I built an AI Shopify audit tool, then decided to take it back into my own hands

A developer built Vetta, an AI-powered Shopify audit tool that bluntly critiques stores for conversion issues, using Emergent.sh in a single session. Dissatisfied with the interface and token costs, the developer took the code into their own hands, deploying it to Git to gain more control over design and iteration. The project highlights the balance between AI acceleration and hands-on development.

Most Shopify audit tools give you a polite report. Vetta roasts your store instead and tells you exactly how to fix what's broken. A few weeks ago, I built Vetta an AI-powered tool that audits Shopify stores for conversion issues. Instead of a standard “here are some insights” report, it bluntly calls out where your store is bleeding money, then gives you actual solutions to fix it. I built the first version in a single session using Emergent.sh. The core functionality was there, the app was running, and I could get surprisingly far without building everything from scratch. But there was one problem: The interface worked, but I wasn't completely happy with where it was. I wanted more control over the design and the ability to iterate without constantly spending tokens. So instead of continuing to spend tokens on changes I knew I could eventually handle myself, I decided to take a different route. I deployed the code to Git and started working on it directly. That shift has been interesting. Emergent helped me get from idea → working product quickly. Git and my own development workflow give me more control over what happens next which matters a lot for a project where the tone and design carry as much weight as the logic behind it. I'm still learning the codebase and figuring things out as I go, but that's kind of the point of this project. I'm not trying to build the perfect app on the first attempt. I'm trying to understand what I can build, where AI can accelerate the process, and where I still need to get my hands dirty. Vetta started as an AI-built project. Now I'm turning it into a project I actually understand.