As someone who’s worked alongside coders (or, as they used to be called, programmers) for most of my professional life, I’ve often found myself admiring the trance-like, keyboard-pummelling fluency of ostensibly unremarkable colleagues as they slip seamlessly into a state of profound creative union with their computer.
The whole concept of no-code must be a bitter pill to swallow for the coder community. But this is what paradigm shifting technological progress looks like, right? And, ultimately, I’m very conscious that, however accessible so called no-code tools make previously complex dev tasks to technological dilettantes like me, it will almost always take the hard won technical expertise of a skilled coder to make the most of them.
Feel the vibes
In case you’ve somehow missed the hype, vibe coding is the AI-adjacent trend of building software by describing what you want in natural language – no rigid specs, no structured prompts, just vibes. It’s about leaning into conversational inputs and having an AI generate the scaffolding, design, and even functionality based on the mood and intent of your prompting. Powered by multi-agent AI systems or LLM copilots, vibe coding aims to lower the barrier to building by replacing technical planning with creative intent – turning ideas into products without ever touching a terminal.
While the hype around vibe coding is unsurprising – AI-driven dev tools like Cursor, Windsurf, and Lovable promise to help ambitious non-technical types fumble their way from concept to product – it’s also quite deceptive. If you’re a preeminent computer scientist like Andrej Karpathy, who coined the term, its wafty evocation of an effortless flow state and miraculous productivity may well ring true. If you’re a relative noob, like me, this promise will more often that not deliver you to the cusp of hitherto inconceivable accomplishments then leave you marooned in a sea of frustrated incomprehension.
Perhaps inevitably, vibe coding’s initial viral moment quickly met with an angry backlash. Established developers were quick to voice their scepticism, even going so far as to denounce it as a ‘dangerous fantasy‘. And, to be fair, the linked post on Namanyay’s blog does a good job of backing up the alarmist tone of its headline.
His first example – a self-declared ‘non-technical’ vibe coder claiming to have built his SaaS business using Cursor with “zero hand-written code” – provides a pleasingly succinct parable. The poster’s (somewhat hubristic) vibe has more recently shifted to a tone of panicked disillusionment. A recent post complains that “random things are happening… ” – maxed out API keys, database corruption, and bypassed subscriptions.
Conceding that he’s not technical (a fact that, not so long ago, had been central to his triumphant no-code use case) the vibe coder is now confronted by a dawning realisation – he’s built something that he doesn’t understand. Namanyay observes that “there’s a growing and dangerous fantasy that technical knowledge is optional in the new AI-powered world.”
Can you really build an app on vibes alone?
The short answer is: sort of – but not without hitting friction.
Vibe coding works best when the AI is asked to build familiar, well-scoped things. A basic to-do list app with a pastel UI? Sure. A rudimentary clone of a classic Atari game? No problem. These are the sweet spots: projects that have been built thousands of times before, with plenty of examples in the AI’s training data.
But when you go beyond that – when you try to build something even slightly novel or more complex – the wheels start to wobble. Vibe coding thrives on confidence and fluency, but it struggles with nuance. AI can generate a decent-looking front end, or scaffold a Node backend, but it’s not going to intuit your business logic or architect your app with any long-term maintainability in mind. You’ll still need to step in and make sense of what’s going on under the hood.
Noobs beware: it’s not (yet) no-code
Much of the hype around vibe coding frames it as a way for non-coders to “just build.” But most of the time, the experience isn’t as frictionless as advertised. You may not need to know how to code to start, but you’ll probably need to learn the basics if you want to finish.
That’s because even when AI writes functional code, it rarely writes robust code. Bugs crop up. APIs fail silently. CSS looks wrong on mobile. And if you don’t know how to read or fix what the AI has generated, you’re stuck. Debugging in vibe coding workflows is still a technical job – and in some cases, a more frustrating one, since you’re dealing with code you didn’t write and don’t fully understand.
Security is a blind spot
Security isn’t a particularly vibey concept is it? If anything, it’s inconveniently rigid and a bit of a ball ache – the ideological enemy of vibe coding. And sure enough, AI-generated apps often ignore best practices entirely: authentication flows are cobbled together, input validation is minimal, and dependencies are pulled in without vetting. The result? Code that runs, but might expose users (and developers) to risk. It’s one thing to build fast – it’s another to ship code that leaks user data or gets you flagged for using unsafe packages.
Will anyone pay for your vibe-coded app?
Probably not – at least not yet. While vibe-coded prototypes can look slick, they often lack the stability, scalability, and polish needed for commercial use. These tools are brilliant for idea exploration, weekend hacks, and internal tools – but building a viable startup on top of AI-generated code? The prevailing wisdom suggests it wouldn’t be wise – yet.
There are exceptions, of course. Some developers are using vibe coding to rapidly test product-market fit or mock up MVPs to raise funds. But the road from “working prototype” to “real product” is still paved with a lot of manual refactoring and rewriting.
Still want to vibe code?
If you’re determined to jump into vibe coding, great – but go in with your eyes open. A few pointers:
- Treat it like a collaborator, not a miracle worker. Use AI to get things moving, not to do all the work.
- Review everything. Don’t trust the AI’s code blindly. Check for bugs, performance issues, and security holes.
- Keep it small. Vibe coding works best for tight, focused projects – not sprawling systems.
- Learning a bit of code goes a long way. Even if your goal is to avoid coding, knowing the basics will make you dramatically more effective.
Ultimately, it’s healthy to think of vibe coding as less a revolution and more a creative accelerator – a new way to think about prototyping and ideation. The real magic still comes from knowing what you’re trying to build and being willing to wrestle with the details when the vibes run out.
