Case study
A beautiful, keyboard-navigable web app for exploring the entire Tree of Life — every branch of living things on Earth, from “Life” all the way down to a single species.
Overview
Jahan — Persian for “world” — is a way to wander the tree of life instead of studying a chart of it. You begin at the root, “Life”, and move down through kingdoms, phyla, classes, orders, families and genera until you reach a single species. Every node is a card: a photo, a common name, a scientific name.
The whole thing is navigable from the keyboard, works in dozens of languages, and is completely free. It's live at jahan.life.
The ideas it's built on
Existing tools (OneZoom, Lifemap) lean on zoom-and-pan, which gets disorienting fast. Jahan takes the opposite bet, inspired by neal.fun's “Size of Life”: linear, predictable movement. Up goes to the parent, down into the children, left and right between siblings — the whole tree navigable from the keyboard.
The phylogeny view frames a species by its relatives — ancestors, siblings, and notable cousins. Tiger and lion are siblings; tiger and cheetah are cousins. It reads as a family tree rather than an abstract diagram.
A clade can have hundreds of thousands of children. Rather than dump them all, Jahan scores each by iNaturalist observations, whether it has an image and a common name, and species rank — so the branches you see first are the ones worth seeing.
Real-world taxonomy is messy: hundreds of thousands of species hang directly off high-level ranks where they don't belong. The API filters and sorts by rank priority to keep the descent through the tree coherent.
Where the data comes from
There's no single database of all life, so Jahan builds one — merging several open sources into one clean, deeply-linked tree of roughly 4.5 million nodes and 47 languages.
How it's built
The heavy, multi-source merge and enrichment runs offline in PostgreSQL. The result is flattened into a query-optimised SQLite table and shipped to Cloudflare D1 at the edge — so serving a node is a single fast query, not a pile of joins.
New data is imported into a standby D1 database and then swapped live, giving zero-downtime releases of a dataset that spans millions of nodes.
The navigator is a WebGL canvas (Three.js via Threlte) with a tweened camera panning between levels and custom raycasting for interaction — served from Cloudflare with Smart Placement so the Worker sits next to its database.
Scientific names are hard to say, so Jahan speaks them via Cloudflare Workers AI and Deepgram's Aura-2 voices, with common names and descriptions in dozens of languages and full RTL support.
Stack