← Index Case study / 01 2026

TV-first recommendation app · Amazon Vega / Fire TV

YourNextBinge

A D-pad-only Fire TV recommender that turns 83 titles and seven moods into five watchable picks, with a reason for every one.

Year
2026
Role
Interface, recommender, test suite
Context
Team project
Stack
React 19.2 / React Native 0.83 / Amazon Kepler 4.0 / TypeScript 5.8
83 titles in the catalog
7 moods
4 screens, D-pad only
0 network calls

A TV remote is a bad search tool: endless grids, no keyboard, and too much scrolling for a simple question.

Built a 960×540, D-pad-only app that turns 83 titles and 7 moods into a five-item shortlist, with a one-line reason for every pick. The catalog, scoring, and explanations run locally.

  1. Designed a deterministic recommender with weighted scoring and a fixed relaxation ladder; the UI never shows an empty list.
  2. Wired predictable first focus and explicit focus maps for row edges.
  3. Covered scoring, relaxation order, navigation, and rendering with a Jest suite.

83 titles, 7 moods, 0 network calls. A complete TV flow that answers the question in one screen.

Next / 02

Rogue Protocol

A fault-versus-noise classifier for ship logs, tuned to 0.6832 validation F1 and shipped through a checker-gated submission path.