A cartography of music before it breaks

Every artist is a constellation

Picture an artist not as a row in a database, but as a constellation — a scatter of individual sightings, each one small, none of them proof by itself. A DJ plays them at a club in Berlin. A record store in Tokyo puts their vinyl on the shelf. A radio host in Seattle spins the track between two others. Someone points their phone at a speaker on a night bus and never thinks about it again. A festival lineup lists their name in small type, fourth from the bottom.

None of those sightings, alone, means much. Hosaka's job is to notice when the same artist keeps showing up across genuinely different, independent kinds of attention — the DJ set and the record-store shelf and the radio spin all pointing at the same constellation — and to say: this one is starting to cohere.

The idea underneath that is worth stating plainly, because it is the whole trick in one sentence: hosaka does not need anyone's permission slip to do this. It does not wait for an artist to have an official entry in some industry database before it starts tracking them.

The identity primitive, in plain terms

Every artist, label, and song gets its own internal marker from the first time hosaka spots them — internally, hosaka calls it a cluster_id, but you can just think of it as the artist's shape, made of sightings instead of paperwork. If that artist later turns up in an official music catalogue, that catalogue entry gets attached as a note on the shape that already existed — it never replaces it. The sightings were always the real identity.

This matters more than it might sound like it should, because the artists without an official catalogue entry yet are disproportionately the newest ones. The unresolved cases — the ones no clean database has caught up to — are where the freshest talent tends to live. A system that only trusts an artist once they are catalogued is a system that is structurally always a little late. Hosaka is built not to be late.

You can read more about how the identity spine works — including why cluster_id is nullable in API responses for unresolved artists — on the method page.

The scale, made tangible

None of this works as a handful of anecdotes. It works because of scope. Right now, hosaka is tracking roughly 2.2 million artists, built out of sightings across 16 different kinds of places music surfaces — DJ sets, radio airplay, music press, record-store shelves, festival bookings, club bookings, reviews, podcasts, phone-recognition moments, reposts, official release catalogues, and more.

The live-scene side of the map alone spans 422 areas around the world (counted by RA area id), pulling from 79 live radio stations — not just the handful of markets that usually get to decide what counts as "happening."

That is the map: not one feed, not one city, not one kind of attention, but 16 different lenses on the same underlying reality, cross-checked against each other, everywhere at once. The detection-time ledger — how hosaka knows, and since when — holds 3.16M rows as of 2026-08-13. Developers can query it directly via crate.

Ask it yourself

You do not have to take any of this on faith. You can ask it directly.

The explorer queries the same records everything else here is built on — the dossier for an artist, the independent sightings behind it, and when each one was observed. No key, no account, no sign-up.

It is the plainest possible answer to "what does hosaka actually do": name an artist you care about, and read what has actually been seen of them, with the dates attached.

Public traces, honest measurement, and killing ideas that do not hold up

A project like this earns trust the way any careful observer does: by being scrupulous about what it looks at, and honest about what it finds.

Hosaka only works from public traces — things an artist, a store, a station, or a crowd already put into the world for anyone to see. Nothing private, nothing taken from behind someone else's login, nothing gathered by convincing someone to hand over data they did not mean to share.

It also holds itself to the standard a good scientist would: measure first, believe second. When someone on the team has a hypothesis — this kind of signal should predict a break, this pattern should matter — it gets tested against real outcomes before it gets trusted. Plenty of ideas that sound right on paper do not hold up. When that happens, they get killed, not quietly buried: the team writes down what did not work and why, so the map does not accumulate wishful thinking alongside real signal.

A map that only ever adds and never subtracts stops being a map of the world and starts being a map of what its builders wanted to be true. That discipline is slower than shipping every plausible-sounding feature. It is also the only way a map like this stays worth trusting. The evidence behind the methodology is on the method page.

Where this goes

Music has always broken somewhere before it broke everywhere. A city has always known first. A shop, a station, a DJ's crate has always known first. That was true before the internet, and it is still true now — it is just harder to see, scattered across a thousand small, independent moments that never talk to each other.

Hosaka does not create that pattern. It just draws the map of it — constellation by constellation, city by city, sighting by sighting — so that "somewhere, first" stops being a rumor you hear about after the fact, and starts being something you can actually go look at.

The substrate is built and running. The map is live. If you are a developer, label, or record store, the relevant door is below. If you want to see it for yourself, the explorer needs no key: crate.hosaka.fm/explore.