Unread: the truly personalized book discovery

Year

2026

Industry

Books, Startup, ML

Client

Unread

My role

Co-Founder | Lead Product Designer | FE

Why we built this?

Imagine walking into a store filled with 100k unlabeled mystery boxes and having to pick just one. To get your money's worth, you are committing to read whatever is inside for the next week or two. This is a dilemma readers face when searching for their next book.

Standard algorithms push the exact same sponsored bestsellers to everyone, but Unread surfaces hidden gems that align with an individual's unique taste profile.

Business Strategy

UX/UI

AI Coding

Web App

Database

Currated Feed

How does it work?

01

Not an LLM

Proprietary two-stage machine learning model: Two-Towers retriever + FiBiNET ranker.

02

Better than LLMs

Benchmarked as 12 times better at predicting your next favorite book than Gemini 3.1.

03

Semantic fit

Unread takes into account more than 20 types of levers to find the best personalized fit.

Currently in Beta

Mastodon