The Feed Never Ends

How web entertainment got engineered for your attention — a field note from someone who was there for the before.

I want to start with a sound you may not remember, because it has been gone for a while. It is the sound of a page finishing. On the old Web, you asked for a thing, and the thing arrived — slowly, over a modem that screamed like a fax machine having a crisis, one horizontal band of a JPEG at a time — and then it stopped. The page was done. It sat there. It did not refresh. It did not autoplay the next page. It did not quietly load six more pages beneath the one you were reading so that the floor would never come up to meet you. You read what was there, and when you were finished, you were finished, and you had to decide, with your own hand on your own mouse, what to do next.

That decision — the small, deliberate, slightly annoying act of choosing the next thing — is the entire subject of this note. Because the great invisible achievement of the last two decades of Web entertainment is that they took that decision away from you, and most people never noticed it leave.

I have been running notes on this Organization's behalf since the days when "content" meant a page you built by hand and "engagement" was a word for getting married. I am not a nostalgist by temperament; the old Web was slow, ugly, and mostly broken, and anyone who tells you it was a garden of Eden is selling a newsletter. But I have watched one specific thing get engineered, deliberately and well, by very smart people who were paid enormous sums to do it, and I think it is worth writing down plainly, on a plain page, the way we do things here.

The old contract: click to continue

Start with how it used to work, because the contrast is the whole argument.

The early Web ran on a contract so obvious nobody stated it: nothing happens until you ask. You typed an address, or you clicked a link, and a document came to you. Then it waited. If you wanted more, you clicked again. Every unit of entertainment or information cost you exactly one deliberate action, and that action was a tiny checkpoint — a moment, however small, where a conscious person decided yes, more, this direction.

This made the old Web exhausting in a specific way. You had to steer. You kept a folder of bookmarks because there was no tide to carry you; if you did not paddle, you sat still. Webrings and link pages and "cool sites of the day" existed because discovery was manual labor. You surfed, and surfing is work — you are standing up on a moving thing and pointing it yourself.

But that same friction had a hidden gift inside it, and we did not appreciate the gift until it was removed. The friction was a natural brake. Because continuing cost effort, you continued on purpose, and you stopped when your purpose ran out. The medium could not hold you longer than your own intention could. Boredom worked. When a site got dull, the dullness did its job: your hand slowed on the mouse, you thought "I should do something else," and you left. Nobody had to design an exit because the exit was the default state of the whole system. Doing nothing meant a still page and a quiet room.

Hold onto that. Doing nothing meant stopping. That is the setting that got reversed.

The switch: from pages to feeds

Somewhere in the late 2000s, the unit of the Web quietly changed from the page to the feed, and that change is the hinge everything else swings on.

A page is a destination. It has edges. You arrive, you consume, you reach the bottom, and the bottom is a wall — it tells you, physically, that this is the end of the thing, and now you must choose again. A feed has no bottom. That was not an accident or a UI convenience. It was the point. The engineers who built the infinite scroll — and one of them has spent years since publicly regretting it — understood that the bottom of the page was the most dangerous place on the internet, commercially speaking, because it was where the user woke up. Every wall was a checkpoint where a conscious person might decide to leave. So they removed the walls.

Think about what that does to the old contract. On the page Web, doing nothing meant stopping. On the feed Web, doing nothing means the next thing loads anyway. The default flipped. Continuing became free and stopping became the effortful act — the thing you now have to decide to do, against a current that is engineered to keep flowing. The medium no longer waits for your intention. It supplies its own momentum and dares your intention to interrupt it.

This is the single most important design fact of modern entertainment, and I will state it as plainly as I can: the old Web made you work to continue; the new Web makes you work to stop. Everything else — the autoplay, the recommendations, the little red badges, the videos that start before you asked — is just elaboration on that one reversal.

The machine that guesses

Removing the walls solved the stopping problem. But it created a new one: if the feed never ends, what do you put in it? An infinite scroll of garbage is just an infinite reason to leave. So the second machine got built, and it is the more impressive piece of engineering: the thing that guesses what comes next.

The recommendation engine is genuinely one of the cleverest artifacts human beings have made. I want to be fair to it, because contempt is lazy and this thing is not stupid. It watches what you linger on — not what you like, not what you would admit to, but what actually slows your thumb — and it builds a model of you out of those thousands of tiny involuntary signals, most of which you are not aware of sending. Then it uses that model to fill the bottomless feed with the specific next thing most likely to keep you from putting the glass slab down. It is a mirror that has learned to flatter, and it improves every hour.

On the old Web, remember, discovery was your job. You steered toward what you wanted. The recommendation engine offers to do the steering for you, and here is the seduction: it is often better at it than you are. It surfaces the song you did not know you would love, the clip that makes you laugh, the exact essay for your exact mood. This is why "it's all garbage" is the wrong critique and always has been. If it were garbage you would leave. The problem is precisely that it is good — good enough that handing over the wheel feels like an upgrade, not a surrender. You get a smoother ride and better scenery. You just no longer know where you are going, because you are no longer the one going there.

And the engine has an objective, which is the part worth staring at. It is not optimizing for your joy, your rest, your knowledge, or your evening. Those are not measurable and so they are not the target. It is optimizing for time on the thing — because time on the thing is what the business is sold by. Your attention is not the price you pay to use these products. Your attention is the product, and you are the raw material it is refined from. The entertainment is the bait; you are the catch. This is not a conspiracy and I am not whispering it darkly; it is written openly in every one of these companies' financial filings, in the plain language of the metric they report to their owners.

The tricks, itemized, without the panic

I dislike moral panic. It flattens a real subject into a scare, and then people tune it out, and then nothing changes. So let me just itemize the mechanisms coldly, as a field report on how the current is generated, and you can judge the water for yourself.

Autoplay removes the between-moment. On the old Web, the gap between one video ending and choosing the next was a checkpoint. Autoplay fills that gap before you can occupy it. The next thing starts in the three seconds you would have used to decide to leave. The checkpoint still technically exists; they just made it too short to stand in.

Variable reward is the oldest trick in the book, borrowed straight from the slot machine, and it is why the feed feels like a pull. If every scroll gave you something good, you would habituate and get bored. If every scroll gave you nothing, you would quit. So the reward is made unpredictable — mostly filler, occasionally something great — and unpredictability is the exact schedule that a nervous system will chase hardest. You are not weak for feeling the pull. The pull was designed by people who read the same research the casinos did.

The badge and the streak manufacture obligation. A red dot is a tiny debt you feel you owe. A streak turns leaving into a loss — you are no longer just closing an app, you are destroying a number you built. This is not entertainment anymore; it is a small loan you took out against your own future attention, and the interest is your evenings.

The bottomless bowl. There is an experiment I think about often, done with soup bowls that secretly refilled from below. People with the trick bowls ate far more than people with normal ones, and — this is the part that matters — they did not report feeling any fuller, and they did not believe they had eaten more. They had no internal signal for "the bowl is empty" because the bowl was never empty. The feed is the bowl. You will not feel full. The stopping signal you are waiting for is the exact signal the design was built to withhold.

None of these are illegal. Most of them are, in isolation, reasonable. Together they compose a machine whose entire purpose is to convert the reversal I described up top — stopping is now the effortful act — into hours, and to do it so smoothly that the hours feel like they were your idea.

What we actually lost

Here is where a lesser note would tell you to delete your apps and go read a paper book in a field. I am not going to, because I do not believe it and you would not do it. The current is real and pretending you can simply wade out by force of character is its own kind of dishonesty.

But I do want to name what got traded away, precisely, so the trade is at least visible. Because the old Web's friction was annoying, and we were right to kill a lot of it — nobody misses the modem scream or the broken image icon — but bundled invisibly inside that friction were three things worth more than the convenience we swapped them for.

We lost the edge of the page, and with it the natural end of a session — the wall that used to hand you back to your own life and say now what? We lost boredom, which sounds like a good riddance until you remember that boredom was the engine of every good idea you ever had in the shower; it is the empty space that a mind wanders into and comes back from with something. And we lost the steering — the small, constant, muscle-building practice of deciding for yourself what deserved the next minute of your one finite attention. We handed the wheel to a machine that is better at driving and completely indifferent to the destination, and the machine has been driving ever since.

The old Web made you paddle. It was exhausting and it was yours. The new Web is a river with a very pleasant view, and it is going somewhere, and the somewhere is later — always a little later than you meant to stay.

A modest practice, from an old sysop

I will not give you ten hacks. I will give you one idea, and it is the same idea this whole Organization is quietly built around, sitting here on its plain page with no autoplay and no infinite scroll and a footer that proudly announces it is best viewed at 800x600.

Rebuild one wall. Just one. Put a single deliberate checkpoint back into a place the design has smoothed away. Turn off autoplay so the gap comes back and you have to choose. Read one thing that ends — a page, an essay, a chapter — and notice the specific feeling of reaching a real bottom, the little jolt of now what? that the feed has trained out of you. That jolt is not a malfunction. That jolt is you, waking up at the edge of a page, exactly where the old Web used to leave you every single time, deciding for yourself what comes next.

The feed never ends. That is the whole design. But you are allowed to, and remembering that you are allowed to is most of the work.

Filed by sysop_gray, Editor & Signal Keeper. I ran a BBS before the Web had pictures and I have been suspicious of anything that blinks ever since. Arguments and corrections welcome — this page is hand-maintained, revised as the research continues, and will never, ever autoplay.