I talked to PJ Wehry on the Chasing Leviathan podcast, and now you can hear us: www.youtube.com/watch?v=iCTd...

The genre system we created at the Echo Nest and expanded for a subsequent decade at Spotify had, by its unnecessary demise, over 6500 genres, every one of them picked and guided by humans. The automated ML system that replaced it has 708 tags.

Sometimes I come across a very long playlist and I want a quick sense of how my own listening intersects with it. There are lots of ways to quantify this, but here's one of them in Curio query form (you can copy and paste the query from the alt-text):

If you're curious about how different streaming-royalty allocation schemes would affect you and your artists, or could be tempted into curiosity if you thought there was a way to see this, there's a way to see this: furia.com/log/517

People complain about Spotify's shuffle mode, and I never listen on shuffle so I don't care, but if you do, you can make your own in Curio: tracks....shuffle/artist;.1:@<=100 Run that on the query page, and then save the results as a playlist.

Do not let the world's deluge of grievous wrongs numb you to your own tiny useless delightful errors.

If you want to scrutinize algorithmic playlists Spotify makes for you, you can copy their tracks into normal playlists and then use Curio to run queries on those. E.g.: playlists:~<[DW 2025-06-27].other playlists .tracks.(.artists:@1).other artists /label=(.artist catalogs.catalog.label)#count

Here, for example, is ?? in a Curio query that gets a playlist and some albums, annotates the track, album and artist overlaps between them (in that order of priority), and flags the tracks in each album that were previously released (by that artist) before their album appearances.

DACTAL changes that might affect you if you're doing Curio queries: filter negation was :! now :- reverse index-number filtering was :@- now :@@ repeat was ?? now !

Today's tiny DACTAL tweak that probably affects nobody: in an aggregation, semicolons explicitly identify aggregators. So this query is a count of counts; the first "count" is a property (of groups), the second "count" is the aggregator. tracks/artist...count;count

Made some improvements to playlist comparisons. (Also, did you know Curio does playlist comparisons?)

Here's what you get when your software is built by people who don't live in it. This is Spotify's useful display of what you've recently played, by playlist. But if you tap a song here, it doesn't resume that playlist, it starts playing through THIS view, which is your play-history in reverse.

The Spotify iOS app has adopted the annoying habit of starting my playlists over again, so in defense recent Listmarks in Curio can now be expanded to show your recent play history within them, so you can jump back to where you really were.

Happy NMF. Some notes about a geeky Curio feature. www.furia.com/log/507 NMFinder 2025-04-18 open.spotify.com/playlist/1j4... the new Flight of Icarus song is out (and some others) open.spotify.com/playlist/2hz...

"Best practices" are incurious. Practice is most interesting when you don't predetermine what it is allowed to reveal. You can now annotate playlists in Curio. And do what else with those annotations? Nothing, yet.

Hooked up the DACTAL export function. Particularly useful in Curio for copying Listmarks' recent_history to other browsers.

I wanted some Curio track-flagging features that were too specific to justify adding as general features, but that in itself justified adding a custom-flagging feature.

Curio is mostly intended for desktop use, but I've now made some very basic style tweaks for rudimentary phone-screen support. The only thing specifically tuned FOR phone use is the Listmarks page.

If you create a playlist from a query in Curio, like NMFinder for example, you can now rerun the query to update it from the playlist's Curio page.

Album-cover scan-previewing in Curio is now on Option-hover instead of Command-hover, to eliminate jump-scares if you cycle into a tab or window with your pointer in the wrong place.

If the amount curious you are about my book is $4.99, the ebook version has dropped to your bargaining point. a.co/d/4lUJ1zp

I added a "lookup" connector to Curio so that queries can navigate into any api.spotify.com links the Spotify API returns. E.g. playlists:@1.tracks.href.lookup.results.items:@1.added_by.href.lookup

I spent almost 5 minutes searching the web for liner-notes on this compilation to explain which tracks came from where, and then I remembered that I have a tool for that.

Can we collate our collective knowledge ourselves? Not as much as should be our right, but sometimes, a little bit. open.spotify.com/playlist/2BV...

The first-letter song links in NRbG were originally a placeholder idea, but I became instantly fond of them, and I'm always pleased when they reveal some pattern that is less obvious in full context, like the way this album was maybe not entirely sequenced for artistic effect:

Finding songs by matching playlist-title words is a bad method, but bad methods should at least be easy. In Curio: power ballads.matching playlists.matches /track=tracks:count>=10 #rank=count,(.track.popularity) /artist=(.track.artists:@1).(.of:@1) #rank .track open.spotify.com/playlist/1sH...

In today's big Curio news, I fixed a bug in a feature that I've never explained, and which might be too specific to my own habits to be useful even if I did. But I use it once a week, and I broke it between last week and this week, and now it's fixed again.

My favorite kind of feature development: Last night I added Play buttons to the related-artist page in Curio. This morning I woke up realizing that they would be useful in all the other artist pages, too. And when I went to add them to the others, I discovered that I already had.

I added Play buttons to the related-artist view.

Somebody needs to build a system for collective ongoing maintenance of genre definitions, and probably that somebody will be me, eventually. But in the meantime, as a small token of possibility in a time of loss, I added three missing genres because I needed them. everynoise.com/nrbg.html?ge...

Spotify seems to be quietly removing genres from the API now. I'm going to wait and see where this goes before I change the everynoise.com/research.cgi connections. NRbG is not affected.

It's more Unraveled than Wrapped, but here's a tour of the things you can do with your Spotify extended streaming history if you request it and load it into Curio: www.furia.com/log/498

NOW you can go to www.spotify.com/us/account/p... and request your Extended Streaming History (scroll down to "Download your data", check the box on the right) for data fun in Curio.

Seems like @everynoise.bsky.social ought to resolve to @everynoise.com after I switched the handle. Can I make it do so, @bsky.app?

2024 is done in the world, but Spotify data-days are on GMT, so if you live west of the Atlantic, wait until the 2nd to request your Extended Streaming History.

When I did the Listening Personality for Spotify Wrapped a few years ago, I reduced four measures of listening to single numbers, so I could produce a Myers-Briggs-esque 4-letter coding, because that was buzzy enough to get it shipped. But every number is a spark dreaming of being a heatmap.

Some ways of looking for the stories your listening tells. everynoise.com/curio.html

Curio query flex: let's say you live in NYC, but you spent the first weeks of April and September in LA, and one long fall weekend in Iceland. Here's how to convert all the GMT timestamps (ts) to the right local times.

Whoops. Accidentally reset my DNS nameservers while trying to switch my Bluesky handle. furia.com/everynoise_public until everynoise.com resolves again.

Curio is private. It uses your listening data, and stores all its own additions in your browser's local storage. I mention this because I just added a heatmap of listening by day and hour, and realized that from it you can completely reconstruct the dates of a decade of my trips to Sweden.

Curio will make a playlist out of your year of listening. Spotify already made one out of 10/12ths of it, which is one reason I bothered. But also, your listening is yours, and the playlists Spotify makes "for" you are always Spotify's. Curio has no retention metrics. Your playlists are yours.

Top-album banner.

Duration heatmap.

It's a pretty basic tenet of mine that "the year in music" means the music released this year, not just whatever I played this year. But you aren't required to care about this just because I do.

Do you listen to albums? Me too! Sometimes. A little more complicated to calculate if you also listen to singles that end up on albums, which me too, but grouping is the key to data analysis, and if your query language doesn't think groups are as real as anything else, you should use mine instead.

Update: the year is still not over. But I'm spoiling some of my own surprise to get things ready for yours.

Oh, I was wondering when the French translation of my book was going to come out, and the answer is that it already did! amzn.eu/d/a7cu9jt

Haven't done the dayglo burger graphics yet, and won't ever, but the data views are starting to data.

If you like your human culture ground into low-intention AI mush, good news: you're going to get a lot of that. If you're interested in how data and algorithms might work for curiosity and joy, instead, you might like my book. www.furia.com/log/487