I Shipped a Booking App Without Writing Code

A friend runs a small yoga studio. She was paying $80/month for a booking system that she hated. “Can you build something better?” she asked.

I’d never built a booking system before. But I said yes anyway.

Two weeks later, she had a custom booking app. Her students use it daily. She canceled the old service. And I learned a lot about building real software with AI.

What she needed

The requirements were simple enough on paper:

Display upcoming classes on a calendar
Let students book spots
Send email confirmations
Let her manage classes from a simple admin panel
Limit class sizes to avoid overbooking

Nothing groundbreaking. Standard features that exist in a thousand apps. But that’s actually what made it a good test — these are solved problems with known patterns. AI should handle them well.

The build process

I didn’t try to build everything at once. Started with the absolute minimum: show classes on a page. Then: let someone click “book.” Then: record that booking somewhere.

Each step built on the previous one. The first version of each feature usually needed tweaking. Sometimes the styling was off. Sometimes the logic had edge cases I hadn’t anticipated. But the foundation was always solid enough to iterate on.

Biggest challenge: email integration. Getting confirmation emails working required understanding infrastructure stuff I wasn’t familiar with. Transactional email is its own domain — deliverability, authentication, formatting. Took some learning.

What took longer than expected

The admin panel. It sounds simple, but the UX decisions piled up. How should she add a new class? What if she needs to cancel one? What about recurring classes?

Each question spawned more features. And each feature had edge cases. What happens when you cancel a class that has bookings? How do you notify everyone? What if someone already showed up?

This is where the real work happens. Not in the code generation — in the decisions about what should exist and how it should behave.

The result

A deployed web app that runs on $0/month hosting. Students can view classes, book spots, get confirmations. She can manage everything from her phone. It’s not fancy, but it works.

The client is happy. More importantly, her students are happy — several mentioned the new system is easier than the old one.

Total time invested: maybe 20 hours across two weeks. Not full-time work — a few hours here and there between other things.

“This would have taken me months if I had to code everything myself.”

What I’d do differently

Start with the admin experience. I focused on the student-facing side first, which meant the admin panel felt rushed. Should have flipped that — she uses the admin daily, so that UX mattered more.

Also: more planning upfront. I added features as we thought of them, which worked but wasn’t efficient. A simple spec document would have reduced the backtracking.

Could you do this?

If you can clearly describe what you want built — yes. The AI handles the implementation. You handle the decisions.

The hardest part wasn’t the code. It was knowing what questions to ask and recognizing when something was done well enough to ship.