Your Taste Didn't Stop Growing. Your Feed Did.

Your Taste Didn't Stop Growing. Your Feed Did.

At some point — and you probably can't pinpoint exactly when — your recommendations stopped surprising you. Spotify started recycling the same forty songs. Netflix kept surfacing the same three genres. Your TikTok For You page started feeling less like discovery and more like a mirror you'd already looked at a hundred times. You didn't change. The feed just decided you had.

This is the part nobody talks about when we celebrate how smart recommendation engines have gotten. They're optimized to keep you engaged, not to keep you growing. And those two things are not the same.

How Platforms Build a Version of You (And Then Get Stuck With It)

Here's the uncomfortable reality: every platform that learns your behavior eventually stops treating you like a person and starts treating you like a profile. A cluster of signals. A set of weights in a model trained to predict what you'll click on next.

The technical term is collaborative filtering — the idea that people who liked what you liked also tend to like certain other things, so the system nudges you toward those things. It works great early on, when the model is still learning. That phase feels like magic. The algorithm finds your niche, surfaces the obscure stuff, makes you feel seen.

But collaborative filtering has a ceiling. Once the model has enough signal, it stops exploring and starts exploiting. In machine learning terms, it shifts from trying new things to doubling down on what already works. From your side of the screen, that feels like the internet quietly deciding you're done becoming someone new.

Spotify researchers have actually published on this — the phenomenon where heavy users end up in what they call "filter bubbles" that shrink over time rather than expand. You're not imagining it. The math is real.

The Ghost of Your Past Self Is Running Your Feed

What makes this weirder is the lag. Most systems weight recent behavior heavily, but they never fully let go of older signals. So if you spent three months really into lo-fi hip hop in 2021, there's a decent chance some ghost of that phase is still haunting your recommendations. Platforms build what's sometimes called a "taste fingerprint" — and like actual fingerprints, they're surprisingly hard to change.

The result is something like digital creative stagnation. You're boxed into a persona assembled from your past clicks, your 2 a.m. rabbit holes, your comfort-watch habits. The algorithm isn't mean about it. It's just doing its job, which is to serve you content that statistically resembles content you've engaged with before. It has no incentive to take a risk on you.

And here's the kicker: the more you engage — even negatively, even just by watching something all the way through to hate it — the more you reinforce the box.

Practical Ways to Actually Break Out

So what do you do? Deleting your account and starting fresh sounds satisfying but it's mostly theatrical. You'll just rebuild the same profile faster because you're still the same person. The real move is subtler.

Starve the signal. Most recommendation engines interpret passive behavior as a vote. Autoplay is the enemy. Letting something run in the background while you do something else tells the system you love it. Start being deliberate: search for things instead of accepting what's served. The act of searching registers differently in most systems than passive consumption.

Use the platform wrong. This sounds glib but it's actually effective. Follow accounts or artists that are adjacent to your interests but not inside them. If you're into indie folk, follow a bluegrass label. If you watch a lot of prestige drama, deliberately seek out one foreign-language film per week. You're not tricking the algorithm — you're giving it new inputs to work with.

Leverage your browser like it's 2009. Go to actual websites. Use RSS if you can stand it. Read a music blog. Check out a Substack from a critic you've never heard of. Discovery that happens outside a platform's ecosystem doesn't feed that platform's model, which means you're building taste that the algorithm genuinely doesn't know about yet. When you bring that back — when you search for something you found on your own — the system has to reckon with a version of you it didn't predict.

Make the algorithm work for someone else for a while. Log into a streaming service and spend a session watching or listening to things you'd recommend to a friend who has totally different taste than you. It feels weird. That's the point. You're salting the model with unfamiliar data, which forces it into a more exploratory mode.

Actually use the feedback tools. Most platforms have them — thumbs down, "not interested," "don't recommend this channel." People almost never use these because the friction feels annoying. Push through the friction. A few dozen deliberate dislikes can meaningfully shift what a system thinks you want.

The Deeper Problem Nobody Wants to Admit

None of these tactics are a permanent fix, because the underlying incentive structure doesn't change. Platforms make money when you stay on the platform, and the easiest way to keep you there is to serve you things you already know you like. True discovery is risky — you might bounce, you might get bored, you might leave. Algorithmic comfort is the safer business bet.

This is why the most culturally alive people you probably know tend to have deliberate offline discovery habits. They read physical magazines. They go to shows. They have friends who text them weird links. They treat curation as something they actively do, not something that happens to them.

The algorithm isn't your enemy. But it's also not your friend. It's a system optimized for retention, not growth. Your taste can still expand — it just has to happen somewhere the model can't fully see.

Start there.