Programmers and developers traditionally build applications through time-consuming, complex coding that requires patience. This involves spending hours reading documentation and debugging code line by line to solve problems that inevitably arise.

Vibe Coding, however, simplifies the development process by allowing developers to prompt AI LLMs (Large Language Models) to generate code and build applications entirely from scratch. Through conversational prompts with AI chatbots, these requests are translated into functional code and applications.

AI can create development environments, handle frontend and backend development, manage UI/UX design and logic, debug issues, and significantly speed up the development process. But is it worth all the hype?

I've been watching this whole vibe coding phenomenon unfold, and honestly, I'm not sure what to make of it yet. On one hand, the idea sounds amazing anyone can now describe what they want to an AI, and it builds them an entire app. On the other hand, well... that may not be the reality. Let me share what I've been discovering.

What Got Me Curious

Here's what initially caught my attention: AI coding tools are everywhere now. Stack Overflow's latest survey shows 62% of developers are using AI in their work—that's up from 44% just last year. That's a massive shift in a short time.

I've experimented with these tools and tried to involve them in my developer workflow, using LLMs (Large Language Models) like ChatGPT and Claude to create code and build app components for me. The concept is simple: you write a prompt, and the AI builds you an entire app or component from scratch. Sounds incredible, right?

Reality Check

Here's where it gets interesting, though not necessarily in a good way. I tried building a few applications, and while the results looked impressive at first glance, they felt more like smart templates than genuine applications. The functionality was limited, and there were underlying issues that weren't immediately obvious.

Different Perspectives

I've been listening to what other developers are saying, and there's quite a range of opinions. That seem to align with my thinking.

Salma Alam-Naylor has a great analogy about AI applications being "slop". She makes the analogy of using AI is like hiring unqualified kitchen fitters who use the latest tools to fit your new kitchen. The tools might be advanced, but without proper knowledge, you get "cowboy builder" results. She argues that relying on AI for app development often leads to disposable, unreliable products when something inevitably breaks, there's no accountability or expertise to fix it, resulting in a cycle of throwaway software that ultimately fails to serve real users.

David Bushell talks about what he calls "ensloppification" this rush to implement AI everywhere without really thinking about whether it's solving real problems. He critiques the current trajectory Big tech is driving us towards saying: "Frankly, I'd rather quit my career than live in the future they're selling. It's the sheer dystopian drabness of it. Mediocrity as a service."

Kevin Powell raises the same concerns about developers becoming over-reliant on AI tools without understanding the underlying code they're implementing. He warns that this dependency can lead to lower coding standards and insufficient quality checks on AI-generated code.

His concerns are backed by real-world examples: Daniel Asaria from Palantir demonstrated how he could hack into multiple Lovable-launched sites (One of these prompt app builders) in under an hour, accessing sensitive data, API keys, and even sensitive information.

The underlying issue is that these AI systems, while trained on millions of lines of code and getting better at predicting what should come next, also hallucinate they create code that looks right but isn't. They reference functions that don't exist or miss the bigger picture of what you're trying to build.

But then there are developers reporting 26% productivity increases, so clearly some people are finding genuine value in these tools.

The Hype Question

What's fascinating is how much investment and excitement surrounds this space. Everyday we learn of new AI tools we can use to improve are daily lives. Like, having AI summarise our emails, and sms messages. Who Cares? Code-generated startups are getting sky-high valuations with promises of replacing expensive human engineers. The buzzwords are everywhere: AI agents, streamlined workflows, MCP servers, large language models.

Recently, Builder.ai, the London-based startup that was supposedly AI-powered but was actually just 700 Indian engineers pretending to be AI bots, filed for bankruptcy after the fraud was exposed. The company was once valued at $1.5 billion and backed by Microsoft. It makes you question what's real and what's marketing hype.

Where I'm Landing

After experimenting with these tools and reading about others' experiences, I'm in a cautious but curious place. I think there's real potential here, but maybe not in the way it's being sold.

The idea that anyone can prompt their way to a functioning application feels like fantasy to me. You still need to understand what you're building and why. These tools seem most useful when you already have a solid foundation of knowledge they're assistants, not replacements.

I'm going to keep experimenting with AI coding tools because I believe they'll continue improving. But I'm also keeping my traditional coding skills sharp. Maybe the future isn't about replacing developers entirely, but about augmenting what we can do in specific areas where these tools genuinely add value.

The security and quality concerns are however are real, though. Until those get sorted out, I'm treating these as assistants rather than replacements, and I'm definitely verifying everything they produce.

What I'm Watching

I'm curious to see how this evolves. Will these tools become more focused and reliable? Will the security issues get resolved? Will we find the sweet spot between human expertise and AI assistance?

Right now, the hype definitely seems bigger than the reality, but that doesn't mean there isn't something valuable emerging here. It's just going to take time to figure out what that actually is.

What's your experience been? Are you seeing value in these tools, or are you running into similar issues? I'd be curious to hear other perspectives on where this is all heading.