In November 2017 the founding president of Facebook stood at a public event and described, on the record, what he had helped build: a machine designed to answer the question "how do we consume as much of your time and conscious attention as possible?" He called its core mechanism "a social validation feedback loop", said it worked by "exploiting a vulnerability in human psychology" and added that the people who built it "understood this, consciously, and we did it anyway." No leak produced this. No whistleblower carried it out in a folder. The architect said it into a microphone, and the machine he described now mediates how several billion people encounter the world.

The canonical redpill line is Morpheus: "You take the blue pill, the story ends. You wake up in your bed and believe whatever you want to believe. You take the red pill, you stay in Wonderland, and I show you how deep the rabbit hole goes."

Start somewhere unserious, because the unserious place is doing serious work. Cinema produced zombie films at a trickle for seventy years. Then, in the early 2000s, the trickle became a flood: 28 Days Later in 2002, the Dawn of the Dead remake, the games, the parodies, and by 2010 The Walking Dead arriving as the most successful series in American cable history, premiering the same year Instagram launched. The scholarly literature on the zombie renaissance dates the turn to the early 2000s and generally files it under 9/11, contagion anxiety and institutional collapse, which is almost certainly right, and which is why this piece will not pretend the zombie boom is evidence for what follows. It is not evidence. It is a frame, and frames should be declared at the door rather than smuggled through it.

Here is the frame. A zombie, clinically, has no memory of the past and no conception of the future. It is entirely absorbed in the present moment. It is driven by a single appetite that can never be satisfied, and it converts everything it touches into more of itself.

Now hold that description against a machine whose builders have told us, on the record, what it wants: your attention, now, in a loop that feeds on its own output and recruits the people it captures into producing more of what captured them. The fit is not proof of anything. It is the reason the image will not leave once you have seen it.

The confession record

The case against the engagement economy does not rest on critics. It rests on confessions, and the confessions are unusually good.

Sean Parker's is quoted above, and the full passage is worth reading because of how little it leaves for an investigator to do. The dopamine hit, the social validation feedback loop, the deliberate exploitation of a known psychological vulnerability, the closing line about what it might be doing to children's brains: the design intent is stated by the designer, in public, with the names of colleagues attached.

Tristan Harris, formerly a design ethicist at Google, supplied the engineering detail in Senate testimony and talks across the following years. Pull-to-refresh works like a slot machine lever. Infinite scroll removes the stopping cues that tell a person they are finished, the way a bottomless glass would remove the cue to stop drinking. His phrase for the competitive dynamic deserves wider circulation than it gets: the race to the bottom of the brain stem. Calm, useful and respectful of your time loses the attention auction to whatever fires the older, faster circuitry underneath thought, so the bidding moves relentlessly downward, past reasoning, past reflection, toward alarm. Nobody chose fear as the product. The metric chose it, by watching billions of clicks and keeping whatever held the eyes.

Note what kind of claim this is. It is not a claim about how much damage the machine has done, which is genuinely contested territory addressed below. It is a claim about what the machine was built to do, and on that question the builders have already testified against themselves.

Presented, interrogated, absorbed

To see why design intent matters even where damage is hard to measure, you need a model of what stands between a person and the world. A workable one has three stages: reality as it is presented, interrogated and absorbed.

Everything you encounter arrives presented: framed, selected, worded by someone or something. Between presentation and absorption sits a checkpoint, the moment of asking what this is, who is saying it and whether it survives examination. What passes the checkpoint gets absorbed, and what is absorbed is what you become.

The checkpoint matters because frames are not decoration. Cognitive scientists George Lakoff and Mark Johnson argued in Metaphors We Live By that metaphor is the architecture of abstract thought itself, and the experimental literature has since shown frames steering conclusions directly. In a series of experiments by Paul Thibodeau and Lera Boroditsky, people given identical crime statistics reached different policy conclusions depending on a single word: describe crime as a beast preying on a city and they reach for punishment, describe it as a virus infecting a community and they reach for intervention. Same data, different frame, different politics. The frame did its work beneath the checkpoint, which is what frames are for.

So now describe the engagement machine precisely. It is a system that selects, from everything that could be presented to you, whatever is least likely to wait at the checkpoint. Moral-emotional language travels measurably further on these platforms: one widely cited study found each additional moral-emotional word raising a message's retransmission rate by around 17 per cent, an effect strongest inside ideological in-groups, which is worth saying plainly since the finding is often quoted without the caveat. Outrage, alarm and tribal threat are not the content of the machine. They are its selection pressure, because they are what gets absorbed without examination and absorption without examination is what the metric rewards.

This gives us a definition that will carry the rest of the piece. Call a system red when it is built to bypass the checkpoint, to be swallowed before it can be examined. Call it green when it is built to survive interrogation, when it does not mind waiting, when its sources are open and its incentives do not depend on your staying. The test is structural, not editorial. It does not ask what the content says or whether its politics are agreeable. Anything optimised for retention is red by definition, whoever runs it and however virtuous its content, because the optimisation target is absorption before scrutiny. A sourced argument that invites you to check it is green by construction, including when it is wrong.

What this argument is not allowed to claim

An argument about systems built to bypass interrogation has an obvious obligation: it must survive interrogation itself. So here is what the strong version of this case cannot honestly say.

It cannot say the machine has been proven to drive mass polarisation. The largest experiments ever run on the question, conducted inside Facebook and Instagram around the 2020 US election and published in Science, found that switching users to a chronological feed for three months produced no significant change in polarisation or political attitudes, though critics have argued the platform's emergency algorithm changes during the study period muddy the result. Earlier work by Boxell, Gentzkow and Shapiro found polarisation rising fastest among the oldest, least-online Americans, which a story about feeds cannot explain. The honest summary is that the harm-magnitude question is open.

It cannot recycle the popular statistic that YouTube radicalised half of the far right. The source usually mangled into that claim is a Bellingcat study of seventy-five fascist activists' own conversion stories, in which thirty-nine credited the internet broadly and YouTube was the single most-mentioned site. That is a real and disturbing finding about self-reported pathways. It is not a prevalence estimate, and pretending otherwise is exactly the checkpoint-bypassing move this piece is against.

What survives the cull is the part that was never in dispute, because the builders conceded it first: the design-intent layer. A machine engineered, by the public admission of its own architects, to exploit a vulnerability in human psychology for commercial gain is an indictment that needs no inflated damage estimate attached. We do not wait for a body count before objecting to a product designed to be swallowed unexamined. The design is the objection.

The re-greening

Once the problem is stated structurally, the two standard responses reveal themselves as the wrong shape.

The first is exit: smash the machine, delete everything, go offline, build the cabin. Exit is unavailable at the scale that matters. The machine now mediates employment, family, civic information and emergency communication and advising several billion people to log off is advising the tide to stop.

The second is awakening, which is worse because it has been captured. The vocabulary of seeing through the system, of red pills and sheep and the unplugged few, now belongs to the movement Bellingcat catalogued and its structure was always self-flattering: I see truly, you are asleep, my contempt for you is insight. Any critique of the attention economy that ends with the critic awake among sleepers has reproduced the machine's favourite content type, which is tribal superiority delivered as alarm.

There is a third position and the old metaphor everyone reaches for actually contains it, unused. In the source material the code rain was green but green never meant freedom. Green meant the system as it ran, ambient and indifferent, a world that did not care about you in particular. The early internet was green in exactly this sense. It was not liberating and nobody should pretend it was. It was indifferent: pages in chronological order, sites you went to rather than feeds that came for you, infrastructure with no opinion about whether you stayed. What changed was not that a neutral tool fell into bad hands. The system itself was re-engineered around retention and the colour of the code changed. The corruption has a date range and named engineers, several of whom have apologised.

"The Matrix is everywhere. It is all around us... It is the world that has been pulled over your eyes to blind you from the truth. That you are a slave, Neo. Like everyone else, you were born into bondage, born into a prison that you cannot smell or taste or touch. A prison for your mind."

The demand that follows is narrower than revolution and harder than awakening: re-greening. Make the systems indifferent again. Strip the retention incentive out of the infrastructure of presentation and let content compete at the checkpoint instead of being engineered around it.

This is not utopian, because pieces of it already exist in the most boring possible form, which is how you can tell it is real policy rather than a vibe. The EU's Digital Services Act already obliges the largest platforms to offer users a feed not based on profiling: a legally mandated green mode, sitting one toggle away, almost never surfaced by the platforms because surfacing it would cost retention. Auto-enrolment pensions correct a known cognitive bias structurally instead of lecturing individuals about it. Wales runs a statutory commissioner for future generations whose job is to make long-term consequences legally present in short-term decisions. None of this is exciting. Seat belts were not exciting. They were the admission, built into the machine, that humans predictably fail in known ways and that systems should absorb the failure rather than harvest it.

Who decides what green is? The definition above was built to answer this, because "restore the system to neutrality" has served as cover for plenty of people whose neutrality was somebody else's red. The test is not editorial and nobody's politics are consulted: a system is green when it carries no incentive to keep you there. Chronological order is green. An open archive is green. A sourced argument with its evidence hyperlinked is green even when you hate its conclusion and an engagement-ranked feed is red even when every item in it is true. The question is never whether the content deserves to be absorbed. The question is whether the system was built to let you ask.

The zombie stories, whatever drove their renaissance, all agree on one thing about survival: the survivors are not the strongest or the most awake. They are the ones who plan, who ration, who stay in the argument, who keep the checkpoint manned while the horde runs on appetite. Nobody gets to leave the world where the machine runs. The whole demand is that the code falling through it stop being aimed at the bottom of your brain stem and there is a word for that now. Re-greening!

Sources