> I strongly disagree. Spotify has dramatically increased the variety of music that I listen to and introduced me to lost of tiny artists with great music.

I've used a number of music recommendation applications over the last decade.

With Spotify, I've noticed that it will become "stuck" in what I describe as similar artist troughs and valleys. In other words, the Spotify recommendation engine will pigeonhole your recommendations into certain subgenres, niches, or related artists more so than say Pandora or Last.fm would.

Yup, I've had to work around this. I'll pick a song I like, and do "song radio" to find more like it. But this makes a playlist which is basically 6 more similar artists, and no variety beyond that. But I want to branch out more, several steps beyond that. So I'll open a particular song from that playlist, go to "song radio" for that, and drag a bunch of songs I don't recognize back into the first playlist. Repeat this for a few more songs. Then once I've added a bunch more variety, do "playlist radio" again.

It's super obnoxious. I wish they had a slider for "be more adventurous" versus "hew close to the line". Do I remember Pandora having something like that?

Spotify automatically creates a playlist from songs you've "liked" while they're playing on radio. So just start a certain song radio, like the kind of songs that go into the direction you want, and then start playlist radio on the "Liked from Radio" playlist.

You can simulate the slider behavior you describe by thumbing up everything for more variety, or thumbing down everything for more restriction.

Wholeheartedly agree. While my musical tastes are a bit all over the place, I can say that for the past month or so almost all my generated suggestion playlists consist of music I've not only listened to recently, but many I already liked and saved to my songs.

Spotify made a Tastebreakers playlist at the end of the year to recommend music I might like but is outside the genres I typically listen to.

I thought it was decent.

I can second this. Spotify has really helped me discover artists and bands I enjoy.

Years ago, I would have spent a non-zero amount of time linking through friends on Last.FM / Audioscrobbler. While this can be enjoyable, having it done by an algorithm is perfectly fine for me and I'm willing to pay for it.

> Spotify has dramatically increased the variety of music that I listen to and introduced me to lost of tiny artists with great music.

I can't speak to Spotify, but many years ago I picked up some agent software that does a rough analysis of the music I enjoy, then crawls the web looking for music that resembles it. I use that for recommendations.

The software is a bit crude in its analysis, but I think that's a feature, not a bug, because it ensures that in the mix of music that I certainly will like are random tunes that I would never have thought to listen to, and often those turn out to be gems.

Most of the music the crawler finds is from indie artists that aren't well-known at all. When I find a new one that I enjoy, I make it a point to seek them out and buy their music directly from them.

Anyway, this has been my solution for about 15 years now, and it works great for me!

Sounds like you made a little Music Genome Project (by Pandora) like system. Lately I've been thinking about building something like that using machine learning and the hopes that it would scale better than Pandora's effort did.

What's the name of this crawler software? Is it open-source?

It's a freeware script that I picked over a decade ago. I don't remember where I got it or what it's called, though. I'm not able to access the server it's on at the moment, but I'll comment here once I can.

What's the name of this crawler software?

>I strongly disagree. Spotify has dramatically increased the variety of music that I listen to and introduced me to lost of tiny artists with great music.

This has been my experience as well.

There was even a post on HN recently by a small artist talking about how the Discover algorithms brought listeners to his music that probably never would have found him otherwise.

I'm a big fan of the recommendation algorithms, so I set up a super basic site to wrap the Spotify API and let anyone get recommendations based on a single song link, rather than waiting for the Discover Weekly recs which are based off the blend of everything you listen to.

https://www.spotifinder.com/

Warning: the site is really basic right now

How does this compare to Spotify’s ‘Song Radio’ feature?

Honestly not sure since I don't know which APIs + secret sauce they put into Song Radio. I personally found Song Radio a little lacking - I never usually liked the most of the related songs, it didn't allow me to reshuffle the recommendations.

The /recommend API I wrapped for the site is similar, in that I let it spit out 12 "similar" songs, but I'll probably only like one or two songs, but you can also re-submit to shuffle the recommendations. The site is definitely more for active discovery than passively listening. I made the site because:

1. I wanted to more actively discover songs

2. I'm a big Spotify fan and was curious about the guts of the recommendations algorithm

3. I had never deployed an actual site (I exclusively do backend programming at work) so it was good learning experience about things like setting up a little static site

Spotify has good suggestions. All you need to pick is the genre and then you go to song/playlist radio and you are good. I have found many bands and it often recommends bands with foreign languages( which I don't want). Then there is also your discover weekly which usually gives me at least 1 or 2 good songs.

I can confirm this. Pre-Spotify it was either listen to top 100 music or use up a lot of time downloading VS tracks. Now most my of music doesn't even chart.

I neither have the space, budget, or time to trade my music in the cloud for physical media. My online collection is much larger than either my Mp3 or physical media collection.

I only keep Spotify because of "Discover Weekly" playlist generated every Monday based on your taste. The algorithm is truly incredible. There are weeks I find myself saving half of the songs.

found a lot of greeat stuff on spotify. it has never sent me some topN track as an ad... :s never. jut dont search for these things if u dont want it to suggest it to u ;)