I’ve been writing code since 1998. Hand-coded HTML on a 56k modem, FTP uploads, Netscape Navigator testing — the whole thing. I’ve watched every wave of “this changes everything” come through: Flash, Ajax, responsive design, mobile-first, WordPress, the SaaS explosion. Each one did change something. None of them changed everything.
Then AI coding tools showed up. And I’ll be honest: this one feels different.
Not because AI writes perfect code — it doesn’t. Not because it replaced the craft — it hasn’t. But because of something that’s hard to explain until you’ve experienced it: the feeling of being able to just try things. To say “what if we did it this way” and have an answer in 30 seconds instead of 30 minutes. That’s what people mean when they say vibe coding.
So What Actually Is Vibe Coding?
The term got popularized by Andrej Karpathy (ex-Tesla, ex-OpenAI) earlier this year. His version is pretty radical — he described mostly ignoring the actual code and just vibing with the AI, accepting whatever it produces, not really reading the output. I respect the man’s brain but that’s not how I use it.
Here’s my working definition: vibe coding is AI-assisted development where you stay in a creative, generative state instead of constantly switching into deep debugging mode. You describe intent, the AI scaffolds it, you review and direct, iterate fast. The vibe — the flow state — lasts longer because you’re not spending 45 minutes tracking down a semicolon in a framework you’ve never touched before.
For me, it usually looks like this: I open Cursor, I describe what I’m trying to build in plain language, and I treat the AI like a very fast junior developer. It writes the first draft. I read every line. I tell it what’s wrong. We iterate. It’s collaborative, not magical.
What Changes (And What Doesn’t)
Here’s what actually changes with AI-assisted development:
- Prototyping speed. I can have a working proof of concept in an afternoon that would have taken me a week to research and scaffold from scratch. This is legitimately transformative for solo operators like me.
- Language and framework flexibility. I’m primarily a PHP/Laravel dev. With AI help, I can write competent Python, Go, or TypeScript without spending two weeks on tutorials first. I still wouldn’t call myself a Python developer, but I can build something useful and not embarrass myself.
- The boring 20%. Every project has a boring 20% — writing unit tests, scaffolding CRUD operations, setting up boilerplate configs. AI handles this well. I can stay focused on the interesting architectural decisions instead.
Here’s what doesn’t change:
- Judgment. The AI doesn’t know your client’s requirements, your server constraints, your existing codebase architecture, or why you made the weird decision you made in 2019. You still have to know what good looks like.
- Debugging real problems. When something breaks in production at 2am, AI is helpful but not magic. You still have to understand your stack.
- Security. AI-generated code has security blind spots. Always review anything that touches user input, authentication, or data storage. Full stop.
Why It’s Different If You Have 28 Years Behind You
I think vibe coding is actually most powerful for experienced developers — and most dangerous for beginners. Here’s why.
When I look at AI-generated code, I can immediately see when it’s wrong. I recognize bad patterns. I know when a database query is going to be a performance disaster at scale. I know when an authentication flow has a hole in it. A developer two years in might not catch those things. They’d ship them.
But because I have that foundation, I can move fast. I can let the AI do the scaffolding, the boilerplate, the research — and I can focus my attention on the 20% of decisions that actually matter. The leverage is enormous.
This year I built a complete SaaS application — ADHDCal, a calendar and task tool designed for ADHD brains — mostly using this approach. What would have taken me six months of traditional development took about six weeks. Not because AI wrote my app for me. Because I stopped wasting time on the stuff AI is genuinely good at.
Is This Going to Kill Developer Jobs?
Maybe some jobs. Probably the ones that were already precarious — offshore boilerplate work, template-based freelance projects with no real problem-solving. If your whole value proposition was “I can write HTML faster than you can think,” that’s a tough spot to be in.
But if your value is understanding a client’s business, making architectural decisions, building trust over years, and delivering something that actually solves a real problem? AI makes that person more productive and more valuable, not less.
I’m not worried about my job. I’m excited about what I can now do with it.
How I Actually Use It
My current toolkit:
- Cursor — AI-augmented code editor. This is my daily driver now. The inline AI chat and autocomplete is genuinely useful once you learn to work with it.
- Claude — For longer architectural discussions, content, and situations where I want to think through a problem out loud before writing code.
- GitHub Copilot — Still useful for quick completions in languages I use less often.
What I don’t do: paste my entire codebase into ChatGPT and ask it to fix everything. That’s not vibe coding — that’s chaos.
What I do: treat AI like a fast, knowledgeable, sometimes-overconfident collaborator. Give it clear context. Review everything. Stay in the driver’s seat.
28 years in, I’m more excited about building things than I’ve been in a decade. That’s the real story here — not that AI is replacing developers, but that it’s making development fun again in ways I didn’t expect.
Want to see what AI-assisted development looks like in a real project? Check out my AI Solutions page or drop me a line — I’m happy to talk shop.