Should recommendation systems adapt to listeners and artists, or the other way around?

Music streaming solved access. Choosing what to listen to can still feel like work:

• As a listener, you search, compare, skip, manage playlists and decide what should come next.

• Artists face a different version of the same problem. Releasing music now often means learning how feeds work, when to release, which formats perform, how often to post and what behaviour a platform rewards.

Both sides end up doing work for the recommendation system and I keep wondering whether AI gives us a chance to change that relationship.

For listeners, the interaction could be much simpler. Tell the system what you need from the moment, then let the music continue. Change direction when you want to without starting another search or building another queue.

For artists, the system could do more of the matching. Instead of asking musicians to shape their work around recommendation mechanics, it could get better at finding the listeners who are likely to care about the music they already make.

That question is a big part of why we built musen as a personal AI Radio. You press play, then steer the radio with text, voice or a picture. It keeps going while adapting to what you ask for and how you listen.

But I think the question goes well beyond music.

How much work should a recommendation system ask from the people using it? And when artists or creators are involved, should they have to learn how to satisfy the algorithm, or should the algorithm get better at understanding who their work is for?

We are launching the new musen mobile apps here on Product Hunt on October 1, so I would especially like to hear how other people experience this, whether you are a listener, artist, builder, or just someone who has felt recommendation systems asking a bit too much from you.