I built this because of a book called "1913" by Florian Illies — a month-by-month portrait of the last year before the world changed. It covers Louis Armstrong picking up a trumpet, Coco Chanel opening her first shop, Kafka's love letters, Stravinsky writing The Rite of Spring. When I finished it I didn't want one next book — I wanted a dozen. A cookbook from that world. A Russian political history. A French novel set in exactly that moment. Every search tool I tried gave me one lane: pre-war history, then WWI. None of them could do what my brain was doing, which was moving sideways across the same moment in time.

So I built OpenedShelf. It uses a three-tier Boolean taxonomy — cross-genre thematic tags, genre identities, and community-specific tropes — to let readers stack objective, verifiable facts about a book rather than vibes or bestseller rankings. The engine is intentionally strict: it shows you exactly where your search runs out of results. That zero-result state isn't a failure. It's a discovery — proof that the book you're imagining either hasn't been tagged yet, or hasn't been written yet. The URL is shareable. The gap is the point.

On the technical side: the app runs on Cloudflare Workers + D1 (SQLite at the edge), server-renders plain HTML with no client-side JS frameworks, and collects zero behavioral data. The 17GB OpenLibrary dump is processed offline through a 7-stage pipeline and uploaded to D1. The taxonomy is licensed CC0. The engine is AGPLv3. The whole thing is designed to be self-hostable on a weekend with Python, SQLite, and Bash.

The taxonomy is governed by a community Library Council — genre-literate readers who review tag submissions, vote weekly on promotions, and publish every veto with a written explanation. The goal is objective facts about texts, not algorithmic popularity.