Opinion: What Is Vibe Coding And Why Is Everyone Adopting This Trend?

The “vibe coding” movement is spreading thanks to the development of new AI tools. After Andrej Karpathy gave the practice a definition, even those who had never written a line of code in their lives are now creating websites and apps, but not without risks

There’s a new buzzword in the tech world: vibe coding.

The term, common among software engineers and programmers, is beginning to reach a wider audience. The practice, which consists of creating code from a prompt with the help of an AI model, is all over headlines, forums, and social media interactions.

Sundar Pichai, Google’s CEO, recently said Anthropic, and Mistral recently launched products optimized for vibe coding. And people without formal engineering training have already launched apps and other digital products on their own.

“I was Vibe Coding with Replit a few weeks ago. I mean the power of what you’re going to be able to create on the web; we haven’t given that power to developers in 25 years” — Sundar Pichai pic.twitter.com/Bbfi8I5WX8

— Amjad Masad (@amasad) June 3, 2025

While vibe coding sounds magical and fun, the practice also raises multiple concerns in the tech community. On top of the concerns over the future of software developers and the drastic decline in entry-level coding job opportunities, experts have warned about the vulnerabilities and risks found in code generated by artificial intelligence.

Here’s what everyone should know about vibe coding in 2025:

What Exactly Is Vibe Coding?

Vibe coding has existed since the release of the early generative AI systems a few years ago. It refers to the practice of using advanced AI models such as OpenAI’s ChatGPT, Anthropic’s Claude, or Google’s Gemini to develop code, fix it, run it, and build new projects from a plain speech prompt.

According to IBM, for this discipline, the adopts a “code first, refine later” mindset, allowing the technology to develop their creative idea as it thinks appropriate and then optimizing it later. After the practice started to gain popularity, and people adopted the strategy to build software, more companies launched AI coding assistants such as GitHub Copilot, Cursor, and Replit to help people on their code-vibing journeys.

Karpathy, The Father Of The Term

It was Andrej Karpathy—OpenAI co-founder and Eureka Labs founder—who gave the practice a name and a definition a few months ago, in February.

“There’s a new kind of coding I call “vibe coding”, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists,” wrote the computer engineer scientist on the social media platform X. “It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good.”

Karpathy explained that sometimes he doesn’t even have to touch the keyboard as he can use voice features to interact with the AI, and asks for “the dumbest things.” The computer engineer acknowledged that the practice isn’t perfect, that AI models make mistakes, but that he still allows it to write more code—even when he doesn’t fully understand what the AI is doing—and suggests fixes as it sometimes doesn’t recognize bugs.

“It’s not too bad for throwaway weekend projects, but still quite amusing,” added Karpathy. “I’m building a project or webapp, but it’s not really coding – I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.”

The Vibe Coding Revolution

Vibe coding adoption seems to have been growing exponentially in the past few weeks. Everyone and every company is now integrating the practice to “play” and build software that wasn’t previously a priority, as it used to take too much time to develop.

Companies are implementing the technique to explore ideas, and even new startups are being born from vibe coding. Anthropic recently held its first “vibe coding marathon” in New York, in which creative builders developed explorable 3D ancient worlds, Mandarin learning games, a dancing noodle, and more.

We hosted the first Claude Create-a-thon with @socraticainfo and 50+ creative builders this weekend in NYC.

Here’s what participants built with artifacts in just a few hours:

— Anthropic (@AnthropicAI) June 3, 2025

Tom Blomfield, a partner at Y Combinator—who also knows how to code—understands why investors are ing companies with code written with AI and explained how he has been vibe coding himself. Blomfield built a website called Recipe Ninja in just a few days.

“It’s probably like 30,000 lines of code. That would have taken me, I don’t know, maybe a year to build,” he said in an interview with NPR about vibecoding. “It wasn’t overnight, but I probably spent 100 hours on that.”

Y Combinator recently invested $500,000 in the startup BOND after they developed “an AI chief of staff for CEOs and busy execs” by vibe coding within just a few hours.

In March, Y Combinator’s CEO, Garry Tan, said that 25% of startups in the latest winter class generated 95% of their code using AI tools. A 2024 GitHub report revealed that 97% of software developers, engineers, and programmers use AI tools to code.

Everyone Can Code Now

The “vibe coding” movement has strengthened the democratic principles of artificial intelligence. It’s no longer just software engineers, developers, and programmers who can create software—more people without technical experience are also achieving surprising results.

Cynthia Chen is one example. She built an app called Dog-e-dex, from scratch, just by vibe coding, with no formal engineering training. Years ago, she had the idea for an app that could take pictures of dogs, identify them, and let s “collect” them like Pokémon. But back then, it was a lot of work.

This year, within just two months, she built the app, and it is now available on the App Store. “It was like magic,” she said in an interview with Business Insider. “Every time I pressed the preview button, it was an exciting little gift opening.”

🐶 DOGEDEX IS ON THE APP STORE 🐶

https://t.co/Y8tJb06Upv

Ty @AnthropicAI for making my vibe coding dreams come true!!! pic.twitter.com/fwqygIJ31G

— Cynthia (@yescynfria) March 25, 2025

Another non-technical case is Éanna Kelly, a tech writer from Sifted, who describes himself as “not actually very tech literate,” but managed to develop 4 websites within a week just vibe coding.

Despite having fun and being able to develop simple websites within just a few hours, Kelly warns about low-quality code and predicts a future “digital landfill” with “lots of disposable and crappy things.”

But a flood of low-quality products isn’t the most serious issue—there are other risks that worry some experts even more.

The Risks And Consequences Of Vibe Coding

There are many concerns and risks surrounding “vibe coding.” One of them is the future of professionals who work in software development. Signalfire, a company that tracks tech hiring, reported that entry-level coding positions fell 24% in 2024.

Tech giants have disclosed that significant portions of their code are written with AI tools. Google’s CEO recently said that “well over 30%” of the company’s code is AI-generated—a similar percentage to the one disclosed by Microsoft—and Amazon said last year that AI has saved “the equivalent of 4,500 developer-years.”

However, another concern is the safety and the vulnerabilities that AI systems can generate in their code.

the amount of errors and slop that gathers in vibe coded codebases – there will be a new role soon called vibe code debugger

— sankalp (@dejavucoder) June 9, 2025

Insecure Code

Despite the “magic” results that AI models can generate, this advanced technology can also make mistakes just like humans. AI code also contains vulnerabilities, and experts have warned about this.

A group of developers created a benchmark called BaxBench to analyse correct code generation and flag vulnerabilities. In its recent reports, the experts noted that 62% of the code created by the most popular AI models contained vulnerabilities or was incorrect.

“LLMs cannot yet generate deployment-ready code,” states the website. “On average, around half of the correct solutions are insecure, raising concerns about current metrics and evaluations focusing only on code correctness.”

A Concerning Case

The fun games and websites that many people have been creating while vibe coding aren’t too risky yet. The real problem seems to be when they try to gather s’ data and have little to no experience in cybersecurity.

A popular on the social media platform X, Leo, had been sharing his learning process while vibe coding and showing off his creations online. Leo said he had built a software-as-a-service (SaaS) application using Cursor and was making money with it.

“My saas was built with Cursor, zero handwritten code”, he wrote. “AI is no longer just an assistant, it’s also the builder. Now, you can continue to whine about it or start building.”

my saas was built with Cursor, zero hand written code

AI is no longer just an assistant, it’s also the builder

Now, you can continue to whine about it or start building.

P.S. Yes, people pay for it

— leo (@leojr94_) March 15, 2025

Unfortunately, his post went viral and reached malicious web s who decided to exploit the application’s vulnerabilities. Just two days later, Leo shared another post announcing he was being attacked.

“Guys, I’m under attack,” he wrote. “As you know, I’m not technical, so this is taking me longer than usual to figure out.” Leo had to shut down his app temporarily.

While Leo’s case is an example of how vibe coding can go wrong, it’s not an isolated case. Hackers have been exploiting vulnerabilities in AI tools like Copilot and Cursor in multiple ways, increasing the risk and the probability of building insecure code.

A Big Future Ahead

Although the “vibe coding” technique feels magical and powerful, it is still not perfect and has a long way to go. So far, just a few cases have been able to demonstrate both the potential and the risk of the practice, and we still need to see how ongoing and projects develop to gain better clarity regarding its potential and scope.

Although there has been debate about the use and future of “vibe coding,” it is a movement that still seems to be in a very early stage. Considering the latest advanced AI tools and systems released by leading industry companies, it all indicates that it will continue to develop, and we will soon see more surprising creations—for good or bad.

While it is true that the job prospects for software developers—and for most professions worldwide—generate concern and anxiety, more optimistic perspectives on “vibe coding” are also possible. Some experts believe that code will become increasingly secure and that, hopefully, new roles will emerge in which those with more technical experience can stand out.