A physicist can spend years learning how to ask a better question. Anton Osika learned that lesson at CERN, among people working on difficult problems in science. Later he found another question, much less imposing on its face: why did a person with a useful software idea so often need to find someone else to build it? The answer eventually became Lovable, a company whose interface begins with a sentence instead of a line of code.
There is a lovely irony in the route he took. Osika spent years acquiring the skills that his product aims to make less necessary for everyone else. He studied engineering physics at KTH in Stockholm and worked in particle physics at CERN. In a later interview, he recalled admiring the people there while deciding that he wanted to apply technical ideas more directly in the world. It was the sort of career turn that looks neat on a timeline. Living it probably felt more like leaving a very good party because you could hear another one starting down the street.
He went toward AI. Around 2015 and 2016, as deep learning was gathering momentum, he helped start the Stockholm AI community. He joined Sana as its first employee in 2016, then became co-founder and chief technology officer of Depict in 2020. Sana worked with AI for learning; Depict applied recommendation technology to online commerce. The settings changed, but Osika remained close to the difficult work of turning a promising technical system into something people could actually use.
The expensive distance between idea and app
Osika has described a recurring encounter: people with ideas would ask him to help find an engineer who could make them real. The request was reasonable. Software had become the usual answer to almost any business problem, and writing it was still a specialized skill. For the person with a sketch, the next step could mean recruitment, fundraising, months of specification or all three. A small idea often arrived with the administrative burden of a large one.
That gap was also a design problem. Development tools generally assumed the builder already knew how to code, even when the valuable knowledge belonged to the person closest to the problem. Osika's premise was that AI might change the starting point. If a model could translate a description into code, a founder, teacher or shop owner could test an idea while it was still vivid. The software might require repair and judgment, but the first version would no longer have to wait for a meeting with a development team.
“I think I think best by talking together with someone else.”Anton Osika, in conversation with Sequoia Capital
The line says something about his method as well as his temperament. The product he eventually helped build is conversational. It asks what someone wants, shows an answer and invites a correction. That rhythm is closer to working beside a patient collaborator than to filling out a technical brief. Osika did not arrive at it by pretending that code had vanished. He arrived after years spent writing and managing it.
A public experiment gets an audience
In 2023, Osika released GPT Engineer, an open-source project that used an AI model to generate software from instructions. It ran from the command line, which was familiar territory for developers and hardly a welcoming doorway for everyone else. Yet the project attracted a large audience on GitHub. The reaction offered more than applause. It showed that the basic request - tell a machine what you want to build and see it take shape - had arrived at exactly the right moment.
The public repository mattered because it let the idea be tested in the open. People could try it, argue with it, improve it and see its shortcomings. Open source can be a harsh focus group. It is also unusually honest. The revealing signal was that curiosity extended beyond a faster coding workflow. People wanted a route from intention to a usable product. Osika's next step was to make that route friendlier to people who did not live in a terminal window.
He founded Lovable with Fabian Hedin in Stockholm in 2023. Hedin brought his own engineering and founder experience, including work at Depict. The company first carried the GPT Engineer name, then adopted Lovable as its product identity in December 2024, after a November launch. The new name was a clue about the intended audience. A command-line experiment could prove that code generation worked. A product people wanted to return to had to make the entire experience feel possible.
Osika introduced the startup publicly on Valentine's Day in 2024, in a LinkedIn post that also wished his father a happy birthday. He described a plan in two steps: help anyone prototype a web app through natural language, then work toward more autonomous generation within a chosen technical stack. It was an unusually precise launch note for an ambitious claim. Even before the product had its final name, the founders were clear about the direction of travel.
The numbers, and what they measure
Lovable reached $10 million in annual recurring revenue two months after launch, according to the company's account of its early growth. It is a number that explains why investors paid attention. It does not, by itself, explain why users stayed. The founders' central wager was that the person who understood a problem should be able to shape the first working answer to it. A designer could make a tool. A small team could replace a spreadsheet. A founder could let customers touch an idea before turning it into a pitch deck.
The growth brought a familiar temptation: to tell the story as if it were a stopwatch. Lovable's milestones are striking. The company announced a $330 million Series B at a $6.6 billion valuation in December 2025, followed by a $400 million Series C at a $13.3 billion valuation in August 2026. In the latter announcement it said people had created more than 60 million projects since the November 2024 launch, and that apps built with Lovable were receiving more than 900 million visits a month. Those are the company's reported figures, and they describe activity on a platform, not a count of finished businesses.
The distinction matters. Making a prototype is now easier; making a lasting product is still an exercise in understanding users, maintaining software and handling real-world obligations. Lovable itself has moved beyond the initial thrill of generating an app. Its 2026 funding announcement emphasized features for finding customers, managing operations and running software securely. The next version of Osika's question is therefore harder: once people can build, can they keep what they built useful?
Osika is unusually well placed to know the difference. He has held the technical roles that deal with reliability and the founder roles that deal with customers. His years at Depict put him inside a company selling AI for a specific commercial purpose. Lovable asks a broader audience to cross the same bridge from interesting demonstration to dependable tool. That is where a conversational interface must meet the less charming duties of permissions, testing and maintenance.
The Stockholm part of the story
A company with global users could have a biography that begins almost anywhere. Lovable's begins in Stockholm, and Osika keeps returning to what that location means. In a Sequoia conversation with ElevenLabs co-founder Mati Staniszewski, he spoke about the talent available in Europe and his decision to keep building there. He has also spoken of wanting Stockholm to grow as a startup city. This is a founder making a claim about where ambitious technology can be made, not only about who can use it.
KTH gave Osika and Hedin its 2025 Innovation Award for their work to widen access to technology. The recognition completed a pleasing circle: the school where Osika studied engineering physics honored a company trying to make engineering output available to people without that training. In the same KTH conversation, Osika advised students to start small and work close to someone with a real problem. It is practical advice from a founder with a very large ambition. A billion-dollar valuation does not make a vague problem any clearer.
He and Hedin also signed the Founders Pledge, committing to give away half of their profits if Lovable succeeds. Osika describes himself on LinkedIn as someone who cares about having a positive impact. Such declarations are easy to print and harder to measure. The pledge gives one part of that ambition a specific form, while the product gives another: a wider group of people getting a chance to make tools for themselves.
There is plenty about Lovable's future that no one can write in advance. Competing products will change, AI models will improve and businesses built on the platform will reveal where its promises hold up. Osika's career so far suggests he is comfortable changing the form of the work while keeping the question intact. A physics lab gave way to early AI teams; those gave way to an open repository; the repository led to an application with a name that sounds less like software and more like a hope.
The revealing image is not a founder beside a valuation chart. It is a person with a problem describing the thing they need and seeing a first version appear. Osika spent much of his career being the engineer someone hoped to find. At Lovable, he is trying to make that first conversation available to everyone who has an idea worth testing. The question is deceptively small. Its consequences, if the tools keep improving, could be anything but.