A band catches your ear. You work your way through its albums, then go looking for someone else who sounds a little like it. Before long, you’re surrounded by names: Punk, Post-Punk, Alternative, Grunge, Shoegaze.
The names are easy to find. The connections take more untangling. Did one style grow out of another, or borrow from it? Do two bands belong together because of their sound, or because they came out of the same scene and shared the same routes to an audience?
The idea came to me one sunny afternoon, over a cup of coffee, with Pink Floyd’s “Comfortably Numb” playing. I wanted to build a knowledge graph of rock’s family tree.
Rock already has its stars. Why not give them a sky?
That was the starting point for Rock Atlas, a project I built for the pleasure of it.
Start with a style, follow it to an artist, then spend some time with the records and their sources. Or begin with a band you know and see where the connections lead. I wanted to gather the clues scattered across music catalogs and reference pages into a place you could wander through, with something to follow when curiosity takes over.
01. Making the Pieces Connect
The project began as Music Wiki. I wanted to bring together rock styles, artists, works and the evidence linking them.
Finding material was never the problem. Getting it to agree was. The same artist might appear under several spellings. One album could have a long trail of editions. A genre in one database might be divided quite differently in another. A table of names was a start, but every link raised another question: are we looking at the same thing, and what exactly connects it to the next?
So sources and evidence had a place in the data from the beginning. A gap could stay a gap. If two sources disagreed, both accounts had to remain visible. Filling in a guess just to make a path look finished would defeat the point.
The same rule applies to the picture. Two points sitting close together on a canvas do not establish an influence. A line can be beautiful and still have very little to support it.
02. A Small Beginning
The early experiments used a fixed sample: 20 styles, 100 artists, 200 release groups and 9 representative recordings. I put the same material into three kinds of interface: 3D, 2D and a more editorial, reading-led version.
Keeping the content the same mattered. If the 3D view had bands people loved and the list had names they’d never heard of, I could easily mistake familiarity for good design.
The star map took shape gradually. A broad view of styles led into individual artists, then into their works and the records of how they were connected. The catalog grew alongside it, reaching the limited snapshot available today.
Looking back through the development notes, the main stages were:
| Stage | What changed | What it did not establish |
|---|---|---|
| Early sample | Three interfaces used the same material and actions | A test dataset was not a completed study with real readers |
| Catalog growth | More styles, artists and works, with identities and relationships checked across sources | An artist appearing under several styles was still one artist |
| V3 static catalog | Overview, search and details became separate pieces, loaded as needed | The new reading system did not mean the full data import was finished |
| October 9, 2026 | The frontend became its own repository, keeping the star map, deep links and audio controls | Splitting the code did not add to the catalog’s coverage |
| October 12, 2026 | Crawlable descriptions, bilingual guides and text directories were added | Giving search engines a way in did not mean they had indexed the site |
Step by step, an idea became something I could put online and keep working on. Whether it helps people make sense of the music will have to be judged by people using it.
03. A Sky of Styles, Artists and Records
In Rock Atlas, a style becomes a nebula, an artist a star, and a release group an object in orbit. Light bridges carry the connections between them.
The trick is to let the picture open up a little at a time. From a distance, you see styles. Choose one, and its artists come into view. Choose an artist, and you reach the works. Put all of that on the screen at once and the information soon gets lost in the crowd.
The Styles and Artists menus, along with search, offer places to begin. Eras narrows the time period; Relations lets you choose the kind of connection to look at. You can switch languages or reduce the motion. There are lists and a separate text directory for those who prefer reading, and for devices without WebGL.
The map gives you a sense of where things sit. The details and sources tell you what you’re looking at. Select something on the canvas and you should find that same thing in the list and the detail panel.
For a first visit, try a style you know well. Open an artist, look at a work, then check a connection and its source. Familiar ground makes it easier to notice whether the map has shown you something new, or simply dressed up what you already knew.
04. Where the Map Stands Today
As of October 13, 2026, the published snapshot described in the repository and online guide contains:
| Entity | Count | What is being counted |
|---|---|---|
| Music styles | 828 | Styles in the current limited catalog |
| Artists | 420 | Distinct artists, counted once even if they appear under several styles |
| Release groups | 528 | Works organized as ReleaseGroup entities, rather than every edition or individual track |
| Relationship records | 997 | The repository’s current total; these are not all claims of historical influence |
These figures describe the snapshot on the site. They do not amount to a complete map of rock music worldwide. Full imports from MusicBrainz, Discogs and the other sources are still unfinished.
V3 is running, with search and details loaded as needed, background music controls, a bilingual interface, and guides and style directories in English and Chinese. While writing this piece, I checked the live map and Chinese guide. Their style and artist counts matched the repository notes.
The site is open to visitors. The frontend repository is still private, and work continues. The version you see online is V3.
05. Three Parts Behind the Map
The project now has three separate parts: the frontend, offline data processing, and the tools that build and publish the JSON files.
The frontend draws the map and handles queries and reading. The offline work brings in sources, checks identities and puts the records into a common form. The build step turns those records into static files the site can read.
That means a visitor never has to reach a database on my computer. The homepage loads an overview first, then fetches the relevant pieces of detail and search data as you choose a style or artist. The long-term aim is to keep the whole catalog underneath, while showing only as much of it as you need at any one moment.
V3 also checks data signatures and hashes, and keeps each browsing session on the same snapshot. Those checks help ensure that a file belongs to the published version and that old and new pieces do not get mixed together. They cannot tell us whether a claim about music history is true.
The data comes from Wikidata, MusicBrainz and Discogs. Permission to use it depends on the fields and snapshots involved. The code’s license does not automatically cover artwork, audio or every upstream record.
Music comes with a few practical limits of its own. Audio files stay out of Git, and the site has play and pause controls. Your browser may refuse to play sound until you’ve clicked or tapped. Once you pause it yourself, clicking elsewhere should leave it paused.
06. Leave Room for What We Don’t Know
The connections are the part I most want to keep improving.
One style being a subgenre of another is a different claim from having been influenced by it. Two styles sharing tags is different again, as is an association someone thinks worth looking into. They should not all disappear into the same line. Even styles that often share a record cannot be assigned an order of influence from tags alone.
When you open a connection, I want you to be able to see what it says, which way it runs, where it came from and what is still missing. Where we don’t know, the map should say so. Where sources disagree, the disagreement should remain. It takes longer than drawing a line. It needs to.
There are still gaps in the catalog, and limits to what has been reviewed. For now, it can point you towards the next record to hear, the next page to read or the next question to follow. A disputed account of music history still calls for a look at the source.
07. What Comes Next
The next round of work follows the same concerns that have shaped the data and the site.
7.1 A Larger Catalog, with Its Sources Intact
I want to keep working towards the full Rock catalog, adding styles, artists and works without losing their sources, identities, disagreements or licensing records. That still means doing the imports and checking completeness. The present catalog will not become complete just because I rename it.
If the data outgrows the current static host, the plans allow for moving it into separate object storage. There would still be work to do on addresses, cross-origin access, caching and rollback. That option belongs to the next stage; it is not a capability the site already has.
7.2 Easier Browsing, Better Corrections
There is more checking to do on phone gestures, reduced motion, readable text and what happens when data fails to load. When someone spots a problem with a name, date or connection, a correction needs a source we can check. It then needs to reach both the published data and the page.
Can you find what you came for? Can you make sense of the connection? Can you keep going when the information runs out? Those questions matter more to everyday use than another visual effect.
7.3 A Picture That Holds Up as It Grows
With the small dataset in use today, I haven’t noticed problems with visual density or the colors. A full dataset, more than five orders of magnitude larger than this sample, will be another test altogether. Rendering speed, crowding and the ease of telling colors apart still have to prove themselves at that scale.
08. Bring a Band You Know
If you like rock music, start with a style or artist you know and follow the connections for a while.
Try Rock Atlas · Guide and Data Notes
The code is still private, but you can reach me through the contact details here. If you’ve found an error, send the name, the page and a source I can check. If something is awkward to use, tell me where you began and where you got stuck.
And if you’re a developer who’d like to help work on it, get in touch through GitHub.