Beyond the hype cycle: Why retention and community will decide the winners in vibe coding

Vibe coding is a relatively new term people are using to describe building software through natural language prompts instead of traditional programming. The term first entered commercial consciousness in a X post by OpenAI co-founder Andrej Karpathy in early 2025, marking a shift in how people interact with code and software creation.

Instead of writing lines of code, you type what you want an app or feature to do in plain English (or another language), and an AI system generates the code or even the full application for you. It's often seen as the next evolution of low-code/no-code platforms, powered by AI. It promises to democratise software creation, allowing anyone to build products, reduce costs and time, as AI handles repetitive coding, and shift competition from pure technical ability to product quality and user experience.

However, it also presents challenges such as technical debt, enterprise readiness, and the risk of a market flooded with replicable tech without commercial traction.

Thomas Cuvelier is a partner at early-stage VC RTP Global and was an early angel investor in the red-hot coding startup Lovable, which in July 2025 raised $200 million in a Series A round, reaching a valuation of approximately $1.8 billion just eight months after founding.  I spoke to Cuvelier to learn more. 

Where are we in the hype cycle for vibe coding?

According to Cuvelier, we're still early. 

"This isn't hype based on nothing — it's hype based on revenue. Companies are scaling quickly, going from one to 10, one to 20. The growth is real, so I don't think we're at the peak yet."

According to Cuvelier, most vibe coding users are non-technical people:

"We see a lot of school kids creating simple apps and products with zero technical knowledge."

The second group consists of technical or semi-technical users who utilise it to speed up coding. They might still go into the code to make adjustments for more complex applications, but the repetitive work is abstracted away. While Cuvelier predicts that most markets will be big, his interest lies in giving "superpowers" to the 99 per cent. 

"These tools let anyone become a developer. No one should have that moment of thinking, 'I have a great idea, but I need to find a developer and I don't know how.' That's a huge problem to solve."

However, AI code tools can’t escape technical debt or the need for review

I wondered about the problem of vibe coding generating technical debt, in the context of the hidden "cost" of quick coding decisions that trade long-term stability for short-term speed. "That's the biggest issue with vibe coding today," shared Cuvelier. 

"These products are great for simple apps, but enterprise use is another matter. They can create a significant amount of technical debt, and the code still requires review.

Tools like GitHub Copilot have been generating code for years, but you still need to review it. That won't go away anytime soon. Right now, the apps aren't entirely enterprise-ready. This will get fixed in time, but today it's a limitation."

What about explainability?  Coding is also about communicating and reasoning through ideas—pair programming, design processes, and peer reviews. Does a vibe coding approach change that?

"You'll always need an audit trail," contends Cuvelier. 

"These tools will increasingly include features showing how something was built — like an ingredient list on food packaging. Transparency will remain essential, even as more gets automated."

Vibe coding to fundamentally change how entrepreneurs work

While Cuvelier contends that technical skills are indeed necessary for startups, he argued that vibe coding democratises entrepreneurship.   

"Anyone — a business leader, a biochemist — can now get started.  Over time, people won't compete on who has the best technical team, but on who builds the best product. User experience will be the differentiator, and the winners will ultimately be the users."

Cuvelier contends that the core concepts will remain, such as building products and adding features, but developers will spend less time on repetitive tasks and more time on design, user experience, and creativity. 

He suggests that education must change, too.

"If you're learning to code today, you should also learn how to use vibe coding tools alongside traditional coding."

Looking forward, Cuvelier predicts that as coding becomes accessible to everyone, the cost of building software will fall close to zero, like the cost of storing data

"We'll see more apps, more products, and much better quality overall, because competition will push companies to focus on user experience. Incumbents with average, clunky software will be disrupted.

This shift will be visible within the next 12 to 24 months."

Cuvelier: In 24 months, software costs could fall to zero — and incumbents should worry

However, more access to product-building tech means more products and even with easy tools, teams still need to focus on building something users actually want.

Cuvelier admits, "That's always the hardest part. It will come down to empathy — understanding how consumers think and giving them a great experience. You won't be able to get away with forcing users to adapt to bad software or locking them in."

It's somewhat reminiscent of how "plug-and-play" and no-code transformed industries like IoT. Suddenly, traditional companies could implement digital tools in-house, without huge deployments. AI is finally making that possible.  

However, Cuvelier cautions that incumbents should be worried:

"If you're a large company with mediocre software, and you've survived by locking in customers, you're in trouble."

Cuvelier sees consumer applications as particularly ripe for disruption, but also enterprise tools like CRM, accounting, and productivity software — "basically anything you use every day that you don't like."

"These tools will be disrupted quickly. Some areas, though, are more protected, like apps with strong network effects. 

For example, it's going to be hard to build something better than WhatsApp because all your contacts are already there."

Retention is the ultimate metric

As an investor, Cuvelier looks for startups with retention.

"Do users stick with your platform? That's the key metric," he shared. 

While a few dominant players are already emerging with strong communities, and they're hard to catch up with, there's still room for niche players—startups that specialise in verticals or unique use cases. 

"If you can build community and keep users engaged, that's a good sign."

Ultimately, Cuvelier contends that Europe's strength lies in its talent. 

"We have deep pools in AI and machine learning. The startup ecosystem is improving, with more funding available, and companies like Lovable show that European startups can attract strong backers. 

But Europe still lacks the large-scale funding rounds —€10, 20, 50 million — that Silicon Valley can provide.  That needs to change.  At the same time, the US is a huge part of the market, so European companies must think globally from day one."