Last night I built an app from my phone, while watching a movie. By the time the credits rolled it was online, and I was already using it.
The movie wasn't great. Hence the side project.
The itch
Picking what to watch is something I'd wanted to fix for years. I'd built lists on IMDb and TMDB, tried a handful of apps. Nothing ever did exactly what I wanted.
So, mid-movie, I started building the thing I actually wanted: a private movie tracker with my own taste baked in. I call it Movee.
Building it from the couch
The trick is that the project lives on my computer at home. I just drive it remotely from my phone.
I opened Claude on my phone, described what I wanted in plain English, and the agent did the rest: wrote the Python, designed the database, generated the pages, and pushed it live. Every change I asked for rebuilt and redeployed the site automatically, so the version on my phone was always current.
Not a single line of code written by me. I described behaviour; the agent wrote and shipped the implementation.
What it does
By the end of the film I had a working app:
- 2,000+ movies preloaded, with all the details (director, cast, runtime, genres, posters, age ratings).
- Filters and watched-lists to track what I've seen and rate it 1-10.
- AI recommendations, tuned for three contexts: me solo, me and my son, or a family Friday night.
The recommendations are the part I care about most. They lean on my own ratings and favourite directors, and stay in a separate "to watch" list per context, so the family picks never get mixed up with mine.
Under the hood
I asked for a stack that is boring and durable on purpose:
- SQLite as the single source of truth, one row per film. No server database to babysit.
- TMDB for metadata, so I never type a title or hunt for a poster.
- A small Python build step that renders the database into a static site (just HTML, CSS and a little JavaScript). Static means fast, cheap, and nothing to break.
- Deployed to a private, password-protected site, with the build and upload triggered automatically on every change.
One detail I like: the AI recommendations run inside the coding session itself, on my existing subscription. No separate AI service to wire up, no per-request bill, no extra key to manage.
Why this matters beyond movies
This was a personal itch. But the capability behind it is the real story.
With AI you can now build the documents, dashboards and custom internal tools that used to take days or weeks, or never got built at all. The bottleneck was never the idea. It was the time and the hands to make it real. That part barely takes effort now.
I've already used Movee more than every other solution I'd tried, combined. It took one evening and zero code from me.
Still amazed, every single day. 🚀