I have spent evenings scrolling a music app and gone to bed having listened to nothing. Thirty minutes of looking at music. That is a genuinely strange way to spend a life, and it is not an accident — it is the product working correctly.

An infinite feed is optimised for time spent looking, because looking is where the advertising is. It is very good at what it does. What it is not good at is the thing I actually want from music, which is to be somewhere else for forty minutes.

So discovery here is built to end. That is the whole design goal. Every path is supposed to run out somewhere useful, with you listening to something, rather than loop forever.

Practically that means a few things. There is no endless feed. The mixes are a finite list and you can reach the bottom of it — twenty-eight of them, which is a number you could actually finish. The releases are seventeen. You can see the whole catalogue in one sitting, and I think knowing the edges of a place makes you more likely to go into it, not less.

It also means the filters are about you rather than about me. You can come at the catalogue by how something feels, or by when you would play it — late night, the commute, Sunday morning — instead of by genre, which as I wrote yesterday tells you almost nothing useful.

And there is search, which is deliberately boring and works across everything at once. Mixes, records, series, writing. One box. I have never understood why a platform would make you choose which kind of thing you are looking for before you look for it.

The part I keep coming back to is that recommendation systems are extremely good at telling you what you already like. That is what they are for. The problem is that a system which only ever agrees with you is not a discovery system, it is a mirror, and I have been in enough algorithmic mirrors to know they get boring in a way that is hard to notice while it is happening.

I do not have a clever solution to that. What I have is a small catalogue made by one person, which means the connections between things are real rather than statistical. If two mixes sit near each other here it is usually because they came out of the same six months of my life. That is not a better algorithm. It is just a different kind of true.

There is one thing I will defend without reservation: if you start something and leave, the platform holds your place. Come back tomorrow and it is where you left it. I want the friction to be at the beginning, when you are choosing, and nowhere after that.

So my ask is slightly odd for a launch article. Please spend less time here. Come in, find one thing, put it on, and go and do something else while it plays. If this place ever starts competing for your attention rather than your forty minutes, I have built the wrong thing.

Start here, pick one, and close the tab.