Vibe Coding Gets You Started. Systems Get You Results.

The 80/20 trap

There’s a term floating around called “vibe coding.” The idea is simple: you tell an AI what you want, it writes the code, and you ship it. No engineering knowledge required.

The exciting part is that it genuinely works, up to a point.

AI is impressive at getting you 80% of the way there. You describe what you want, the model produces something that looks right, and you think “this is incredible, I just built an app.” But then you try to change something. Or it breaks in a way you don’t understand. Or you realise the last 20% is where all the actual complexity lives.

That remaining 20% doesn’t take another ten minutes. It takes twenty hours. And the final stretch after that? Easily another forty. That’s the gap between a demo and something you can actually rely on.

The tools are powerful, with the right direction

I use AI every single day to write code, plan systems, and debug problems. It’s extraordinarily powerful. But it’s powerful because I know what to ask for. I understand the architecture. I can spot when the AI gives me something that looks right but isn’t. I know when to push back and when to accept its suggestion.

That’s not vibing. That’s engineering with better tools.

The difference isn’t technical ability. It’s knowing how to direct the AI. Even experienced engineers need to learn this. It’s a different skill from writing code, and it’s one that matters just as much for non-technical people automating their workflows.

Understanding what AI is good at

Here’s something worth knowing about how these models work. They’re trained on the internet. All of it, more or less. That means they’ve absorbed the collective knowledge of millions of developers, writers, designers, and everyone else.

The result is that they’re competent at almost any topic you throw at them, including coding. But competent and expert are different things. A competent developer can’t build you a flawless CRM in a week. An LLM isn’t going to build it in a day.

This is where your direction makes the difference. With poor guidance, you get generic results. With clear, structured guidance, you can push the output well beyond what the model would produce on its own. But you need to know what good guidance looks like.

The foundation comes first

Experimentation is valuable. Play around with it, explore what’s possible. That curiosity is how most people discover what AI can do.

But your time is better spent learning the foundations alongside that experimentation. Learn how to communicate with AI effectively. Understand what it’s good at and what it isn’t. Build the skill of giving clear, structured instructions that get consistent results.

Once you have that, everything else gets easier. The tools make more sense. The outputs get better. You start making real progress instead of circling the same problems.

This applies whether you’re a developer trying to be more productive or a business owner trying to automate parts of your workflow. The skill is the same: learning to lead the AI rather than hoping it leads you.

Beyond the shortcut

Vibe coding is a great starting point. It gets you to that exciting 80% fast, and that momentum is genuinely useful. But if you’re trying to build something you can rely on and grow, the next step is building the systems and skills that close the remaining gap.

That’s where the real results come from.

If you want help building that foundation, that’s where I start with every engagement.