part of the reason everyone has a million tabs open all the time is because browser bookmarks are legitimately an extremely awful way to save links. the underlying design metaphor hasn’t changed in, what, over two decades? it comes from a much smaller and simpler internet
Conversation
folders are completely the wrong structure. what you really want is to be able to throw a link into a service that will spit it back out at you *when it becomes relevant*, and/or *when you’re in the mood to read it*. folders don’t capture context- and mood-dependence
6
1
62
i’ve started thinking about “context- and mood-dependence” using the word “tuning,” so far only in conversations with - the original motivation was wanting to “tune” twitter to specific moods and contexts but i increasingly realized i want tuning *everywhere*
2
32
i want to tune spotify to specific moods and contexts, i want to tune youtube to specific moods and contexts, etc. etc. etc. i think a decade ago we weren’t good enough at ML to do stuff like this but surely we must be by now?
Replying to
i was really tickled by independently observing that his twitter feed seemed “mistuned.” the time is ripe for tuning!
Quote Tweet
My feed seems weirdly mistuned. Like a radio tuned to the wrong channel. But I haven’t been able to retune it. 
Not sure what I’m looking for. I’ll know when I find it.
Not sure what I’m looking for. I’ll know when I find it.Show this thread
1
1
22
currently the way i actually save links is to treat browser bookmarks as an inbox where links go to be sorted later, usually into roam with keywords. ’s memex looks promising but i haven’t played with it much yet:
getmemex.com
9
Replying to
Quote Tweet
Replying to @QiaochuYuan @sam_havens and 3 others
stay tuned
1
2
this problem doesn’t need complicated ml to do well imo
1
2
Show replies
Replying to
working on this rn with spotify and ml, research in associating mood with musical features in data science started a decade ago what was missing then was the widespread adoption of streaming services
1
1
suspect that the reason it hasn't happened already is the ability to resolve and control context as a user undermines viral content promotion.
you can't hookworm me into a content stream designed to legibilize my behaviour until i want to buy x if i'm not feeling that vibe rn
1
2
Show replies
Replying to
yes, it makes me sad that ML gets used to optimize "engagement" (bad on purpose to make you click) but not as a way to quickly do bespoke/custom classification and tuning for that sort of stuff
5
Replying to
I suspect any operationalizable definition of "mood" and "tune" and "context" will be wrong in some important ways.
(Setting aside incentive alignment: will allowing users to tune/customize their experience result in more engagement and ad clicks?)