Re:Why the Substack hype is much ado about very little
by fragmede on Tuesday July 27, 2021 @05:34AM
(#27970429)
Pandora.com, which launched in 2000, predating Spotify by 6 years, did a really good with discoverability. That was actually their primary feature - customizable radio stations, that you'd seed with a specific suggestion, and then further curate with a thumbs up/thumbs down. If you put in some time and effort, the stations would get really good at knowing what you liked, and would introduce new music. Spotify's radio feature is a poor shadow of this feature. (Soundcloud's is even worse.)
Because it's music, and not written text, and thus on in the background, having an 80/20 (or whatever) split for old/new music was really great for discovering new music that's similar.
TikTok solved discoverability for brief video clips, but they demand full attention. Basically the opposite of music.
Discoverability for the written word is a bit more complex, especially for long form articles.
(Pandora's backend implementation irrelevant here. Pandora predates the popularity of ML, so I presume there was a lot of human effort going on behind the scenes to make it work so well.)
Because it's music, and not written text, and thus on in the background, having an 80/20 (or whatever) split for old/new music was really great for discovering new music that's similar.
TikTok solved discoverability for brief video clips, but they demand full attention. Basically the opposite of music.
Discoverability for the written word is a bit more complex, especially for long form articles.
(Pandora's backend implementation irrelevant here. Pandora predates the popularity of ML, so I presume there was a lot of human effort going on behind the scenes to make it work so well.)