If you’ve been paying attention to the AI hype machine lately, you’ve probably heard terms like “vibe coding” and “agentic engineering” thrown around. Maybe you’ve tried them yourself. Maybe you took someone else’s code and “fixed it”. Maybe you’ve watched someone on YouTube build an entire SaaS app in 20 minutes and thought, “Well, I guess I’m obsolete now.”
You’re not. But the way we write code is changing, and it’s worth understanding what’s actually happening versus what the influencers and Reddit want you to believe.
Here’s how I think about it: imagine you’re building a house.
The Architect and the Builder
Traditional software engineering is like being both the architect and the builder. You draft the blueprints, you understand the load-bearing walls, you know why the plumbing goes where it goes. It’s slow, but when something breaks at 2 AM, you know exactly where to look.
Agentic engineering is like being an architect working with a skilled builder. You provide detailed blueprints, specifications, and context. The builder (the AI) does the actual construction. You review the work, catch issues, and course-correct. The builder might even suggest improvements you hadn’t considered. It’s collaborative. You’re still in charge, but you’re leveraging someone else’s hands.
Vibe coding is like being a 10-year-old with crayons trying to direct a construction crew. “Make it blue! No, bluer! Add a slide! Why isn’t there a pool?” You’re technically giving instructions, but you don’t understand what you’re asking for, you can’t evaluate if the result is actually sound, and when the roof caves in six months later, you’ll have no idea why.
I’m not trying to be insulting here. That’s just the reality of how tools work. You can hand anyone a hammer, but the results depend entirely on what they bring to it.
The Friend Who Learned the Hard Way
A friend of mine spent weeks building a web app with AI. He started with an HTML template, fed it to an AI, and iterated until he had something that mostly worked. I was genuinely impressed at first glance. It looked good, it functioned, he was excited.
Then he wanted to change a color.
He didn’t have CSS experience, so he started doing find-and-replaces in local Bootstrap files. When that broke things, he asked the AI to “clean it up.” The AI made changes he didn’t understand. He asked it to fix those. Repeat.
By the time I got my hands on it, I couldn’t make heads or tails of what was going on. No MVC pattern, no consistent structure, single files with five JavaScript functions that all did the exact same thing. The whole project had to be rolled back.
Watching him work with the AI was painful. Not because he was doing anything wrong, but because I could see the pattern… he was giving instructions he couldn’t evaluate, accepting output he couldn’t verify, and making changes he couldn’t trace.
Garbage in, garbage out. Again, that’s not meant to be an insult. It’s just how it works.
This Isn’t an Anti-AI Post
I’ve been shipping web applications and SaaS products for years. These days, AI is a core part of my workflow and I wouldn’t go back. I recently used it to refactor old projects I built years ago, and it found architectural issues I’d completely overlooked. It catches things I miss. It’s like having a code reviewer on call 24/7.
The difference is that I understand what I’m building, so I can evaluate what the AI gives me. That’s the whole game.
Vibe coding is fine for quick prototypes, learning experiments, and low-stakes projects. The danger is when people vibe code their way into production systems they can’t maintain or debug.
If you’re building something that matters, be the architect. Understand your blueprints. Review the work. Know why the walls are where they are.
AI can take you places fast. Just make sure you’re driving, not riding in the trunk hoping for the best.
Good post!