Popcorn
Recommendation engine
A taste engine that ends the what-should-we-watch scroll
- Role
- Design / Build
- Stack
- Next.js · Supabase · pgvector
- Year
- 2026
- Status
- Private beta — nightly use
Streaming solved access. Popcorn solves choosing.
Streaming recommendations run on what you half-watched, and only inside their own service. Popcorn learns your real taste from pairwise picks — this or that, ten seconds each — not a prompt wrapped around a catalog. Then one confident call across everything you can watch, tuned to runtime, mood, and who’s on the couch. Together mode blends two people’s taste into a pick neither has to apologize for.
Refine — this or that, ten seconds each; the model learns your taste
Tonight — one confident call, tuned to runtime, mood, and the room
Together — two people’s taste blended into one pick
Taste compare — where you two align, and where you collide
Profile — how well it knows you, your all-time top shelf, and your watch stats