Okay, here is my take on “vibe coding.” I’m building a small pet project that has been in my mind for the past year, but I’ve been procrastinating for a long time. Now, I’ve made a firm decision: I will not write even a single line of code myself. Everything must be generated by AI — that is the boundary I’ve set for this project.
What do I expect as the outcome? A nice-looking UI and working functionality are not enough. I also expect industry-standard code quality, complete with proper unit tests, end-to-end test coverage, and thorough documentation.
For this, I’ve decided to use Google Antigravity and ChatGPT integrated with VS Code.
My Initial Experience
As of now, I’m focusing only on the APIs, and I’ve almost completed all of them. Through this entire experiment, I realized that vibe coding has both Pros and Cons.
The Positive Side — The Nitrous Effect
Vibe coding gave me a feeling I never expected. Back in my school and college days, I was a huge gaming freak. When I played racing games like Need for Speed or Paradise City, the car would already be speeding down the road. But still, I’d wait for the nitrous bar to fill up — and the moment it did, pressing the Shift key would send the car roaring forward:
Wooooo WHROOOOMMMM!
During vibe coding, I felt that same rush. The moment the AI-generated code started falling into place, it felt like hitting nitrous — a sudden acceleration that pushed the project forward at incredible speed.
That “nitrous feel” is one of the best parts of vibe coding.
The Negative Side — Wings Required
But vibe coding also has its downside. It will extend your wings only if you already have them — it does not create new wings for you.
It works best for experienced developers who understand:
how things work under the hood
how to debug when something breaks
how to design a deployment strategy
how to judge and structure high-quality code
We cannot expect a fresher with no real-world experience to produce industry-standard code purely using vibe coding. Without foundational knowledge, they won’t know whether the AI-generated code is correct, secure, scalable, or even functional.
Vibe coding amplifies skill, but it does not replace it.
My Experiment to Prove This
To prove this point, I ran a small experiment.
I built the same API in two different programming languages:
Node.js — a language I know inside and out
.NET — a language in which I have never written even a single line of code
Same task. Same expected output. Only the language changed.
In Node.js, I completed the API with vibe coding in 30 minutes, even with complex logic.
In .NET, it took me more than 2 hours, and even then, I was not fully confident about its performance.
This made something very clear:
Vibe coding accelerates you only in domains where you already have expertise. If you don’t have the wings, the nitrous won’t help.
Why This Matters in the Software Industry
The main reason I conducted this experiment is because, at Grids and Guides, almost every client meeting eventually leads to the same question:
“Hey Vasanth, can’t we reduce the development cost? AI and vibe coding are popular nowadays, right? Buhhaaaa…”
No matter what the meeting is about, someone brings this up.
But if you understand the experiment above, the answer is obvious:
AI-generated code does NOT magically reduce the cost of development.
Because the developer behind the AI is still the same developer. Their time, experience, and judgment still matter.
AI is just an additional tool — nothing more.
The Macro-Economics Question
This brings us to a bigger question: Can we still estimate development cost using the Time & Material model?
Let’s take a simple example:
A developer with 10 years of experience takes 8 hours to complete a job and charges $100.
With AI, the same developer finishes the job in 4 hours.
So should the developer charge $50 or still $100?
Some may argue:
“Hey, the time reduced, so the cost must reduce.”
The developer hasn’t changed. Only the tool has changed.
Just like a carpenter doesn’t charge less because he upgraded from a manual screwdriver to a power drill, a developer shouldn’t charge less because they upgraded to AI-assisted development.
The value lies in the expertise, not in the number of minutes.
Conclusion
Vibe coding is powerful, exciting, and transformative — but only when used by someone who already knows how to build real software. AI doesn’t replace developers; it simply gives them a nitrous boost.
So if clients still insist, “AI will make development free, right?”
My final answer is ready:
“Sure! I’ll let the AI do the work — as long as it also attends your meetings, handles your change requests, and explains your requirements to itself!”
Until then… vibe coding or not, developers are still very much needed. 😄