew the competition. The hard part was making as much noise as rivals running million-dollar hackathons. He chose acquisition because the market was moving faster than anything he had seen, and he felt Wix, with the same DNA and customer base, gave the best chance to lead the category and "punch down, not only punch up".

He thinks that domain knowledge and an interesting angle now matter more than funding, giving the example of his first user, a friend and former restaurant manager who built an invoice SaaS for restaurants. Lenny relates this to his own "anti-empire" newsletter. Shlomo says Base44 was the first time he was not trying to build the biggest thing: on a plane he told his girlfriend they would buy a nice car at $1.5 million ARR by the end of 2025, and they hit it within about four weeks.

Shlomo got his first users by asking close friends, starting with three, two of whom were unemployed at the time. He sat with them around a table every other day: they tried to build something, it broke, he checked the logs, fixed it and shipped it. He did not market anything until users started sharing the product unprompted, and the 11th user, someone who did not know him, was the signal to invest in launching.

His first Product Hunt launch went badly and brought about 15 new users, but he treated it as a tool to reach the next 30 or 50 rather than as make-or-break. He also got his first paying customer, who churned within hours because the product was not good yet. The second launch, supported by a very active community, was so strong that Product Hunt's algorithm treated the votes as bots until the community contacted support and it was fixed; Base44 then moved to first place with roughly a 500-vote lead.

Shlomo says he does not believe in releasing an MVP that is not good enough, since people's attention spans for new software are short. Even past 20K, 50K and 100K users he gathered 20 to 30 people in a room every other week for focus-group-style feedback. He also thinks a rigid ICP matters less than what people are trying to do, which in his case is build a tool.

After influencer posts and paid ads failed, a founder friend suggested sharing his bootstrapped, non-VC journey. Shlomo began posting on LinkedIn, which fit because his audience was builders. He wrote honestly about the good, the bad and the ugly, from the technical stack to growth numbers, and tested posts on friends first as if they were a product.

The second lever was a credits program: users who shared posts about building an app, even not about Base44, got extra credits. Early on they emailed him links and he granted credits, later automated. A week after growth was about 20 users a day, Base44 saw about 4,000 new users a day and broke, and scaling taught him things like databases and virtual CPU costs on the fly. Shlomo adds that velocity itself became a growth engine, since shipping features every other day pulled people in.

He had a few thousand LinkedIn connections to start, and says the key was picking the channel that worked and putting all his weight on it. Posting on Twitter was a waste of time for him, although he now has a decent following there. Lenny compares this to growth engines like SEO, paid or sales, where most companies grow mainly through one.

Shlomo ran the community first on WhatsApp, which he found surprisingly good for feedback and spotting slowness or bugs, before moving to Discord, Reddit and email updates. His first hire came about a month and a half before the acquisition: a technical product person who could read LLM logs, write Python analysis scripts and add analytics. He later asked him to take over growth, showing why he wanted people who can wear many hats.

At roughly 5,000 to 10,000 users, Shlomo launched a hackathon for building apps that do good in the world, funded by a 5K prize from Base44's profits. It drew 3,000 teams and sponsors including Amazon, Google, MongoDB and Deloitte. He says it is one of the top moments of his career, citing someone who built a memory game with photos and names for her grandmother with Alzheimer's.

Shlomo runs infrastructure on Render.com, which he praises as easy compared with the large DevOps teams at his old company, and uses MongoDB because schemas change often when vibe coding. He spent 20 to 30 percent of his time making the repository suitable for LLMs. His principle was to have the LLM write as little code as possible by providing a very opinionated infrastructure that handles CRUD, authentication and the database, so there are fewer places to make mistakes or lose context.

His controversial advice is to use plain JavaScript and JSX instead of TypeScript because models write it more easily. He also keeps front end and backend in one repository so the AI sees both, and writes the backend in Python, which has held up under heavy traffic and attempted DDoS attacks. Base44 uses a mix of models: Claude for the initial app and UI, Gemini for complex algorithms or when Claude gets stuck, with a router choosing between them.

He uses a heavy-and-light pattern in which strong models propose high-level changes and smaller, faster models such as Flash or o4-mini patch the code in the file. Asked about activation, he explains he dropped a step where the LLM first showed user flows, an almost PRD-like plan. It produced better apps but fewer users reaching the aha moment, "Holy shit, it actually understood me", so he removed it.

Wix reached out after community members kept saying that Wix should buy Base44, partly because of the Israeli ecosystem and partly because people built websites on Base44. At the first meeting, Wix's CEO Avishai said everyone had been telling them to buy Base44, so it was worth a talk. Shlomo spent several evenings with him over dinners asking for advice, and it was not clearly an acquisition path at first.

He says chemistry with the buyer's team is crucial, especially for small companies, and being fun to work with matters. He thinks the best position from which to negotiate is to be fine with not being acquired, like not showing too much interest on early dates. The $80 million was the initial payment, with a further piece tied to future performance, which Shlomo likes because it aligns his financial and personal interest with making Base44 as big as possible.

The team had pushed to sign on a Thursday night, but at 2 a.m. the lawyers still wanted wording changes, so they went to sleep. At 4 a.m. an announcement came that war had broken out between Iran and Israel, and they signed the papers the next morning. Shlomo calls it the least boring month of his life.

Shlomo thanks the Base44 community and offers his main advice: make sure at least 50% of your time goes to work you are great at and enjoy, because that keeps you showing up every day. He notes that even CEOs of public companies he has met spend a major part of their job on what they love, such as product design or building a marketing machine. He says this is the best time to build and that building something you like leaves you little to regret.

On the name, Shlomo says he wanted "base" in it as the foundation for building software. Base.com was unavailable and he had little money for a premium domain, so he tried numbers from his birthday, February 2: Base22 was available, and doubling it gave Base44. It also evoked Base64, an encoding of data, and he liked the idea of encoding natural language into software.