How Some People Have Basically Cracked the Code on Finding Great Content Online

Somewhere in Portland, a 31-year-old graphic designer named Theo has a Spotify account that looks nothing like yours. He has seventeen playlists, each one named after a specific mood or micro-genre that Spotify's official genre tags don't even recognize. He follows 34 artists, most of whom have under 10,000 monthly listeners. His Discover Weekly is, by his own description, "almost unnervingly accurate."

He didn't get there by accident. He got there by spending about two years deliberately training his algorithm.

"Most people treat recommendation engines like a vending machine," Theo told us over email. "You put in money and you take what comes out. But it's more like a conversation. If you're intentional about what you put in, you get something completely different back."

Theo is part of a loosely connected, rarely discussed group of internet users who have stopped passively consuming algorithmic recommendations and started actively engineering them. Call them algorithm whisperers. They exist across every major platform — Spotify, YouTube, TikTok, Netflix, even Goodreads — and their strategies range from elegant to borderline obsessive.

We spent time with several of them, and also talked to people who actually build these systems, to understand what's really going on in this cat-and-mouse game between humans who want genuine discovery and platforms designed to keep them comfortable.

The Problem With Letting the Algorithm Drive

Before we get into the hacks, it's worth understanding why people feel the need to hack at all.

Recommendation algorithms — at their core — are optimization machines. They're built to maximize a specific metric, usually engagement or watch time, and they do this by surfacing content similar to things you've already demonstrated you like. This sounds helpful. In practice, it creates what researchers call a "filter bubble": a narrowing loop of increasingly familiar content that slowly walls you off from anything genuinely new.

"The algorithm is not trying to expand your taste," explained one data scientist who works on recommendation systems at a major streaming platform and asked to remain anonymous. "It's trying to predict what you'll click on next. Those are very different goals. One serves you. One serves the platform's retention numbers."

The result is something a lot of us feel without being able to name: the sense that the internet used to feel bigger. That there used to be more surprise in it. That discovery used to be something that just... happened.

It still can. You just have to be a little sneaky about it.

The Spotify Approach: Confuse the Machine (On Purpose)

Theo's method on Spotify starts with something he calls "palette expansion" — deliberately listening to music he's not sure he likes yet, in full, without skipping. Skipping, he explains, is a strong negative signal. Listening through is a positive one, even if you're ambivalent.

"If I want to explore a genre I know nothing about, I find the most-followed playlist in that genre and I listen to the whole thing like it's homework," he said. "I'm not always enjoying it. But I'm teaching the algorithm that I'm interested. Within two or three weeks, Discover Weekly starts pulling from that space."

Another technique: following micro-artists strategically. When Spotify sees you following artists with small but highly engaged fanbases, it starts connecting you to their fan networks — which tend to have overlapping taste in things the algorithm hasn't yet categorized neatly. It's a backdoor into recommendation territory that most users never see.

A music blogger in Nashville named Keely has taken this even further. She maintains what she calls a "sacrifice account" — a second Spotify profile where she deliberately listens to mainstream pop and nothing else, specifically to keep her primary account's algorithm clean. "My main account thinks I'm a very specific person with very specific taste," she said. "And it's right, because I've been very intentional about what that person likes."

The YouTube Rabbit Hole, Engineered

YouTube's algorithm is famously aggressive, but it has a vulnerability: it responds strongly to watch completion rate. A video you watch all the way through sends a much louder signal than one you clicked and abandoned after thirty seconds.

Daniel, a 27-year-old in Austin who runs a personal blog about niche internet culture, has developed a system he calls "signal stacking." He opens a new incognito tab when he wants to explore something genuinely new, watches several videos in that space to completion, then — and this is the key step — logs in and likes them before closing the tab.

"YouTube's recommendation engine weighs likes heavily," he explained. "If you like a video from a creator with 8,000 subscribers who makes incredibly specific content about, say, vintage synthesizer repair, YouTube starts to understand that this is a space you value. It doesn't know how to categorize it yet, so it starts pulling in adjacent content from similar micro-communities. That's where the interesting stuff lives."

He also recommends following what he calls "cultural connectors" — creators who sit at the intersection of multiple niche communities and tend to reference each other. Following one often leads, within a few weeks, to a web of interconnected creators the algorithm would never have surfaced on its own.

The TikTok Long Game

TikTok's For You Page algorithm is widely considered the most powerful recommendation engine currently running on a consumer platform. It's also, paradoxically, one of the easiest to deliberately shape — because it responds so quickly to behavior.

Amara, a 24-year-old in Brooklyn who creates content about sustainable fashion, has spent the last year carefully curating her TikTok experience through what she calls "active not-interested management." Every time a video appears on her FYP that pulls her toward the algorithmic mean — the big creators, the trending sounds, the content that's designed to be universally appealing — she taps "Not Interested" immediately.

"Most people never use that button," she said. "But it's incredibly powerful. The algorithm learns what you don't want just as fast as what you do. Within about a month of being really aggressive with it, my FYP looked completely different. I was seeing creators I never would have found otherwise."

She pairs this with a hashtag strategy that goes one level deeper than the obvious tags. Instead of following #sustainablefashion (enormous, noisy), she follows #slowfashionmaker and #thriftflipfinds — tags with smaller, more passionate communities where the signal-to-noise ratio is much higher.

What the People Who Build These Systems Think About All of This

We asked our anonymous data scientist whether the platforms are aware of these kinds of deliberate manipulation strategies.

"Oh, absolutely," they said. "And honestly? The people doing this are making the algorithm better. When a user gives really consistent, intentional signals, the model has more to work with. The problem is most users give extremely noisy signals — they watch something they hate because it autoplay, they click things they'll never finish. Intentional users are almost a gift to the system."

The irony, of course, is that the platforms aren't specifically trying to help these users find more interesting content. They're just benefiting from the cleaner data. The discovery is a side effect of the user's own effort.

The Bigger Point

What all of these people have in common is something pretty simple: they decided that finding great content was worth treating like a skill. Not a passive experience that happens to them, but an active practice they get better at over time.

That's a very GoLike kind of energy, honestly. The whole point of this place is that discovery — real discovery, the kind that makes you text a friend at midnight because you just found something incredible — doesn't happen by accident. It happens because someone cared enough to go looking.

The algorithm isn't your friend. But it's not your enemy either. It's a tool. And like most tools, it works a lot better when you actually learn how to use it.