I'm one of the developers on the SameTunes team. We've spent the whole summer working out of a house to build this platform. We're really proud of the work we've done and hope you all enjoy using it.
With SameTunes you can view interesting statistics about your taste, find your music compatibility with friends, and discover new music!
Nice work, it looks really polished! What I would suggest improving is the single user experience: what I can do without involving friends. Right now I only see a huge list of strangers in alphabetical order. Picking someone at random seems pointless. However, if strangers could be sorted by match, that could be really interesting. That could also serve as a benchmark for my matchscore with friends.
If this is prohibitively costly in your current setup, I would suggest mapping each user profile to an embedding vector and find best matches using approximate nearest neighbors, that scales really well.
I think the embedding idea in specific is quite good and there are several potential approaches! One is, if you let xᵢ be the song vector for user i (whose kth entry is 1 if user i has song k in their library and 0 otherwise). Then you can compute the "overlap" by something like xᵢ ⋅ xⱼ (with some normalization of course!) where ⋅ is the inner product. [0]
A simple approximation of this inner product would be to generate a random (potentially sparse!) matrix S whose nonzero entries are i.i.d. Gaussian, for example, and whose number of rows is much smaller than the number of columns [1], then you can instead store and compute
(Sxᵢ) ⋅ (Sxⱼ)
which gives you an approximate overlap, whose storage and computation requirements are much smaller for each user (since Sxᵢ is much smaller in number of entries than xᵢ).
-----
[0] Of course, there are many other similar methods! This is a particularly simple, but often fairly effective one.
Through the Spotify API we are given locational and positional data on some music objects. Your proposal to structure songs as vectors opens up a lot more flexibility for storing, weighting, and manipulating that data.
The way in which the overlaps are calculated is crucial to our platform, something like feeding the dot product into a weighting matrix would leverage this structure really well.
Your proposal is very similar to how comparisons are currently computed on SameTunes. But, leveraging existing mathematical theorems and restructuring the algorithm with linear algebra in mind would help a ton with some of the noise and normalisation issues we've experienced.
I really appreciate you taking the time to draft that out!
For sure! It’s a cool product (I got two people who messaged me about it this morning, before I even saw it on hn) and it’s pretty fun. Good luck with it! :)
I find the marketing message a bit hard to swallow. The perfect answer for “what kind of music do you listen to” in person shouldn’t be something where we stop the flow of conversation and pull out our phones. It’s unclear to me why I’d want to use this. It would be helpful to see some example screenshots on the home page for be to better understand the value.
I appreciate the feedback! I definitely think adding some examples to the homepage is a good feature.
This idea was something we, a group of college students, felt was missing from our social interactions. I agree, it's a bit clunky to set this up mid conversation.
I, personally, find that this works really well with existing friends or when forming new friends online through things like Instagram or Tinder.
To your point, we're always looking to expand our use cases and reduce the first time friction.
Perhaps a better way to go is to have something that does analysis of your own music library standalone as an entry point. That doesn’t require another person, which means most people on HN can try it on their own. If that’s shareable as a link, then not only can you send that as a reply on tinder (versus all this permission mutual stuff), it gives the viewer the ability to hit some big magic button that then finds the overlap with their own library, while also giving them their own stats separately.
Very good point. I think that's where something like last.fm really shines.
Starting off, we focused heavily on comparisons but have branched out since. For example, the stats page you're first taken to gives you some analytical information about your own music. Strengthening out this page with things like music evolution, recommendations, rankings, &c is foremost priority.
Currently, we have friend links that deep links into creating a friendship (e.g. mine is https://smtn.es/125.ferda-bruh). Pivoting this to share your statistics and music tastes is an interesting idea -- it definitely helps introduce the platform to new users and personalise the platform for existing users.
Lastly — is there a reason for the shortened url here? Personally that link looks very much like malware to the untrained eye, especially if I have no idea what sametunes is. At least if it was sametunes.com based I could go to that host to find out.
I guess the point is that it compares your favorites with friends, so you can listen to what your friends are listening to? My friend link is https://smtn.es/4973.99538c0985 if anyone wants to try.
Any app that wants my personal info to "connect me" from the get go creeps me out. Spotify already connects me with others. Although I find spotify UI a bit obtuse and their music suggestions moronic. I would be more interested on a mobile platform that connects me with independent artists (aka fresh music)...like Soundcloud combined with music guaranteed to get it right on suggestions not because some computer algorithm but because they have competent curators...Like Calmradio.
That's totally fair. I wish Spotify gave us more granularity in the data we ask for, I don't think we need the blanket permissions a user has to grant us (as someone who just deleted their Facebook and soon Google accounts, this is really important to me).
We've reached out to a lot of local artists here in Atlanta and hope to use the platform to recommend and share their content to those interested.
Hand curating playlists is also a big feature we've been toying with. That's where something like Apple Music's hand crafted playlists are usually a lot stronger than Spotify's computed playlists.
I haven't heard of Calmradio, it's exactly what I've been looking for! I've resorted to listening to music through a cheap FM radio to try and break out of the algorithm hell Spotify tends to put me in.
I have to agree with your points. I'm someone who uses Soundcloud because a lot of the genres I'm into are not that popular and the producers have between 500 to 5k followers. Because of this, a lot of them simply aren't on Spotify.
I would definitely use an app that connects me with other people that are into the same niche genre because it's so hard to find people into this music anyway. The cool thing though is that when you do meet someone in the same hard-to-find niche, the friendship is much more stronger.
That's definitely something we're focusing on more now. Like qnsi mentioned, last.fm does a pretty good job of that -- I've been a fan since high school.
We, SameTunes, have an advantage when it comes to calculating a user's affinity for songs, artists and albums. Last.fm ranks your music by tracking how many times you've played (or "scrobbled") a song, songs by an artist, etc. SameTunes uses additional data Spotify publishes (like position in a created playlist, occurrences in your liked music, etc) to build a more well rounded representation of your music taste in the past and present.
There are definitely features we can, and plan to, add to help better portray this data to a user.
It's interesting that these kind of false premises come up often on HN. Don't get me wrong, I don't like PHP either, but nowadays. PHP is just as "modern" as many other languages in use. And with frameworks like Laravel and Symfony, it stacks up well against the likes of Ruby on Rails, Django or any NodeJS based stack.
And let's not forget that the majority of websites and webapps runs on PHP.
I think that’s totally fair. I have a personal project that I’m getting started on and I’m having trouble picking a stack to use. I hear a lot of conflicting opinions on here. I’ve looked through the documentation for newer frameworks like React and they seem really easy to get started with.
I'm in no place to recommend technologies to chose (I'm no expert) but the documentation argument is a good one. There are a lot of articles from the past few decades introducing different odds and ends with PHP. Additionally, PHPDocs are great but they're nothing in comparison to the tutorials and examples provided by things like ReactJS' or Vue's official docs.
When I first joined, the existing codebase was already built out in PHP on a LAMP stack. I was really opposed to it at first but I've learned to love it since. The safety, speed, and size of a compiled language on the web is something most modern frameworks can't compete with.
I, personally, have built a ton of in-house APIs here aiming to "modernise" php and am really proud of the outcome. It's nowhere as safe, readable, or dynamic as something like a Rust or Swift but PHP is definitely a strong competitor for the modern web (IMO).
I've been in a similar position at my current job. I was not excited at all to dive into PHP code simply because of how I perceived it externally. But once I started I was pleasantly surprised by how much "just works".
I'm one of the developers on the SameTunes team. We've spent the whole summer working out of a house to build this platform. We're really proud of the work we've done and hope you all enjoy using it.
With SameTunes you can view interesting statistics about your taste, find your music compatibility with friends, and discover new music!
Let us know what you think
If this is prohibitively costly in your current setup, I would suggest mapping each user profile to an embedding vector and find best matches using approximate nearest neighbors, that scales really well.
When building out the platform, efficiency was a foremost consideration. Because of that, each comparison calculation flies.
That's a good implementation suggestion, it reminds me of the adjacency matrix Twitter uses for mutual friends.
I'm drafting up proposal for your idea now, it serves a really strong introduction to the platform.
Thanks!
A simple approximation of this inner product would be to generate a random (potentially sparse!) matrix S whose nonzero entries are i.i.d. Gaussian, for example, and whose number of rows is much smaller than the number of columns [1], then you can instead store and compute
which gives you an approximate overlap, whose storage and computation requirements are much smaller for each user (since Sxᵢ is much smaller in number of entries than xᵢ).-----
[0] Of course, there are many other similar methods! This is a particularly simple, but often fairly effective one.
[1] More specifically, it goes like O(log(n)/ε²) where ε is the error you wish to achieve. Often, a fairly large choice of ε actually will suffice. See https://en.wikipedia.org/wiki/Johnson–Lindenstrauss_lemma
Through the Spotify API we are given locational and positional data on some music objects. Your proposal to structure songs as vectors opens up a lot more flexibility for storing, weighting, and manipulating that data.
The way in which the overlaps are calculated is crucial to our platform, something like feeding the dot product into a weighting matrix would leverage this structure really well.
Your proposal is very similar to how comparisons are currently computed on SameTunes. But, leveraging existing mathematical theorems and restructuring the algorithm with linear algebra in mind would help a ton with some of the noise and normalisation issues we've experienced.
I really appreciate you taking the time to draft that out!
This idea was something we, a group of college students, felt was missing from our social interactions. I agree, it's a bit clunky to set this up mid conversation.
I, personally, find that this works really well with existing friends or when forming new friends online through things like Instagram or Tinder.
To your point, we're always looking to expand our use cases and reduce the first time friction.
Starting off, we focused heavily on comparisons but have branched out since. For example, the stats page you're first taken to gives you some analytical information about your own music. Strengthening out this page with things like music evolution, recommendations, rankings, &c is foremost priority.
Currently, we have friend links that deep links into creating a friendship (e.g. mine is https://smtn.es/125.ferda-bruh). Pivoting this to share your statistics and music tastes is an interesting idea -- it definitely helps introduce the platform to new users and personalise the platform for existing users.
I'll propose it to the team, thanks!
We've reached out to a lot of local artists here in Atlanta and hope to use the platform to recommend and share their content to those interested.
Hand curating playlists is also a big feature we've been toying with. That's where something like Apple Music's hand crafted playlists are usually a lot stronger than Spotify's computed playlists.
I haven't heard of Calmradio, it's exactly what I've been looking for! I've resorted to listening to music through a cheap FM radio to try and break out of the algorithm hell Spotify tends to put me in.
Thanks for sharing!
tldr; OP add SoundCloud support please!
We, SameTunes, have an advantage when it comes to calculating a user's affinity for songs, artists and albums. Last.fm ranks your music by tracking how many times you've played (or "scrobbled") a song, songs by an artist, etc. SameTunes uses additional data Spotify publishes (like position in a created playlist, occurrences in your liked music, etc) to build a more well rounded representation of your music taste in the past and present.
There are definitely features we can, and plan to, add to help better portray this data to a user.
When I first joined, the existing codebase was already built out in PHP on a LAMP stack. I was really opposed to it at first but I've learned to love it since. The safety, speed, and size of a compiled language on the web is something most modern frameworks can't compete with.
I, personally, have built a ton of in-house APIs here aiming to "modernise" php and am really proud of the outcome. It's nowhere as safe, readable, or dynamic as something like a Rust or Swift but PHP is definitely a strong competitor for the modern web (IMO).
This article https://mcfunley.com/choose-boring-technology also really helped me come around early on.