PROJECT 19 / VENTURES & INDEPENDENT WORK
Find your next movie
A movie discovery app where you name two films you love and it recommends what to watch next, over a browseable catalog.
- Role
- Designer and builder
- Where
- Self-directed
- When
- Ongoing
- Links
- MovieApp
CONTEXT
Recommendation as a black box
Streaming recommendation engines are opaque. They decide what you should watch next and never explain why, so a recommendation arrives without a reason and without trust.
MovieApp inverts the signal. Instead of a long onboarding quiz or an opaque algorithm, it asks for two movies you already love and uses those as the reference point for what to watch next. The input is concrete and low-friction, and the result reads as an answer to a question you actually asked.
WHAT IT DOES
Built as a Next.js app on Vercel, over The Movie Database catalog.
- Name two movies you love
- Get a recommendation for what to watch next
- Browse the catalog directly when the recommendation is not enough
- Sign in to keep a personal list
KEY DECISIONS
The calls that shaped it.
TWO MOVIES AS THE SIGNAL
Tension
The obvious model is either a long taste profile or a pure popularity feed, both of which ask for work or give nothing personal.
Decision
Make the whole input two movies you love, a signal that is concrete, personal and takes seconds to give.
Tradeoff
Two data points is a thin signal. The browseable catalog is the fallback that catches what the recommendation misses.
A BROWSEABLE CATALOG AS THE FALLBACK
Tension
A recommender that misses has to leave the user somewhere useful, not dead-end them on a wrong answer.
Decision
Keep a full browsable catalog beside the recommendation, so discovery never depends on the model being right.
Tradeoff
It is two surfaces to design and maintain. The catalog earns its place every time the recommendation is wrong.
OUTCOME
MovieApp is live and deployed on Vercel. The design lesson is the inversion: asking for two movies you love is a better onboarding than any quiz, because it collects taste in the form the user already thinks in.