Lovecraft in Silicon Valley

15 Sep, 2026

Roblox turned a dial that makes millions of children spend more time playing games. The same dial works on you.

Won't someone think of the children?!

Over the past year, Roblox, a video game company (think YouTube for video games), has lost about $70 billion in market value. About $10 billion of that went in a single day, after the company's second-quarter call in July.¹ On that call the CEO and CFO explained why they had made less money than investors expected. Two reasons were given. The viral games that had driven 2025 were fading, and, on top of that, the company had changed its recommendation algorithm, the thing that decides which games appear on a player's "for you" page. That second reason is the one this article is about.²

The change moved players away from games with high bookings per hour (read: games that get you to spend money fast) and toward games with high long-term retention (read: games that get you to come back, over and over, for a month or more). Roblox did not hide this. It announced the change to its developers on June 15, six weeks before the call, and published the signals the new ranking uses.³

On the surface this looks benign, even admirable. A company choosing long-term loyalty over short-term cash. That is exactly how the CEO and CFO framed it.

It stops looking benign when you ask what the change does and to whom. Roblox's players are mostly children. By the company's own count, about three quarters of its age-verified users are under 18 and about a third are under 13.⁴ Fortnite is for older kids; Roblox is for younger ones.

In straightforward terms: they turned a dial that makes little Timmy show up to play more often. Not for much longer; an afternoon on Roblox runs about two and a half hours whatever they do. More afternoons. And they did not turn it for one Timmy. They turned it for millions of them at once, and they know it works because they tested it. In the CFO's words, the effect on how often players came back showed up "very quickly."⁵

The problem is not their ability to make more or less money. Companies wanting to make money off their users is arguably the reason companies exist. It is in their ability to reliably change how much of a person's life gets spent on their platform, and to measure that change to the day, in a population that is mostly children.

Now, Roblox has been called to account for its treatment of children, and so, most recently, has Facebook. Roblox set aside $91 million in the first half of this year for settlements with state attorneys general over child safety.⁶ Meta, in August, agreed to pay up to $17.1 billion to settle claims from 51 attorneys general that it designed Instagram and Facebook to addict minors. The remedies in that settlement are worth reading: a default two-hour daily limit for users under 18, a block on access between midnight and 6 a.m., and an option for young users to get a feed that is not run by an algorithm targeting them.⁷

Hearing about these settlements, one may start to think of this as a regulation problem in the vein of alcohol and cigarettes. Keep the children away and let the responsible adults handle it. Every remedy in the Meta settlement expires on a person's eighteenth birthday, so that is evidently how the attorneys general think of it too. This I believe grossly underestimates the scope of the situation. To understand why, let's do a short detour into how Roblox actually exerts its pull on a player.

How they do it

The mechanisms of action are too many to comprehensively list, and any attempt to be comprehensive is doomed to fail. Here is a small sample of how these platforms work to influence how much time users spend on them.

1The 'For you' page
This should be familiar to almost anyone who has used any form of Algorithmic Media in the past few years. When you log into Roblox, exactly like YouTube or Netflix, you are shown a menu of things to choose from, where each game was chosen to be shown to you because a machine predicted you would play it.

2The order is a prediction
Roughly speaking, the system sorts every game on the probability that showing it will mean the player is still on Roblox 28 days later, then lays them out in that order.⁸ The game at the top of Timmy's screen is the one the machine predicts Timmy is least able to leave.

3The guess is made from massive amounts of data
This is the 'Big Data' that you may have heard about. Roblox knows how long you play, which days you play, which days you play with friends, which days you spend money, whether you tap on a game, and whether you quit it in the first minute. It knows the same for 123 million people every day.⁹ Your next afternoon is predicted from a hundred million other afternoons. Roblox says it runs hundreds of AI models, over a million guesses a second, and the "for you" page is one of them.¹⁰

4The recommendation system is constantly improving
In the CEO's own words: the system is "constantly self-improving and self-estimating our best guess at what user-game pairs will drive the best long-term retention."¹¹ In plain English, it guesses which game will keep you, checks whether it was right, and fixes its guess. The CFO said the results show up "very quickly." Every week it gets a little better at bringing you back, and you are the test.

5The system creates the stickiest games
The system does not only choose what you see. It decides which games get seen at all, and it tells game makers what it wants.¹² So the pressure it puts on the players, it also puts on the games. Every developer is competing to build the thing that brings players back day after day, because that is what the machine rewards, and anything else disappears. Over time the whole library bends toward it. The system is not just picking the stickiest game. It's breeding them.

6There's so much more
More than I can cover, and more, I suspect, than the makers themselves realize. The algorithm knows your age, where you are, what your friends play, when you tend to show up and when you drift away. It sees which tiles you didn't tap. It sees which game you quit after a minute. To say nothing of clickbait and dark patterns. Every one of these is a small hook, none of them was designed by a villain, and nobody has an inventory of them all, because the machine finds them on its own. That is the point of the thing. It was built to discover whatever brings you back, and it does not need anyone to understand why.

Notice that none of these mechanisms are about surfacing the "best" game for the player, or the most enjoyable one. They are all about one metric: whether you come back. The assumption underneath is that a game you return to is a game that is good for you. That is an assumption we may have made naively in the early years of the internet. But in the world of clickbait, shocking headlines and oddly compelling horse hoof cleaning videos, we know there can be, and increasingly is, a large gap between content that is good for us and content we will readily go back to.

At this point you may be appropriately horrified. Now consider that nothing about these techniques makes children uniquely susceptible to them. The menu, the guess, the data, the self-correction, the pressure on the games, the friends: every one of them works the same on a forty-year-old. Roblox built it on kids because kids are who it had. It has said, on the same call, that adults are where it is going next, that adults in the U.S. are worth over 50% more per head, and that it now pays developers a premium for money spent by adults.¹³ The Meta settlement caps a sixteen-year-old at two hours a day and hands her the keys at eighteen. On the other side of that birthday the machine is the same machine, and there is no cap.

Enough about Timmy. Timmy has a bedtime. But you, you have no bedtime.

Algorithmic Oblivion

Do you like the relationship you have with the Algorithmic Media you consume? Do you know anyone who doesn't idly say something to the tune of "Oh, I spend way too much time on the phone, lol", "I shouldn't be on twitter/instagram/reddit so much" etc.? These impotent forms of self-critique can start to seem like a cover, allowing us to act as if we know better and yet keep going back to the same experiences over and over again.

The question to ask at this point, though, is why? Why would we do it? Us, with our fully developed frontal lobes? Why do we keep going back to things we profess to want to leave?

To understand, let's consider what it actually feels like to be on one of these platforms. What philosophers call the phenomenology of the experience.

There you are, after a long day at work. You look at your phone and Instagram is open almost before you realize it. A reel is already playing, some catchy music, maybe it catches your attention, maybe it doesn't.

scroll

Something horrible happening somewhere else.

scroll

Ad.

scroll

A funny meme about a show you watched when you were young.

scroll

A video about...

scroll

And suddenly, it's been an hour.

Notice what happened. You start off looking for some easy entertainment, and somewhere around the fourth scroll, you disappear. Time vanishes. You vanish. Your struggles from the day vanish. All that is left is the feed. Consider the texture of this experience, consider how it just slips off of your mind. Like it happens through you instead of happening to you, and while some part of you was present for it, it was almost as if you weren't really there.

Psychologists have a name for this kind of experience. The Hungarian psychologist Mihaly Csikszentmihalyi called it Flow.¹⁴ Humans have experienced flow in many ways: that afternoon where you're just flying through work, the feeling a long distance runner gets after mile 6, the feeling of a player fighting a tough boss in a video game. All Flow. Csikszentmihalyi called it optimal experience, and he meant it literally: when he asked people about the best moments of their lives, this is what they described.

But there is a different kind of Flow, one that has also been called 'The Machine Zone' or 'Dark Flow'. Most prominently, this kind of flow is experienced by gamblers. The anthropologist Natasha Dow Schüll spent years with slot machine players in Las Vegas and found that what they were after was not winning. It was the zone: a state where time, the room, and the self drop out and there is nothing left but the next spin. Psychologists at the University of Waterloo, studying slot players, found a similar state and named it dark flow¹⁵: a pleasurable but maladaptive absorption that depressed players use as an escape. It has the shape of Csikszentmihalyi's flow, the absorption, the lost time, the vanished self, but nothing is being built at the edge of anyone's ability. It enervates and gives nothing in return.

The platforms are giving us what I think can only be called a type of this dark flow. Call it Synthetic flow. It is not the creative flow of the runner or the productive afternoon. It is consumptive. Nothing is made, nothing is learned, nothing is carried out of it. And now we can induce it whenever, wherever, as many times a day as we like, and we do. No slot machine needed.

One may be tempted here to think that the answer is to leave the feeds for something else. But there is less refuge than it may seem. Algorithmic media is the law of the land in the culture industry, and media that isn't administered by an algorithm is increasingly built like an algorithmic product anyway. Movies, TV, big-budget games: the people making them have worked out the formula that yields an "Inspirational Story," a "Gritty Drama," a "Sci-Fi Gritty Video Game World," and they run it. Leave TikTok and you land on Instagram. Leave Instagram, YouTube, and where do you find yourself? Is there any real alternative even available?

Cthulhu in your home

Imagine a world with massive Godlike entities, all of which compete for our devotion. The more devotion we give to an entity, the more it grows powerful and the more we become attached to it. We make altars in our homes to the Gods and spend hours staring at them, we carry totems around that help us devote any stray moment to them. The Gods for their part are not necessarily cruel, they are simply indifferent to what they do to us. They simply want our devotion and they are not above hypnotizing and consuming our minds to get more of it.

This is the world that we find ourselves in. These companies and the platforms they manage are not like entertainment products of the past. The feeds are not simply a menu that is sorted by how much you would 'like' to encounter any specific bit of content. The feed, the content, it is all optimized to keep us coming back over and over.

Human struggle against the compulsion exerted by Algorithmic media is commonly framed as a personal responsibility and discipline problem. After all, one can always choose to not be on Social media. The truth is that the compulsive force exerted by these forms of "entertainment" is almost too strong for an individual human to contend with.

So, perhaps the next time you are thinking to yourself "I really shouldn't be on social media this much" consider what you are up against and then, leave anyway. Personal responsibility is only one piece of the puzzle, but it is the one that you have control over.

More importantly though, if 3 months after leaving you find yourself suddenly at the end of another bewildering doomscroll session and are tempted to kick yourself and think of this as a moral failing, consider: you are up against Alien Computers from the Planet Frypton. Maybe this fight is not yours alone to fight.

The river of attention

So far we have seen how Roblox manipulates its users, how platforms manipulate people in general, how no one is exempt from the effects of these systems, and how the individual, asked to take responsibility, is outmatched.

What then do we do? If personal responsibility is only part of the picture, where does the rest come from? The question is rhetorical; we know where it comes from. It comes, as it always has in the face of exploitation, from people organizing and forcing exploitative businesses to change their tune. It happened to the mills of Manchester, where children worked fourteen-hour days until the Factory Acts said they couldn't.¹⁶ It happened to the cigarette companies, who spent forty years insisting their product was harmless and ended with a $206 billion settlement and a warning on every pack.¹⁷ It happened to the rivers of America, which were so thoroughly used as sewers that one of them caught fire, and two years later twenty million people marched on the first Earth Day and Congress passed the Clean Water Act.¹⁸ It has to happen again. The Meta settlement is a first swing at it.

The trouble with regulating attention is that it is hard to say what it even is. Every past case involved a physical thing: a child, a lung, a river. Attention is psychological, and worse than that, it is the substrate of every concept we have. It is hard to look at through a lens because it is the material every lens is made of. William James said everyone knows what attention is, and nobody has managed to define it since.¹⁹

The only way to talk about a thing like that is in metaphor, and a few good ones are in circulation. Chris Hayes, in The Sirens' Call, treats attention as labor and suggests we borrow the tools of the labor movement. D. Graham Burnett treats it as something like oil, "fracked" out of the crevices of the mind by high-pressure content.²⁰ Both capture something real. I want to propose a third, chosen less for how it feels than for what it lets us borrow.

Treat attention as a natural resource and ask what kind. It is finite: there is only so much in a day. It is renewable: sleep gives it back. And it is diminished by overuse: the more of it is drawn, the less each unit is worth, and the drawing itself does damage. Psychologists have studied that last property under the name attention restoration theory, which found, roughly, that directed attention fatigues with use and recovers in environments that don't demand it.²¹ Nature, mostly.

Now put it next to the resources we already know how to regulate.

Oil and gold run out and stay gone. Sunlight can't be hurt. The one resource that renews itself, can't be stockpiled, can be over-drawn, and is damaged in the drawing is a river. The Colorado is drawn so hard it rarely reaches the sea anymore; its delta is a salt flat.²² The match isn't perfect; a river runs all the time and attention comes back in the morning. But on every axis that matters for regulation, the river is the nearest thing in nature to the thing in your head. And the machine that captures a river is a dam.

That is what algorithmic media is, in the aggregate: a dam. It harnesses the flow of human attention to generate something, in this case money, and in doing so it degrades the river and everything that depended on it. The water still goes through the turbines. But the fish don't run, the silt stops moving, the wetlands downstream dry up, and the people who lived on the river's cycles find the cycles gone. Nobody at the dam meant any of that. The dam does not know there are fish. It was built to turn flow into power, and it does.

The reason to prefer this metaphor is that we already know how to regulate a dam. We license them. We require an accounting of what happens downstream before one is built, not after. We mandate minimum flows, so that some part of the river has to be left to run on its own no matter how much power the operator wants. We build fish ladders. And when a dam does more harm than it is worth, we take it out. In 2024 the four dams on the Klamath came down, the largest dam removal in history, and salmon were back in the upper river within weeks.²³

None of those tools exist yet for attention. But you can see their shapes. A minimum flow is a cap that applies to everyone, not only the under-18s. A downstream accounting is a company having to show what its ranking does to sleep, to school, to the friend who stopped calling, before it ships the ranking. A fish ladder is a way out of the feed that the feed itself has to provide. And a dam removal is exactly what it sounds like.

This is where I am looking for hope. Not because the tools are ready; they aren't, and the metaphor may not survive contact with people who know rivers, or regulation, better than I do. But it is the first frame I have found that makes the problem look like one we have handled before, and that gives the people who might handle it something to hold. If it resonates with you, and especially if you know how to build the thing it points at, I would like to hear from you.