Vibe Coding Is Reshaping Software

For years, building software meant learning how to code, hiring developers or paying someone else to turn an idea into a product.

AI is starting to break that model.

This week, Swedish AI startup Lovable said it had crossed a $600 million annualized revenue run rate, up from roughly $500 million in June.

That number matters. But what is more interesting is what people are paying Lovable to do.

They are not simply asking AI to autocomplete a few lines of code.

They are building products.

From writing code to building businesses

Lovable belongs to a growing category of AI tools sometimes described as “vibe coding” platforms.

The idea is simple.

Instead of manually writing every line of code, a user describes what they want in normal language:

“Build me a booking platform.”

“Create an inventory dashboard.”

“Add payments.”

“Change the homepage.”

The AI handles much of the technical work behind those instructions.

Lovable co-founder Fabian Hedin said at the HumanX summit in Amsterdam that apps built using the platform are now generating close to one billion visits per month.

He also claimed that roughly two-thirds of Fortune 500 companies are using Lovable, with customers including Microsoft, Nvidia and Deutsche Telekom.

That suggests AI coding is moving far beyond hobby projects.

The money is following

Investors certainly seem convinced.

In August, Lovable raised $400 million at a $13.3 billion valuation, according to the company.

That came only months after another major funding round and brought the company’s fundraising over its recent rounds to more than $700 million.

Read Lovable’s Series C announcement

But Lovable is only one part of a much broader change.

Anthropic, the company behind Claude, analysed around 400,000 Claude Code sessions and found that people were increasingly using AI agents for more complete tasks rather than simply asking them for pieces of code.

Its research found that users typically made the higher-level decisions about what they wanted to build while Claude handled more of the decisions about how to execute them.

Anthropic also said the share of GitHub projects showing coding-agent activity had more than doubled since late 2025.

Read Anthropic’s research on agentic coding

The developer is not disappearing

It is tempting to look at these tools and conclude that software developers are about to become unnecessary.

That is probably too simple.

AI can generate working applications quickly, but production software still has to deal with security, databases, permissions, performance, privacy and maintenance.

Lovable itself is investing heavily in those areas.

The company says it now handles hosting, deployment and scaling alongside application creation. It has also introduced enterprise governance and security tools as more companies allow employees to build internal software using AI.

In other words, generating the first version of an app is becoming easier.

Running that app safely at scale remains harder.

And that distinction could determine which AI coding platforms survive.

The bigger shift is who gets to build

Perhaps the most significant change is not what AI means for professional developers.

It is what AI means for everyone else.

Lovable analysed its own users between January 2025 and May 2026 and said 80% identified themselves as working in non-technical roles.

See Lovable’s Build Economy report

That could have major implications.

A marketing manager who understands a workflow problem may no longer have to wait months for an engineering team to build an internal tool.

A small-business owner could potentially create a custom system without paying for a full software-development team.

A founder could test an idea before raising money or hiring engineers.

The barrier between having a software idea and actually building it is getting smaller.

But easier software could also mean more bad software

There is another side to this boom.

If millions more people can create applications, the internet could also fill with poorly secured software, abandoned products and applications built by people who do not fully understand the systems underneath them.

Hedin acknowledged those security concerns at HumanX, particularly as companies allow employees outside engineering departments to create their own tools.

That makes governance one of the next big battles in AI coding.

It is no longer enough for a platform to say, “Look how quickly we can build an app.”

The harder question is:

Can that app still be trusted when thousands of people are using it?

Software development may be becoming something else

The phrase “vibe coding” originally captured the almost playful experience of telling an AI what to build and watching software appear.

The industry is already beginning to move beyond the phrase.

Hedin now prefers “agentic coding” — software development in which AI agents take on increasingly large parts of the building process.

Whatever it is called, the direction is becoming difficult to ignore.

The next wave of software companies may not be built entirely by teams of programmers.

They may be built by designers, entrepreneurs, accountants, marketers, doctors and people who understand a problem well enough to describe the solution.

Coding is not disappearing.

But for the first time, knowing exactly how to code may no longer be the only ticket required to build software.

And that could turn out to be much bigger than the latest AI startup valuation.