Imagine asking an AI to write you a novel.

You describe the feeling you want — the tone, the genre, the vibe. It produces something. And honestly? It's good. The prose flows. The characters breathe. A reader picks it up and enjoys it.

But you didn't write it. You don't know what's on page 47, or why the author made that structural choice in chapter three, or how the subplot in act two was set up to pay off in act four. You don't know the rules the book is playing by — because you never made them.

Now someone asks you to add a chapter.

You go back to the AI. It writes something. But it doesn't know what it wrote before — not really. It guesses at the thread. It gets the tone close. But the new chapter contradicts something on page 47, breaks the subplot, and introduces a character who shares a name with one who died in chapter six.

You don't catch it. Because you never knew the book.

You publish it anyway.

Readers start finding the cracks.

This is vibe coding.

Not the tool — the approach. The pattern of building without understanding. Of shipping what feels right without knowing what's underneath. Of trusting the output without reading the architecture.

The code works. Until it doesn't. And when it doesn't, you have no footing. You can't debug what you never understood. You can't extend a system whose logic was never yours. You can't maintain a structure you never built.

So you go back to the AI. It patches the crack. And creates three more you won't find until they're load-bearing.

The book analogy holds further than it looks.

A novelist doesn't just write words — they hold a world in their head. They know which details are decorative and which are structural. They know what the story can survive and what will collapse it. They make a hundred invisible decisions per page, most of which never appear on the surface.

That knowledge is what makes expansion possible. That knowledge is what makes repair possible. Without it, every addition is a guess. Every fix is a risk.

A system architect holds the same thing. Not just the code — the reasoning behind it. The tradeoffs made, the edge cases considered, the failure modes anticipated. The parts that look arbitrary but aren't. The constraints that look like limitations but are actually load-bearing walls.

You can't vibe your way into that. It has to be built.

AI is a tool. A powerful one. Used well, it accelerates work that a knowledgeable person is already doing — the person who will read what it produces, catch what it missed, and integrate it into something coherent.

Used as a replacement for that person, it produces books nobody can finish writing.

The code ships. The system runs. And somewhere down the road — when the client needs a new feature, when the traffic spikes, when the edge case arrives — the bill comes due.

And the author isn't there to pay it. Because there was no author.

There was only a vibe.

I build systems I can explain, extend, and defend — because I wrote them.