Software development might soon be powered by just "Pure Vibes"

In 2023, ex-OpenAI co-founder Andrej Karpathy made a bold prediction: English might become the hottest new programming language. A year and a half later, and you can’t help but feel like we’re inching closer to that reality.

The vision Karpathy painted of a world where natural language could be used for programming is finally taking shape. And in February this year, a new phrase entered the lexicon: “vibe coding.”

All this from a casual tweet that read: “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. It’s possible because LLMs (e.g. Cursor Composer w/ Sonnet) are getting too good. Also, I just talk to Composer with SuperWhisper, so I barely even touched the keyboard.”

That offhand post lit a spark. Suddenly, a generation raised on TikTok trends and instant creation had their programming “cheat code”: building software without knowing how to code.

What Is Vibe Coding?

If you haven’t bumped into any of the sloppy AI-generated demos on Twitter, vibe coding is the practice of using AI models (LLMs) to generate code all from natural language prompts without painstakingly typing out every line yourself. Instead of, say, memorising JavaScript syntax, you can simply describe what you want like: “Build me a mobile app that reminds me to drink water, with a pastel colour scheme,” and the AI takes it from there.

The AI interprets the request, spits out a codebase, and you, in theory, have an app. No debugging tutorials. No stack overflow rabbit holes. Just vibes. Honestly, I don’t know if there’s anything more appealing to Gen Z than this.

What this implies is that a random person with zero coding experience can now ship code faster than a professional with 20 years of expertise.

The interesting thing is that the tech being vibe-coding isn’t even brand-new when Karpathy coined the phrase. Major players like Microsoft, Google, Anthropic, IBM, AWS, and of course OpenAI were already busy baking natural language capabilities into their developer platforms. GitHub Copilot had been helping devs autocomplete code inside VS Code. Replit was already experimenting with AI integrations. Even newer names like Cline, Cursor, Aider, and Zed have emerged as AI-first developer tools doubling down on the same concept.

But sometimes the tech doesn’t take off until it has a name, or a cultural hook. Karpathy’s tweet gave the idea branding, and like they say, "branding matters." Now, the phrase “vibe coding” has made AI programming sound approachable, even fun, for people who never thought they could build apps.

Why Is Vibe Coding Important?

This past week, I was out jogging with a friend who’s a virtual assistant and does code, and when I asked what he’d been into lately, his answer surprised me: vibe coding. It became apparent that this thing is already pulling in people who would’ve never gone near a code editor before.

The truth is, vibe coding is lowering the barrier of entry in a way that nothing else has. People unfamiliar with code, or even afraid of it, now feel like they can build something. It’s sparking a new kind of excitement: apps that can be built quickly, from scratch, by anyone who can describe an idea clearly enough.

The catch, however, is that the quality of the output depends entirely on the quality of the input. Which means prompting becomes a skill in itself. You have to think through requirements, design patterns, testing implications, and even long-term maintenance before you hand it off to the AI.

But it is not only getting popular among non-coders. Experienced engineers are also using these AI tools as a productivity multiplier. For the first group, it’s liberation. For the second, it’s leverage. Of course, this doesn’t mean AI replaces the traditional coding done by developers entirely. What it does change is where developers spend their energy. With vibe coding, they can hand over the tedious stuff but with oversight and spend more time on higher-level design and problem-solving.

There’s no denying the allure. For many digital natives and non-coders, vibe coding is the perfect gateway drug into tech. Why spend months learning JavaScript when an AI can spit out a full app in minutes? It’s like driving for the first time, but starting with cruise control already on.

Is Vibe Coding a New Era of Coding?

So does vibe coding mean we’ve officially entered a new programming era? In some ways, yes. The software development life cycle is always shifting, and vibe coding is just the latest chapter in that shift.

And if you zoom out, you'd see that the pattern. Waterfall gave way to agile and manual deployments to DevOps. Each iteration has reshaped the developer’s role, introducing new tools and methodologies to tackle emerging challenges. AI’s integration into coding, whether through vibe coding or more structured approaches, is just the latest chapter.

It is the same shift that we've seen happen in web development. We’ve seen it with no-code platforms like Webflow, with website builders like WordPress, and even with Photoshop and image generation. Each time, the purists said, “This isn’t real design” or “This isn’t real .” Yet, those tools didn’t erase expertise; they've just expanded who could participate. Vibe coding is shaping up to do the same for programming.

And some predict it could go even further. Analyst Holger Mueller wrote in a 2023 report that within five years, the developer role as we know it will start to fade — and in 15 years, it could disappear altogether.

“Effectively, the move from keyboard to voice and from code to natural language (NL) means that more software will be able to be built and more business users will be able to put themselves in charge of their automation destiny,” the report reads.

Even Anthropic’s CEO, Dario Amodei, once said, “In 12 months, nearly all code may be generated by AI.”

Call it a hot take, but this shift could democratize software development to a wide range of people who want to build their own applications. Why? Because the barrier to entry won’t be anything sophisticated. It will simply be English.

Think about that: voice over code, natural language over syntax. Software building software. That shift could expand who gets to build applications, putting more business users in charge of their own automation.

But let’s not get carried away. Vibe coding has many limitations — and its vulnerabilities can show up fast.

The Actual Realities of Vibe Coding

For all its promise, vibe coding isn’t just some flawless magic trick.

Yes, it has been established that the benefits are real. It lowers the barrier for anyone with ideas but no coding background. It allows for creative freedom and speeds up prototyping and iteration. It takes care of boring tasks like boilerplate code. But that might just be where the benefits end.

Try to build any scalable product with "vibes", and you will run into a lot of trouble. Just ask Leonel Acevedo, the CEO of EnrichLead. In March 2025, he proudly announced he had built a SaaS product with “zero hand-written code.” Within days, his app was under attack, as attackers exploited exposed API keys in the AI-generated code. He was unable to debug it as debugging the mess without technical expertise turned into a nightmare.

In another case, a developer discovered that Lovable, a “vibe-first” coding platform, was exposing sensitive database credentials in public builds. What was marketed as innovation quickly turned into a security liability.

The risks are obvious. LLM outputs are inconsistent and unpredictable, which makes codebases harder to maintain. It can misinterpret vague instructions and generate entire structures that collapse under real-world pressure. It can introduce subtle bugs or glaring security issues. Without explicit architecture, scaling and maintainability quickly fall apart. And wrestling with prompts can sometimes take longer than just writing the code.

A report from Legit Security in 2025 backs this up. Two-thirds of organizations now use AI tools to help write software, but nearly half use them in risky ways. As they put it, AI has rapidly become a double-edged sword.

That’s why experienced developers aren’t out of the picture. In fact, they’re more important than ever. Someone still has to oversee, test, refine, and make the tough architectural decisions. What vibe coding does is shift their role: less typing, more reviewing, more guiding.

For non-coders, the risk is even bigger. Without the ability to spot errors or think like an engineer, you can quickly hit walls that an AI won’t help you climb. That doesn’t mean it’s useless; it just means vibe coding works best when paired with either strong AI guardrails or human oversight.

And then there’s the cultural side. Some see vibe coding as “cheating,” the equivalent of letting calculators into math class. But we know how that turned out: calculators didn’t destroy math education; they changed it. We spent less time crunching numbers and more time solving actual problems. Maybe vibe coding will push programming in the same direction.

Conclusion

At the end of the day, vibe coding is driving a cultural shift in software development. Whether you see it as revolutionary or reckless, it’s democratizing software creation but also raising new risks. For some, it’s the perfect creative tool; for others, it’s a security breach waiting to happen.

A new generation is growing up believing that building software doesn’t have to be intimidating or gatekept; it can be playful, expressive, and even fun.

The term itself matters, too. “Vibe coding” strips away the intimidation. It doesn’t sound like hard labour or a job requirement. It sounds like a creative process, something closer to making a TikTok than sweating over a compiler. And maybe that’s why it resonates so much.

So yes, English might just be becoming the hottest new programming language. As Karpathy put it, vibe coding might not just be another layer of abstraction; it could be a fundamental shift in how humans communicate with machines.

More importantly, the vibes are here to stay.