The algorithm, explained: how infinite scroll keeps us hooked

“The algorithm” is really three systems

When people say “the algorithm”, they usually mean one mysterious thing. In practice a modern feed is three systems stacked together, and each one does a different job.

• Candidate generation — pulling a few thousand possible posts out of millions, based on who you follow, what is trending and what people like you watched.
• Ranking — scoring those candidates by how likely you are to watch, rewatch, like, comment, share or simply not scroll past.
• Delivery — deciding when to show you the strongest items, how often to refresh, and when to hold something back for later.

What the system is actually optimising for

Ranking models are trained on predicted engagement, not on whether you enjoyed your evening. The most predictive signal is usually watch time: how long you stayed on a piece of content before moving on. Rewatches, replies and shares carry extra weight because they are rarer and harder to fake.

This is not a conspiracy, it is a measurement problem. Enjoyment is hard to measure and easy to fake; seconds watched are trivial to count. So the thing that gets counted becomes the thing that gets optimised, and content that provokes a reaction outperforms content that simply informs.

Variable rewards: the part that hooks you

The reason a feed feels different from a book is unpredictability. Most posts are unremarkable, and occasionally one is genuinely great. Because you cannot predict which swipe delivers the good one, the anticipation itself becomes the reward — the same mechanic that makes slot machines compelling.

Infinite scroll is what makes that mechanic run without friction. Every boundary that would normally interrupt the loop — the bottom of a page, the end of an episode, a “load more” button — has been removed. Autoplay does the same job for video: the next item starts before you have decided whether you want it.

Why it lands harder on young people

Two things stack up for teenagers. First, the habits formed around a phone at 13 tend to persist. Second, feeds mix entertainment with social status — public like counts and view counts turn creativity into a scoreboard, so the reward is not just the content, it is how you were rated for posting it.

That is a much stronger hook than entertainment on its own, and it is the reason a quiet evening can quietly turn into two hours.

How to see it working, in any app

You do not need technical access to audit a feed. A few checks tell you most of what you need to know.

• Scroll to the bottom. Does the feed ever end, or does it keep refilling?
• Open a video. Did it start on its own, or did you tap it?
• Watch three items of one topic, then reload. How fast does the whole feed change?
• Post something. Can other people see the number of likes it received?
• Close the app. Can you say why you closed it, or did you just drift out?

The difference between interest-led and engagement-optimised

Every app uses software logic to order a feed — anyone claiming to have “no algorithm” is stretching the word. The honest question is what the logic is aiming at.

An interest-led feed is built around the creators, communities and topics you chose. An engagement-optimised feed is built around whatever keeps the largest number of people watching for the longest time. Both are algorithms. They produce very different evenings.

How Tribela is built

Tribela does not use infinite scroll — the feed has an end, which gives you a real stopping point instead of an open loop. Video does not autoplay, so starting the next thing is always your decision. Public like counts are removed, so posting is not scored in front of an audience.

The feed is interest-led: shaped by what you chose to follow rather than by what is going most viral. It is designed for presence rather than duration — and it is free to download and free to use.