Vibe coding is a slot machine

I tried making music with Suno AI a while back. Honestly, the output was impressive. Good rhythm, catchy melodies. AI keeps blowing my mind. Two years ago, I never would’ve imagined living in a world where you can make music with just a prompt. But here’s the weird part: listening to these songs “I made,” I felt absolutely nothing.

When I thought about why, I realized I much prefer tweaking each note in a DAW. In that process, every decision is mine: heavier kick drum here, softer synth there. The final product might not sound as “professional” as what Suno spits out, but every note has my fingerprint on it.

When coding becomes a slot machine

I noticed the same feeling recently while using Claude Code.

The tool is insanely powerful. I can casually say “build me a journal app” without any details, and in minutes it generates a fully functional iOS app. Beautiful UI, clean code, error handling included.

But here’s the interesting thing: every time I run the same prompt, the result is slightly different. Sometimes the design leans minimal, sometimes it’s full of gradients and animations. Layout, features, details. Always a different surprise.

It’s basically a slot machine. Or a blind box. Pull the lever, see what combo comes out. Like playing a roguelike game. Every run is a fresh experience. Addictive.

Sounds fun, right? But as a creator, this randomness feels off. And I keep failing to ship.

Who’s the real creator?

A painting analogy might make this clearer.

If I carefully plan the composition, choose the colors, then let AI handle the execution details, AI is just my paintbrush. The real painter is still me. I control the direction; AI just makes the execution more efficient.

But if I just say “paint me a landscape” and wait for AI to conjure something up, then I’m just a person pulling a slot machine lever. The actual creator is AI, not me.

From a results standpoint, both approaches produce a painting. The AI-only version might even look better. But for me, the process matters more than the result.

The journey is the goal

I know this might sound pretentious. In an era obsessed with efficiency, why spend hours on something AI can finish in seconds?

But isn’t the joy of creating in the process itself? In those tiny decisions, unexpected discoveries, and the satisfaction of turning a vague idea into something concrete, step by step.

When I’m adjusting the velocity of a note in my DAW, when I’m agonizing over where a button should go. These seemingly trivial moments are the heart of creation. They can’t be replaced by AI, because they don’t need the “average optimal solution.” They need unique taste, deep understanding of users, and the courage to try something unconventional.

Finding the balance

I’m not saying we should reject AI tools entirely. They genuinely save us tons of time, especially for repetitive work.

Honestly, I’ve come to see code as just a tool. With AI, someone like me who’s always cared more about frontend experience cares even less about things like function naming. As long as it follows best practices, the code is reusable, and it works, I’m happy.

What I actually care about is the user experience. Is that button placement intuitive? Is this animation too flashy? Does the overall flow feel natural? These are what make or break an app.

I need to be a director overseeing the whole project, but I can’t do everything myself either, or nothing will ever ship. It’s about finding a balance between execution speed and how much control I keep over the details.

So I changed my workflow:

Before: “Claude, build me a budgeting app” → wait for result → tweak

Now: Discuss direction with AI thoroughly → write detailed specs → draw my own wireframes → decide design style (colors, fonts, etc.) → decide technical architecture → then ask Claude to implement specific parts one by one.

Most importantly, as a director, when AI produces something beyond what I imagined, I need to reflect: did I not think my vision through, or did my design document fail to communicate it clearly?

This way, AI goes back to being a tool, not the creator. Every important decision is still mine. Execution just got faster.

Slow, but the work is mine

This approach is slower, sure. A project that could’ve taken 10 minutes might now take an afternoon.

But the feeling when it’s done is completely different. Looking at the finished product on my screen, I can say: “I made this.” Every algorithm used has my thinking behind it. Every design decision reflects my taste.

In an age where AI can instantly generate everything, “making it by hand” has become a luxury. But for someone like me who enjoys the journey more than the destination, that luxury is worth having.

After all, if I only cared about results and handed every decision to AI, why did I leave Meta to become an indie developer? Sitting in an office being a cog in the machine pays way better.

But I chose this path because I enjoy the process of creating. So why would I hand the most interesting part to AI?

Alex Hsu

Indie developer, AI music miner, and aspiring writer.
Documenting my journey of personal growth and the pursuit of simplicity.