Yeah. 100% agree here. The Pandora algorithm was an absolute breath of fresh air relative to every other recommendation system. Just constantly surfacing new and interesting music that I had never heard of but was exactly what I liked. I spent so long hunting for info on how they achieved it and if their taxonomies had ever been open sourced.
In a recent thread talking about Flickr URLs I mentioned a couple of interesting technical features that Flickr had also done that I thought were worth celebrating.
When gwern mentions that you should write in more detail about those things I feel you have to take that seriously. This is that attempt.
Flickr deserves a lot of praise for a number of technical advances that I wish had seen wider adoption.
Their API was one of the first and honestly still one of the most enjoyable to actually use as a developer. It's still full of incredibly interesting API calls that you wouldn't expect from it unless you read carefully. Did you know, for example, that flickr API will provide you with the bounding box co-ordinates of different types of places? From a neighbourhood all the way up to a continent?
They implemented the Where On Earth ID (WOEID) which was a super useful way of disambiguating different places that shared latitude and longitude (for example, being able to disambiguate the Sydney Opera House, Circular Quay and Sydney Harbour which all can potentially share the same lat/long co-ords).
They implemented machine tags which are tags in the form of -
namespace:predicate=value
Which, when it was implemented by other sites with machine tags allowed you to get and group all kinds of interesting combinations of content.
Yeah, honestly flickr had some incredible tech the was so much fun to explore and use. That their vision of what the web could be wasn't the one that won is one of the great losses of the web IMO.