Shipping with AI: From scratch to production

I wanted to answer a specific question: with AI tools doing the heavy lifting, how fast can a real, paid product go from idea to production? Seamless Slideshow was the experiment, an app that crops batches of photos into swipeable carousels for TikTok and Instagram, built in spare time over the course of a week.

The constraint I set myself was to use AI end to end: architecture, code, copy, even the ad creative, and see how much of the plumbing I could hand off while I focused on decisions and wiring things together.

The stack

Claude Code for coding and Claude for Chrome to sync up with Google Ads and Stripe.

Next.js hosted to Vercel drives the frontend upload flow, the interactive cropping UI, and the API routes tying everything together. The image processing itself (cropping, resizing, batch export) runs in a separate Python service hosted on Heroku.

Firebase handles the parts I didn't want to build myself: Authentication for sign-in, and Firestore for storing user and job data, including a token-based credit system: Stripe sells batches of tokens, and Firestore tracks how many each user has left as they export. Analytics tracks how people actually move through the app, and a Google Ads campaign sending real traffic to it, so it's real usage against a real product, not a mockup.

The outcome

Spare-time work spread across evenings makes it hard to point to an exact hour count, but having done it once, I'm confident a similar idea could go from scratch to production in a day or two, provided market research and any other pre-development work has been done.

The initial iteration, within 2 weeks had over 300 people sign up, and made $78 in real payments. Analytics worked well and discovered some bottlenecks in the flow, and the app is still live and continues to be used by people worldwide.

Check it out at www.seamless-slideshow.com