Skip to content
Evan Luza
All workNo. 10 / 15

StreamGenie

What are we watching tonight?

Role
Founder · Product, design & build
Year
2026
Status
Live
Focus
Consumer product, Recommendation UX, Search & SEO, Front-end
Three StreamGenie screens on phones: the home screen asking “What are we watching tonight?”, results for “something cozy and funny for a rainy night” filtered to streamable titles, and a where-to-watch page listing subscription and rent options.
Ask, filter to what you can stream, see where to watch.

In short

A streaming discovery app that starts from a mood instead of a catalogue. Type what you’re in the mood for, pick the services you already pay for, and get titles you can actually watch tonight — with exactly where to stream, rent, or buy each one.

At a glance

Context
Everyone has three or four streaming services and still spends twenty minutes choosing something to watch.
Problem
Catalogues are organised by genre and service, but people decide by mood — and then find out the title isn’t on anything they pay for.
My role
Solo: product, interface design, the recommendation engine, and front-end.
Outcome
Live at streamgenie.net, with a where-to-watch page for every title so people can find it from search.

Approach

  1. 01

    Start from a mood.

    The first thing you see is a question — “What are we watching tonight?” — and a prompt bar that takes plain language like “a comfort watch for a rainy day”. Genre chips and a “Surprise me” button cover the nights you don’t want to type.

  2. 02

    Only show what you can watch.

    Pick the services you have, and results are filtered to titles actually streaming on them. Each result says why it matched and where it’s playing.

  3. 03

    Every title answers “where?”

    Each title has its own page listing what’s included with a subscription and what you can rent or buy — and those pages are what people find when they search “where to watch …”.

Execution

Plain language in, a short list out.

Built with Next.js on top of TMDB’s catalogue and streaming-availability data. A deterministic parser turns a prompt into genres, moods, and runtime constraints, with an optional language-model layer for prompts it can’t read on its own. Results come with ratings, trailers, a watchlist, and share links.

Next project

BestPic

Helping people stop overthinking which photo to use.