‘Vibe coding’: The new wave of AI reshaping digital creation
2025
7 mins
Taking the world of coding by storm, the phenomenon of vibe coding is worth paying attention to – but are there hidden risks?
Written by: Jessie Harrison
What is vibe coding?
In 2023, computer scientist Andrej Karpathy, co-founder of OpenAI and director of AI at Tesla, said that “the hottest new programming language is English.” Now, two years later, he’s living out his vision: he doesn’t need to program at all, just speak. Thanks to advancements in AI, software engineers, website developers and laypeople can build apps and websites using only English language prompts inputted into a large language model, such as those built by OpenAI, Anthropic and Google. You can construct code without any understanding of code or, if you do have an understanding, you can ignore it. This is ‘vibe coding.’
The first instance of the term appeared in a tweet posted by Karpathy in February of this year, where he described ‘vibe coding’ as “fully giv[ing] into the vibes, embrac[ing] exponentials, and forget[ting] that the code even exists.”
The concept caught on. A month later, the term appeared under ‘Slang & Trending’ in the Merriam-Webster dictionary, defined as “writing computer code in a somewhat careless fashion, with AI assistance.” In March 2025, Y Combinator reported that a quarter of its Winter 2025 cohort of startups had almost fully AI-generated codebases. Satya Nadella, Microsoft’s CEO, shared that AI generated 30% of all new code used by the company. And there has been a wave of intriguing AI startups driving this new AI function. For example, Google has just announced their new AI-assisted app-making platform, Opal.
The Swedish startup which soared to success
One startup which has already caused an unprecedented stir with its AI ‘vibe coding’ tool for app and website design is Stockholm sweetheart, Lovable.
Lovable was born from Anton Osika’s pastime project, an AI tool which he constructed in 2023 and turned into an easily usable, visual tool with help from co-founder Fabian Hedin.
“Build a beautiful website in minutes without lifting a finger” instructs the website, against the backdrop of a calming, ambient colour gradient. This simple promise has already seduced millions. As of July, the platform had 2.3 million active users, building 100,000 projects per day.
One of their users was London-based investor and Accel partner, Ben Fletcher, who created an app in the space of a weekend which tracked his pickleball matches, inspiring him to build a data-tracking software for startup sales. He then decided to lead the startup’s Series A funding round. In mid-July, just 8 months after their relaunch, Lovable raised $200m and was valued at $1.8bn, becoming Europe’s ninth unicorn of 2025. Less than a week later, they reached $100m in annual recurring revenue, making them the fastest-growing software startup ever.
As AI has become associated with the American giants, such as Microsoft, Google and OpenAI, and other well-known ‘vibe coding’ startups, such as Replit (valued at $1.2bn) and StackBlitz also stem from America, it is exciting to see such a successful AI pioneer from a European country. According to Sifted’s European Tech in Review report for H1 2025, published last month, AI agents is Europe’s fastest-growing sector, with deal count climbing 226% year on year. The possibilities remain endless.
What is the catch?
However, as great as ‘vibe coding’ sounds, successful websites and apps cannot run on vibes alone. Coding is a complex process, rife with potential complications, and AI may not be as well-equipped to handle these as seasoned professionals.
There are concerns that the quality of code generated by AI may not be adequate, as it is simply replicating patterns rather than applying a deep knowledge of code. Some users of app builders will have limited understanding of the required standards, resulting in security vulnerabilities, undetected bugs and other errors. This is especially true for complex projects which AI does not yet have the capacity to execute flawlessly. Going with the vibes may be sufficient for non-coders playing around in their spare time, but is AI up to the task of writing code for apps and websites for professional use and paying customers?
Nevertheless, there is perhaps a happy medium between pure vibe coding and more ‘old-fashioned’ AI-assisted coding. A state where AI is a partner in the creative process, rather than merely assisting or fully taking over.
Software engineers can integrate AI into their workflow to streamline and accelerate the development process, whilst remaining aware of the code and implementing measures to prevent malfunctions and security risks. Aligning with the principle of vibe coding, AI allows them to work more intuitively, leaving more time for problem solving rather than manual coding.
In addition, AI-generated code can be seen as a gateway to rapid testing, rather than as the cornerstone of the finished product. The opportunity for rapid prototyping is a game-changer for startups and product teams who need to quickly prove the functionality and scalability of their ideas to investors.
‘Vibe coding’ and IP
Another key consideration is how the nature of vibe coding impacts intellectual property ownership and protection.
AI-generated software
As with AI photo-generators and AI chatbots, AI vibe coding platforms use LLMs which are trained on vast amounts of existing data and, therefore, there is a risk that the output could be dangerously similar to protected works. For example, if a generative AI model includes a substantial part of a copyright work in its outputs, and there is no licence in place to allow this, the generated content could infringe copyright.
In addition, there may also be uncertainties surrounding the ownership of source code generated by AI. Current UK law includes provisions for computer-generated works “generated by computer in circumstances such that there is no human author”. The author of such a work is the person “by whom the arrangements necessary for the creation of the work are undertaken”. However, this provision has received criticism due to its lack of clarity and the scope of protection remains uncertain. Certainly, there is little consistent recognition internationally. For example, the US Copyright Office’s January 2025 report highlighted that prompts alone are unlikely to result in copyright ownership. Lack of ownership of copyright in the source code has clear ramifications for protection and defensibility of AI-generated software.
Computer-implemented inventions devised by AI
Patenting computer-implemented inventions devised by AI is also likely to be problematic. Under UK law, computer-implemented inventions, including software and AI inventions, may be patentable if they are novel and make a ‘technical contribution.’ However, the UK Supreme Court, and other jurisdictions, have held that an AI system cannot be named as the inventor in a patent application. This does not necessarily preclude human inventors from using AI systems as sophisticated tools to devise and subsequently patent inventions, but, when seeking to obtain patents at the UKIPO it is essential that the inventor is a natural person.
You can read more about the patentability of AI and how to protect your commodities as an AI startup in our previous article on IP advice for AI startups.
Conclusion
Vibe coding’s popularity goes beyond a buzzword to a fundamental shift in how we create apps and websites, focusing on the ideas rather than the technicalities. AI enables people to bring their visions to life and businesses to build useful software themselves, rather than rely on expensive B2B SaaS, as well as giving professional developers more freedom to unleash their creativity.
Still a young concept, it will be interesting to see if ‘vibe coding’ is here to last. Market research and consulting firm, Gartner, for example, has predicted that, in the next few years, AI will generate 40% of new software for businesses based on English language prompts.
Nevertheless, for vibe coding to be successful on a wider scale, including on a business level, developers and non-developers alike need to be careful not to over rely on AI and remain aware of the potential risks.