Developers Using AI: Beyond the “Vibe Coder” Slur

The Loaded Term Nobody Asked For

“Vibe coder” is becoming a slur. And like most slurs that stick, there’s real frustration behind it—but also a lot of nuance getting flattened in the process.

I’ve been using AI since GPT 3 dropped. I’ve watched the landscape shift from “cool toy” to “industry disruptor” to whatever chaotic phase we’re in now. My position on all this? It’s somewhere between where you’d expect a traditionalist to stand and where the AI evangelists think everyone should be.

What “Vibe Coding” Actually Means

Let’s be specific, because the term gets thrown around loosely.

Vibe coding (the problematic kind): You never look at the generated code. You just keep prompting the AI to fix problems, layer after layer, until something works—or more likely, until you have an unmaintainable pile of technical debt that kind of runs.

Should developers do this? No. Full stop.

But here’s where it gets interesting.

The Right Way to Use AI (If You Actually Know How to Code)

If you’re a developer with real coding skills—not just “I did a bootcamp” but actual problem-solving experience—AI becomes something else entirely. A multiplier, not a replacement.

Here’s the workflow that actually works:

1. Planning together — You and the AI architect the solution before a single line gets written
2. Component by component — You build piece by piece, reviewing and understanding each part
3. Approval-based iteration — You check what the AI generates, you don’t just blindly accept it
4. Graduated trust — Over time, you learn what your AI tool is good at and where it hallucinates. You trust it more, but never completely

The goal isn’t to remove yourself from the process. It’s to move faster while staying in control.

The Model Wars: Claude vs GPT vs The Rest

Everyone says Claude is best for coding. That’s only half true.

Claude has this tendency to anticipate what you’ll need next. Ask for a todo app, and it’ll build the todo app plus the user auth plus the database schema plus the deployment config. Great if you’re vibe coding. Annoying if you wanted to build those pieces yourself, in your own way, with your own architecture decisions.

GPT follows instructions to the letter. “Build me a todo app” gets you… a todo app. Nothing more, nothing less. For developers who know what they want and want precise control, this is often better.

Open source models are all chasing Claude’s autonomy. They’re getting better at “doing the thing you didn’t ask for but might need.” Whether that’s good depends entirely on your workflow.

Gemini? The web chat interface is genuinely excellent for research and general knowledge questions. I don’t use it for coding.

Grok? King of general knowledge. Nothing else comes close for staying current with real-world information.

The Real Divide

The vibe coder debate isn’t really about AI usage. It’s about responsibility.

Bad developers using AI become dangerous fast. They ship broken things they don’t understand. They create security holes. They build houses of cards that collapse in production.

Good developers using AI ship faster, explore more ideas, and still own the result. The code is theirs. They could write it without AI—they just chose not to.

The slur emerged because the first group got loud and annoying. But let’s not throw out the tool because some people misuse it.

Where I Land

Use AI. Please use it. But stay in the driver’s seat.

The developers who thrive in this new landscape won’t be the ones who can prompt best. They’ll be the ones who could write the code themselves, choose not to, and still know exactly what’s happening under the hood.

That’s not vibe coding. That’s just coding with better tools.

---

Written by someone who’s been prompting since GPT 3. After this long, I know what to expect.