It seems you had enough labels to train, that's not always the case.

Fingerprinting, without excessive investment doesn't work well for mobile which eats and ever-increasing share of online traffic.

In addition more and more traffic will be agent based.

I don't think Service Model is really hopeful. The society has collapsed and robots are stuck on useless errands.

I have a hard time understanding your opinion. What do you even mean by garbage? There's thousands of legitimate music bands/musicians on these platforms. Naturally you won't care for a majority of it, but an interpret being niche doesn't make them garbage.

FWIW I went through a phase of this (feeling like streaming services were full of low grade stuff) until one day when basically nobody seemed to understand me except Amazon Music, which has an almost eerie recommendation mechanism that I think people don’t talk about much because the rest of the app is so meh.

In one single afternoon Amazon Music served up loads of niche old favourites I had not thought to search for, introduced me to a couple of artists I would not have known existed otherwise, and probably rescued my failing love of music.

I now play a game with it and pick somewhat diametrically opposed aspects of my music tastes in quick succession— Superorganism, Cardiacs, Gavin Bryars, Nanci Griffith, Vivaldi, Anouskha Shankar, say — and then see what it comes up with. Invariably one of the recommendations will be something I already adore and one will change my life.

I stumbled into the Chris Thile multiverse this way, for instance.

Spotify is full of AI slop trying to pass as artists. That’s garbage

Also, the company has a pretty terrible leadership: https://www.latimes.com/entertainment-arts/music/story/2025-...

So dont listen to it? I mean what is the problem here.

I never listen to basically anything spotify recommends me, just songs from bands I already know and love. And if I were to purchase the whole discography of the breadth of music I listen to it would cost the down payment on my house

I do not use the service, I was responding to the comment about garbage

Julia is going to overcome Python in usage, any decade now.

I like how this discussion has focused on Netflix where as the point is something different.

There are 2 useful rules to consider when building stuff: * Your user might not necessarily be like you - this applies even more when we tech people consider ourselves to be like our customers - we're often very different. * While in individual experiences your click of a thumbs up button constitutes a yes, and ...; if we're talking 128k likes, the aggregate dulls some of the edges and has interpretative value.

Interesting. I have had mostly the opposite experience. CEO(s) that changed strategy + focus every month and the middle management having to pick up the pieces. Keep people motivated and deliver.

When was this? MS Copilot has gotten a lot better lately IMO.

This seems to be one of the cases where what people say in a survey and what they do are two very distinct things.

Databricks on Azure went down too yesterday