Vibe coding and the illusion of superpowers

Takeaways

• Vibe coding replaces building skills with negotiating with AI.

• Leaning on the assistant means exercising your own abilities less.

• Smarter models don't make you better — the hours do.

• Tools augment skill; they can't replace it.

I was really curious about all the hype around “vibe coding”. I’ve spent my decades-long career working in software and exploring new possibilities with this amazing tech. Okay, vibe coding, challenge accepted.

I spent months pushing it as hard as I could on genuinely complex projects. Some of it was impressive. Other aspects? Not so much.

My biggest takeaway from the experience is that what’s required to build software at scale is directly—and often negatively—impacted by AI assistance.

Instead of AI expanding my own skills, I spent tons of time negotiating with the model’s “intelligence,” wrangling with rogue decisions, and digging out from issues it had created. It wasn’t a seamless experience.

Vibe coding is not (yet) improving the craft of software . . .

As a software developer-turned UX professional, I thought vibe coding would be a high-return investment of my time and energy.

Fast idea generation plus less manual coding—surely that would give me superpowers as both a designer and a developer.

I felt like a kid in a candy store.

The reality? The experience wasn’t easy. Or magical.

I found myself prompting more than coding or designing. I was operating at a different level of abstraction and routinely wasn’t getting what I wanted from the model.

Worse, it wasn’t clear how to become better at vibe coding, because the more I leaned on the assistant, the less I was exercising my own abilities. And by the way, prompt and/or context engineering do not make you better either. They’re workarounds for a flawed experience.

I’ve spent my career building skills through genuine commitment to the craft: I take classes. I practice. I can’t only lean on tools or technology to make me a better creator.

When the industry announces something that promises to break new creative ground, I’m first in line to get it, learn it, and master it. Sometimes I’m a sucker for the pitch that a new tool actually will deliver on its promises.

Yes, I admit it. I bought VisionPro.

Vibe coding might be making it worse.

I kept seeing: “Pay for a ‘more intelligent’ model, and you’ll get ‘better results.’”

Smarter models don’t guarantee better results. All models hallucinate. And paying for a model doesn’t improve your own ability — it just makes things faster and often more frustrating.

The real question: what skill do you actually gain when you pay for a “smarter” model?

Paying for models is a bait-and-switch. Skills (not tools) are what make you better at a craft.

Growing as a creator isn’t about buying better tools to build things “faster.” It’s about developing the skill to use any tool effectively.

You wouldn’t buy an expensive pencil to become a better artist. The pencil doesn’t make you better. The hours do.

Skills developed with great tools produce great outcomes. That combination—not the tool alone—is what actually augments human capability.

Vibe coding is not quite up to its hype. I still have hope, though. I bought the VisionPro, after all, and I still own it.

---------------------

Jonathan Gordon is the founder/CEO of ReWeaver AI. He has worked as a user-focused software designer leading design and engineering teams at Google, Microsoft, Oracle, Facebook, SAP, and others. ​​​​​​​​​​​​​​​​