How to Find New Music: Push, Pull, and Random

How to find new music comes down to three modes, and almost nobody uses more than one. A system can push music at you, you can pull music toward yourself, or you can take the choice out of your own hands and let a random song generator hand you something you never would have picked.

Push, pull, random. Most people live inside push, a weekly playlist arrives, and then they complain that everything sounds the same. It does sound the same, because any single mode can only reach so far. The fix is not a smarter algorithm. It is using the two modes you have been ignoring.

How to find new music in three modes, not one

Push is when something else decides. Your streaming app builds a playlist from what you already play, a friend texts you a link, a video uses a track under a clip. You did not go looking. The music found you.

Pull is the opposite. You go looking on purpose. You read that a singer you love grew up on a certain label, and you go pull the whole roster. Pull is slow, it takes attention, and it goes deeper than any feed will take you.

Random removes choice from both sides. Nothing is optimizing for your taste and you are not steering either. You get a track chosen with no regard for your history, which is exactly why it can jump a fence that push and pull both respect.

Mode	Who chooses	Best at	Where it fails
Push	A system or a friend	Low effort, steady trickle, near your taste	Narrows over time, circles the same lane
Pull	You, deliberately	Depth, context, real ownership of a find	Slow, tiring, only reaches what you already know to look for
Random	Nobody	Range, surprise, jumping genres and eras	Low hit rate, no context, easy to dismiss too fast

Read that table and the trade is obvious. Push is comfortable and gets narrower. Pull is rich and takes work. Random is wide and mostly misses. No single one of them gives you range, comfort, and depth at once, which is why picking one and living there is the whole problem.

Push: the mode you already overuse

Push is the default, and it works. A recommendation feed can quietly hand you a good song most weeks with zero effort from you, and it is genuinely good at helping you find music you like, because it aims straight at your existing taste. That is a useful thing, and I am not going to pretend otherwise.

The trouble is what push optimizes for. It learns from what you already play, so it keeps handing you the neighbors of songs you like. Over months, the neighborhood shrinks. If you want to understand why your recommended playlist feels smart but small, I wrote a whole piece on how Discover Weekly actually works, and the short version is that it is built to feel familiar, not to stretch you.

The homepage already lays out five quick discovery tips for getting more out of tools you have, so I will not repeat them here. Instead, the useful move with push is to stop treating it as your only source. Skim the weekly playlist, grab the two or three tracks that catch you, and get out. Do not sit inside the feed hoping it will surprise you, because surprise is not what it is for.

Where push fits in a week: ten minutes, once. A quick check, not a home you live in.

Pull: the mode that goes deep

Pull is you doing the digging, and it is where the best finds tend to come from. You hear one track you love and you follow it outward. Who produced it? What label put it out? Who does this artist name as an influence in interviews?

Every one of those questions is a thread, and pulling a thread beats scrolling a feed because it lands you somewhere a feed will not go. Feeds stay near the popular center. A label’s back catalog, a producer’s side projects, a scene from one city in one decade, those live out past where push bothers to look. If you want a concrete method for the digging, start with how to find underground artists, which walks the label-and-collaborator trail step by step.

Here is what one thread looks like in practice. Say you love a single song by a band. Instead of asking your app for similar bands, you check who produced it, find that person made three other records, and go listen to those. One of them points you at a session player, that player has a solo project, and the solo project is on a small label whose entire roster now sits in front of you. You started with one song and pulled a dozen real leads out of it, none of which a recommendation feed would have surfaced, because none of them are close enough to your play history to count as similar.

Pull also has a memory problem worth solving. When you find something by digging, you want to know later that you found it, and how. Keeping a listening log helps, and scrobbling with Last.fm is the low-effort way to build one: it quietly records what you actually play, so your own history becomes something you can pull from later.

The failure mode of pull is fatigue. It takes real attention, and you cannot do it for an hour every night. It also can only reach what you already know to ask about, so pull alone slowly turns into a very deep, very narrow tunnel. Give it about twenty focused minutes a week and it pays off. Ask it to carry your whole discovery habit and it will burn you out.

Random: the mode almost nobody tries

Random is the one people skip, and it is the only mode built to break your pattern on purpose. Push respects your taste. Pull follows your curiosity. Both stay inside the shape of who you already are. Random ignores that shape completely.

That is the point. When a generator hands you a track with no idea what you like, it will cross genres, decades, and languages that your own habits quietly rule out. Most of what it plays will not stick, and that is fine, because you are not fishing for a high hit rate. You are fishing for the one song from a corner you would never have walked into. If your listening has flattened out into the same handful of playlists, the fastest way out of that music rut is to hand the wheel to something that does not know your history at all.

Random has an obvious cost: a low hit rate and no context. You get a name and a song and nothing else. The way to survive that is to decide fast and keep moving, which is its own small skill.

How to decide fast without being unfair to a track

Give a new song about 20 to 30 seconds, but do not only judge the opening. A lot of songs hide their identity behind a slow intro, so jump around: hear the start, skip to about a minute in, then sample a later section. You are checking the whole shape, not the doorway.

The one rule that matters: skip on boredom, not on unfamiliarity. Those two feelings are almost impossible to tell apart in the first few seconds, and unfamiliar is the feeling every good discovery starts with. If you skip everything that feels strange, you skip everything new by definition.

There is a slower signal worth trusting too. Sometimes a track does nothing for you in the moment, then a fragment of it turns up in your head three hours later. That is not random. The mechanics behind why songs get stuck in your head are also, in practice, a decent verdict on which new tracks actually landed. If it comes back on its own, it earned a real listen.

How to actually keep what you find

Here is the quiet killer of most discovery: the song is lost in the gap between hearing it and saving it. You like something in the car, tell yourself you will look it up later, and later never has the title. The find evaporated, and it felt like the music was the problem when the problem was the handoff.

So collapse the gap. The moment a track lands, save it, right then, into one catch-all playlist. Do not organize it. Do not decide which mood folder it belongs in. A single messy inbox playlist that actually captures things beats a tidy system you are too slow to use.

Sorting is a separate job for later, when you have time and a keyboard. Split capture from curation and your keep rate goes up more than any change to how you search. The best discovery method in the world is worthless if the songs it finds leak out before you save them.

There is a second leak worth plugging, and it happens after you save. A track sits in your inbox playlist, you never replay it, and in a month it is a stranger again. Saving is not keeping. Keeping is the revisit, which is why the routine below ends by replaying last week’s finds rather than only collecting new ones. A song you have heard three times on purpose is yours. A song you saved once and forgot is just a title you cannot place.

A weekly routine under an hour

A discovery habit nobody can keep is worthless, so this is deliberately small. All three modes, roughly under an hour, once a week.

1. Push check, about 10 minutes. Open your weekly recommended playlist. Skim it, save the two or three tracks that catch you, and leave. Do not finish it.
2. Pull dig, about 20 minutes. Pick one artist you already love and follow a single thread outward: a collaborator, their label, or a listener with taste like yours. Save whatever you find.
3. Random jolt, about 10 minutes. Run a generator ten times. Keep anything you do not actively dislike. Decide fast, skip on boredom, move on.
4. Sort the keepers, about 10 minutes. Move everything you saved this week into one revisit playlist. That is all. No mood folders.
5. Revisit last week, about 5 minutes. Replay last week’s keepers. Drop the ones that already faded. The survivors are your real finds.

That is about 55 minutes, and it touches all three modes instead of leaning on one. Skip a week and nothing breaks. The point is a rhythm you will still be running in three months, not a heroic session you quit by Friday.

What does not work, and why

Living entirely in push and expecting range. This is the mistake almost everyone makes, and it feels reasonable, which is why it is so sticky.

The logic seems fine: the app knows what you like, so let it find more of that. The flaw is that “more of what you like” is a shrinking target. A recommendation system trained on your plays keeps steering toward the center of your existing taste, so the more you lean on it, the more it circles the same ground. You are not discovering. You are being reflected back at yourself with slight variations, and then you blame the music for getting boring.

There is a real tendency underneath this that is worth naming honestly. Research on what scholars call open-earedness, our willingness to explore unfamiliar music, finds that this willingness tends to peak in young adulthood and then decline with age. A widely cited analysis of how openness changes over a lifetime, drawing on a 2013 study of more than 250,000 people, describes exactly that pattern. It is a population tendency, not a sentence handed down on any one person, and the same work is clear that taste does not have to calcify. It can keep developing when you give it varied listening and a little patience. Push alone does not supply that variety. The other two modes do.

Most people already know, before they finish that experiment, which mode they never touch. It is almost always the same answer: push is the one they overuse, and random is the one they have never seriously tried.

So that is the whole move. The way to discover new music with real range is not a better feed or a secret app. Look at the three modes, find the one you use least, and spend a single week actually running it. If you have thoughts on how it went, or a mode I have framed wrong, tell me about it. And if random is your blind spot, which it is for most people, you already know where the generator is.

The songs you are missing are not hiding in a mode you have mastered. They are in the one you skip.