My First Vibe Coding Project Taught Me I Didn't Know What I Was Building
I manage mobile engineering teams. I review architecture and push back on scope. I have shipped apps to millions of users.
When I started a side project with AI, I thought the hard part would be choosing shortcuts. I was wrong.
I wanted to build a shopping list app. Most apps give you a flat list. You wander through stores like a maze because the list does not follow a path. My idea was simple: a list that sorts items by aisle. You define your store layout once. The app sorts every list to match that layout.
I described the concept to the AI. It built something fast. It was impressive.
Then I looked at it and realized it was wrong.
The code was fine. The problem was my vision. I had a blurry idea, not a plan.
"A shopping list organized by aisle" is not a complete idea. When a product runs in front of you, hard questions appear:
• What happens when an item has no assigned aisle? • Does a completed item disappear or move to a new section? • Does the completed section live at the bottom of each aisle or at the end of the list? • How does a new user set up their store layout? • Does an empty aisle header stay visible or collapse?
I had not made a single one of these decisions.
When I built things the traditional way, I answered these questions while coding. The friction of writing code slowed me down. It gave my brain time to fill the gaps. The code became the specification.
Vibe coding removes that friction. The AI moves too fast for decisions to hide in the implementation. It makes assumptions about every gap you leave open. It shows you the result before you have thought through the details.
People debate if AI can write good code. That is the wrong question.
The real question is: do you know what you want to build?
You need answers for every decision before you start. You need to know how edge cases work. You need to know how the experience feels.
This is a product management skill. It is the ability to describe an experience exactly before a builder starts. I thought I had this skill. Vibe coding proved I was just hiding behind implementation.
The app I built is called By Aisle. It took a long time to finish. It was not because the code was hard. It was because I had to figure out my own product first.
If you use AI to build, do this: Spend one hour writing down every decision your product requires before you open the AI. You will find many gaps.
Optional learning community: https://t.me/GyaanSetuAi