Iain Martin opened one of the most revealing conversations at HumanX Amsterdam with a deliberately simple challenge to Lovable co-founder Fabian Hedin:
“You think vibe coding is dead. What comes next?”
Hedin did not reject what the phrase once represented. “Vibe coding” captured the early thrill of handing a rough idea to AI and watching software appear. But that description, he argued, belongs to a period when people were mostly experimenting. Today, users are building products they charge money for, internal systems companies depend on and entire businesses around AI-generated software.
His preferred description now is sharper: “I prefer the term agentic coding.”
That small change in vocabulary became the thread running through the entire 25-minute HumanX conversation. The barrier to writing software may be collapsing, but the responsibilities around software are not disappearing with it. They are moving somewhere else.
HumanX itself framed the session around that transition. Since Lovable’s late-2024 launch, the platform is now generating more than one million new projects each week, according to the event description, with natural-language development increasingly being used by founders as well as employees inside large companies.
Martin began by putting Lovable’s own speed into perspective. He referenced a year in which the company had moved from roughly $200 million in revenue to a latest financing that valued it at $13 billion, then asked Hedin for the current revenue figure.
Hedin’s answer was brief: “600 million.” The figure was stated by Hedin during the session and should therefore be understood as company-reported.
Yet the more interesting number came a few minutes later. Hedin said roughly 80% of the ideas users bring to Lovable are connected to building a business. What began with websites and landing pages has moved toward SaaS products, customer-facing applications and internal operating software. He described one Brazilian AI education company where the product itself was built on Lovable and roughly 50 employees were also using it to build internal systems as the business pursued what Hedin described as around $20 million in annual recurring revenue.
That example points to a deeper change. Software creation is moving toward the person who understands the problem rather than waiting for the person who knows how to code it.
Hedin told Martin that Lovable usage now cuts across finance, HR, product, design and engineering. His reasoning is straightforward: a finance employee who understands a broken finance process end to end may be better positioned to design the solution than someone receiving the requirement several steps later through an engineering backlog. AI is beginning to close the technical gap between knowing the problem and building something to solve it.
He illustrated that with Uber Eats. According to Hedin, a sales leader used Lovable to create software that generated customised restaurant pitches. A first version could be built within hours or a day and subsequently changed in minutes. Historically, he argued, the same request might have spent months waiting for engineering capacity.
But Martin then pushed the conversation toward the uncomfortable consequence of that freedom: what happens when thousands of people inside a company can suddenly create software?
His question was effectively whether security teams were watching all this new code appear and panicking.
Hedin acknowledged that sometimes they were, and sometimes they had good reason to.
That may be the most important part of Lovable’s story now.
Hedin said people inside roughly two-thirds of Fortune 500 companies are using Lovable in some capacity, while Deutsche Telekom alone has built more than 2,000 applications, according to his account. Once software spreads that widely inside an enterprise, the challenge changes from making creation possible to deciding who should build what, which data an application should see and what permissions follow when other employees are invited into it.
The irony is difficult to miss. AI gives non-engineers more power precisely because they no longer need to understand every technical detail. But those same users are also unlikely to be security engineers.
Lovable is therefore trying to automate some of the responsibility that used to sit with specialists. Hedin said the platform uses AI to scan projects for security issues not only while they are being created but continuously afterward. If a dependency later develops a critical vulnerability, Lovable can detect it and, depending on severity, fix it automatically or propose a repair. At more than a million projects being created each week, manual review would be impossible.
The same philosophy appears in another part of Lovable that users rarely see: model selection.
Martin asked why Lovable does not expose the kind of model picker increasingly common in AI products. Hedin’s answer was that choosing a model is itself becoming a machine problem. A single task can involve more than 100 models under the hood, he said, with Lovable balancing cost, speed and quality depending on what the user is trying to accomplish. A small change should not automatically be sent to the most expensive frontier model simply because a human selected it from a menu.
That reveals where Lovable’s ambition has moved. The company increasingly does not want users thinking about code, model routing, hosting, deployment, security or the infrastructure required to keep an application alive. Hedin described Lovable Cloud as expanding into areas including payments, email, marketing and SEO because customers who successfully build a product immediately encounter the next problem: running it.
And that may be the real meaning of the HumanX session title.
When the barrier to building software falls, software itself does not become trivial. The bottleneck simply moves.
Writing code becomes easier; deciding what deserves to be built becomes more important. Creating an application takes hours; securing, maintaining and governing thousands of them becomes the harder problem. Model choice disappears from the user interface; the platform underneath has to become better at making those decisions automatically.
Even Lovable’s decision to remain in Sweden fits that pattern. Martin asked Hedin why a company with Lovable’s ambitions did not simply move to San Francisco for easier access to capital, compute and the Silicon Valley ecosystem. Hedin said he and co-founder Anton Osika considered it, but concluded that Europe offered advantages in talent and working culture while the company could still adopt the speed and ambition associated with Silicon Valley.
The geography of building companies may be loosening for the same reason the technical barrier is loosening: more of the scarce infrastructure is becoming accessible from anywhere.
Martin began by asking what comes after vibe coding.
Hedin’s answer was “agentic coding,” but the conversation ultimately suggested something broader. Lovable is no longer merely trying to help more people write software. It is trying to make the machinery around software invisible enough that the person closest to a problem can build the solution, while the platform quietly carries more of the complexity that professional software development used to demand.
