The RandTube Blog

There's a particular moment most heavy YouTube users have experienced. You open the app intending to watch one video, and ninety minutes later you're still there — but you've somehow watched six videos on the exact same topic, by the same three creators, covering the same handful of perspectives. You didn't choose that. The algorithm did.

YouTube's recommendation engine is one of the most sophisticated pieces of software ever built for consumer use. It processes an almost incomprehensible amount of data — your watch history, your searches, how long you watched each video, where you paused, where you rewound, what you watched after — and uses it all to predict what you'll click on next. By most measures, it works incredibly well. Engagement is through the roof. Watch time keeps climbing. Users stay on the platform longer than almost any other service.

But there's a cost that rarely gets discussed: the algorithm is optimised for engagement, not for your growth, curiosity, or wellbeing. And those are very different things.

The filter bubble problem

When you watch a cooking video, the algorithm notes your interest and suggests more cooking videos. When you watch political commentary, it suggests more political commentary — typically from creators with similar viewpoints, because disagreement is uncomfortable and uncomfortable feelings reduce watch time. When you watch anxiety-inducing news content, it suggests more, because high-arousal emotions keep people glued to screens.

Over time, this creates what researchers call a filter bubble. Your YouTube homepage becomes a mirror reflecting your existing interests and beliefs back at you, slightly amplified. You see more of what you already watch, less of what you've never tried. Niche interests get reinforced. New ones rarely get introduced.

The effect is subtle enough that most people don't notice it happening. But if you stop and think about the last ten YouTube videos you watched, how many of them were from a creator you discovered in the last month? How many covered a topic you'd never explored before? For most people, the honest answer is: very few.

What we lose when discovery disappears

The early internet was defined by accidental discovery. You'd follow a link from a forum, end up on a strange website, discover a band you'd never heard of, find a community around an obscure hobby you didn't know existed. There was no algorithm deciding what was relevant to you. The chaos was the feature.

That spirit of genuine discovery is increasingly rare on modern platforms. Everything is curated, personalised, and optimised. The result is an internet that's more comfortable but less surprising — and surprise, as it turns out, is essential to genuine learning and growth.

When was the last time YouTube genuinely surprised you? Not with another video in a series you already follow, but with something completely outside your usual sphere — a craft you'd never thought about, a language you don't speak, a sport from a country you've never visited, a scientific concept explained in a way that made your brain fizz?

The case for randomness

This is why tools like RandTube exist. Not to replace YouTube's algorithm — which does a genuinely impressive job of surfacing content you'll enjoy — but to supplement it. To carve out a space where the algorithm has no say, and discovery is left to chance.

There's genuine research suggesting that serendipitous discovery leads to more durable interests than algorithm-driven ones. When you stumble across something unexpectedly, the surprise itself makes it more memorable. You're more likely to seek it out again. You're more likely to tell someone else about it. The memory is richer because it was unexpected.

The algorithm will keep getting better at showing you what you already like. That's fine. But every now and then, it's worth pressing a button and letting chance take the wheel. You might be surprised where you end up.

Read full article →

One of the first decisions you make on RandTube is which category to spin. It seems like a small choice, but it significantly shapes your experience. Each category has its own character — its own range of surprises, its own community of creators, its own potential to take you somewhere completely unexpected.

Here's a guide to what each category tends to surface, and which ones are most likely to genuinely surprise you.

🔬 Science — the best category for genuine discovery

Science is arguably the richest category on RandTube. YouTube has quietly become one of the world's greatest science education platforms, home to thousands of creators explaining everything from quantum mechanics to mycology to the history of the universe. A random spin here might land you on a beautifully animated explainer about black holes, a hands-on chemistry experiment, a deep dive into how vaccines work, or a naturalist narrating a time-lapse of a forest floor teeming with life.

The range is extraordinary. You might get a ten-minute university lecture or a three-minute viral clip of a physics demonstration that defies your intuitions. Either way, you're likely to learn something.

🎵 Music — the category most likely to change your taste

Music discovery is one of the most personal and meaningful forms of stumbling across something new. A random spin in the music category might take you to a live recording of a jazz quartet performing in a tiny Tokyo club, a bedroom producer in Lagos who has ten thousand subscribers and sounds like nothing you've heard before, or a classical guitarist performing a piece that makes everything else feel quiet.

The algorithm will never show you these creators because you've never signalled interest in them. That's exactly why randomness is so powerful here. Some of the most loyal music fans in the world found their favourite artists by accident.

🎬 Documentary — for when you have time to go deep

The documentary category rewards patience. A spin here might yield a thirty-minute mini-documentary about artisanal salt harvesting in the Philippines, an investigative piece on a financial scandal you vaguely remember hearing about, or a stunning nature film shot in a part of the world you've never thought much about.

Documentary is the category most likely to make you feel like you've actually learned something substantial. It's best saved for when you have time to commit to what you find rather than skipping after two minutes.

✈️ Travel — for armchair exploration

Travel content on YouTube ranges from polished tourism videos to raw, unedited footage of a backpacker eating street food in a market at 2am. The random spin reflects that range entirely. You might land on a cinematic tour of Iceland's highlands or a vlogger's honest account of getting lost in a city where they don't speak the language. Both are valuable in their own way.

Travel is particularly good for discovering places you had no idea you wanted to visit — and for realising that the world is considerably larger, stranger, and more interesting than the algorithm's curated version of it.

🎮 Gaming — more interesting than you think

Even if you're not a gamer, the gaming category is worth a spin. YouTube's gaming content has evolved far beyond let's plays and walkthroughs. You'll find cultural essays about what games mean, speedrunning communities that have turned video games into a form of athletic competition, retrospectives on games that shaped a generation, and live esports coverage that's genuinely exciting even without context.

🎲 Any — for the true adventurers

If you're feeling genuinely adventurous, spin with no category filter. This throws the net across all of YouTube, and the results are wonderfully unpredictable. You might get something in a language you don't speak, a video about a hobby you've never considered, or a clip from a TV show broadcast in a country you've never visited. Embrace the chaos. That's the whole point.

Read full article →

Building a random YouTube video generator sounds simple — surely you just pick a random video ID and send the user there? In practice, it turns out to be a more interesting engineering problem than it first appears.

Why you can't just generate random video IDs

YouTube video IDs are eleven-character strings made up of letters, numbers, hyphens, and underscores. In theory, you could generate random combinations of those characters and check whether a valid video exists at that ID. In practice, this is almost completely useless. The space of possible IDs is enormous, and the fraction of those IDs that correspond to actual, publicly available videos is tiny. You'd generate thousands of invalid IDs for every valid one — and even the valid ones you found would be completely unfiltered, potentially surfacing content you'd rather not see.

The YouTube Data API approach

The approach RandTube uses is different. We use the YouTube Data API v3 — Google's official interface for programmatically accessing YouTube's data — to search for videos matching a random search term within your chosen category.

Each category has a set of search terms associated with it. When you hit Spin, RandTube picks one of those terms at random, sends it to the YouTube Data API, and retrieves a set of results. It then picks one of those results at random and shows it to you. This ensures that every video surfaced is real, publicly available, and at least loosely related to your chosen category — while still introducing genuine randomness into the selection.

Keeping the API key secure

One of the trickier aspects of building a client-side web app that uses an API is keeping credentials secure. If you put an API key directly in your JavaScript code, anyone who views the page source can see and steal it.

RandTube solves this using Netlify Functions — small serverless functions that run on Netlify's servers rather than in the browser. When you hit Spin, your browser sends a request to our Netlify Function, which holds the API key securely as an environment variable. The function makes the actual call to YouTube's API and returns the results to your browser. Your browser never sees the API key.

Rate limiting and abuse prevention

The YouTube Data API has a daily quota of 10,000 units. Each search request costs 100 units, meaning we can serve roughly 50 spins per unit of quota (each spin uses two API calls — one search and one stats lookup). To prevent any single user from burning through the quota, RandTube implements rate limiting: each IP address is limited to 20 requests per minute.

What's next

RandTube is a living project. Future improvements we're considering include filtering by video length (short clips vs long-form content), filtering by upload date (recent videos vs classics), and a "surprise me" mode that picks a category at random as well as the video. If you have ideas, we'd love to hear them.

In the meantime — go spin something. You never know what you'll find.

Read full article →