Attention is the Enemy
Disempowering the Attention Economy
From Vol. 5 of the New Machine Cinema
The artist of the future will be surrounded by systems better at capturing his attention than his own judgment. Such a system will know, to the microsecond, what hooks retain attention and what choices are maximally rewarded. It is a cheat available to those with access, granted for now to those whose contributions serve it. While those without access risk being without means entirely; they are forced to shop in the attention economy like merchants, playing at a system that is only incidentally related to their actual pursuit.
You will want to escape the Borg just to have your own thing that hasn’t been digested, hammered and trampled on: illegibility in effect will become a rare and desired virtue.
Engaging the attention economy has become a prerequisite for legibility itself. By the time production reaches full abundance, competence stops being scarce. The artist is squeezed from both sides: the machine pressures the work toward precedent, the audience-selection system pressures it toward familiarity.
We solved AI slop books ago. Bad images are obvious, bad films disappear, and this self-corrects as novelty algorithms gear toward competence. What interests me now is the competence model. The successful work. The beautiful, polished work that travels easily through platforms and audiences.
That work is not slop in the ordinary sense. It is often technically excellent, carrying all the evidence by which contemporary culture catalogues seriousness. But its entire aesthetic life can depend on a series of recognitions that arrive before the work was conceived. The viewer knows the genre, the mood, the status, the intended sophistication. The work becomes superfluous. The need already manifested it.
This is attention slop.
The vocabulary of contemporary image-making reveals how completely this logic has entered the act of creation. Cinematic, moody, epic, cozy, prestige, eighties, anime, Wes Anderson, A24: these terms appear to describe aesthetics, but function more like switches. Pull one and a warehouse of associations arrives at once. Instant reach. Instant recognition.
Step back and what you are often seeing is a vast cultural payload compressed into something fleeting and instantaneous. An aesthetic hat trick.
The attention economy rewards this compression because it optimizes for attention alone. A trope becomes a mechanism for promising reward in advance. This is why the attention economy may be one of the worst systems of artistic selection ever devised. Not because people are stupid, but because we are measuring people in their least discerning state and treating that indulgence as a measure of deep preference. The same person who will sit in a theater for two hours, return to a difficult novel for a week, or stand before a painting until its opacity gives way may scroll past an equally difficult work in bed, because the conditions of encounter have made difficulty indistinguishable from inconvenience.
Aesthetic hooks were never built to last. Today they are treated as the whole of the matter.
Creators who absorb those metrics as verdicts deserve release, and would take offense if you suggested attention says nothing about lasting art. Consider Moby-Dick, which could not have cracked the attention economy of its own time. And yet great art can be immediate, popular, seductive, legible. Hitchcock made popular entertainment that doubles as a structural treatise on cinema itself. Disney’s animated classics are so universal, so magical, that they feel like a separate tradition from cinema entirely, and still do not receive the canonical weight they deserve, standing as both enduring and unfashionable.
Occasionally we bow to sheer craft. The problem begins when craft is mistaken for art’s measure rather than simply one mode of encounter among others. What we need is a category that says: this is not even a film. This is an expression of the attention economy.
Eggleston said he photographed the Red Ceiling at war with the obvious. That notion must return to us: a rebellion against immediate algorithmic pleasure. It was not a photograph as anyone then understood the term, and it became canonical precisely for going against the grain. What happens to the Red Ceiling today? To Duchamp’s urinal in the museum? The coincidence between attention and value has been repeated so often it has started to look like causation.
THE DOUBLE AVERAGING MACHINE
Generative AI makes this old problem stranger, because the pressure toward consensus now operates on both sides of the artwork.
A camera has no opinion about what a photograph should look like. It records what is placed before it, according to the properties of its apparatus. It does not hold a statistical memory of a million earlier sunsets and quietly suggest which one this sunset ought to resemble. A generative model does. It is as if the camera has begun to opine.
It has seen images, organized relationships among them, and acquired an immense internal sense of probability. Ask for a romantic sunset, a frightening monster, an elegant woman, a science fiction city, and the machine produces the consensus answer, because the consensus answer has more statistical support behind it. It can exert a genuinely seductive force. It is cleaner than what you imagined, more evocative, more immediately intelligible. One begins to mistake the machine’s fluency for one’s own intention.
From there, the finished work enters a second system with its own statistical intelligence. The recommendation engine knows, to the microsecond, what has historically held attention and when. It exerts pressure toward the kinds of images most likely to circulate.
System one produces statistical convergence on the image. System two applies the same convergence to the audience. The two systems have different purposes until their pressures converge into a single vector: the familiarity machine meets the legibility hunger, and together they become a double averaging machine.
Variability is the first casualty. It cannot survive contact with two statistics at once. By the time a work reaches you, it has passed through thousands of filters built to weed out variability and maximize predictability. What reaches your senses is the most vibrant, flashy, and hollow facsimile of existence, delivered without remorse.
Worse, there is no culprit. Neither machine is at fault. Each is simply performing its assigned task: predictability, legibility. No one has to tell the artist to conform. He encounters a world in which convention arrives beautifully rendered, and convention is rewarded with superior circulation. A human process has been mechanized into algorithmic shorthand.
The artist who wants to make something singular receives almost every available signal telling him he is doing something wrong. Low views, weak retention, poor recommendation, a model constantly offering prettier and more conventional solutions, other creators visibly succeeding through recognizable tropes. The feedback loop around the future artist has turned actively hostile to originality, in a way it never was before.
Such a system existed before technology arrived to scale it to apocalyptic, almost satirical heights. Perhaps there is no nuance possible when playing to an audience of billions. Artists have always faced hostile feedback. Patrons wanted one thing, studios another, publishers another, audiences another. But those pressures were historically external to the instrument itself, not built into the machine that makes the work. I suspect this undersells the masses; there is more tolerance for the strange among people than our current system allows for.
A painter could resent his patron. A filmmaker could resent his studio. The generative artist recoils against consensus itself, and that is a different order of opponent. A patron is a person. Consensus, mechanized, is closer to a hive.
Artists today occupy an impossible position. The better the machine becomes, the higher the pressure, the more seductive the reward. A crude machine could be resisted easily. A brilliant one exacts real consequences, and at its worst it can turn the hungry masses against anyone who throws a wrench in the works. This is dangerous precisely because the conventional check no longer applies. The machine can make the wrong decision exquisitely well. There is no one positioned to tell it that it’s wrong.
THE LOST CARTOON
The entire internet banded together to find a cartoon in someones old home movie, to no available. It was a world-wide search. Watching the manifestation of collectivized memetic cooperation in action, I felt disturbed by the sheer overwhelming power on display, all to catalogue this missing cartoon.
Eventually the cartoon was found. It was called The Soulmates: The Gift of Light.
A flood of reviews went to its IMDB page, all a parade of pure joy that the cartoon has now been found and that the reviewer is part of this in-joke.
So disturbed, I became the sole dissenting voice. In my review, I said not all mysteries should be solved, there should be pockets the hivemind simply should not get to, and that this paints a frightening picture of the future. Within 24 hours, the review was deleted due to mass reporting.
THE MACHINE THAT JUDGES AND THE MACHINE THAT WATCHES
I once imagined machine intelligence as the great escape from the arrogance of mass taste. I never conceived that it would simply amplify mass taste to the point of nauseating dilution.
The argument was that the one-size-fits-all structure of culture could eventually be broken by an intelligence capable of understanding individuals at extraordinary resolution. Instead of one critic, one studio, one canon, a sufficiently advanced machine might recognize that a forgotten film could matter enormously to one person and almost nothing to another. It could, in principle, create a new sophisticated canon that is personalized for every single person in a way that no mass institution ever could. I still believe that machine is possible. The timescale I imagined however was utopian.
My mistake was assuming this was the machine being built.
The machine I imagined was evaluative. It would need to understand not merely what a person did, but why: the alchemy by which art works on someone. It would have to distinguish pleasure from compulsion, familiarity from significance. It would need to understand that someone who dislikes a film might think about it for years. It would need to recognize that the same viewer needs different things at different moments, and that the work most likely to change them is not necessarily the work they think they want.
The machine that actually governs cultural attention does not concern itself with any of this. Call it a pure mechanized utility of hate, not in the literal sense, but as the precise inverse of sensitivity—to barrage our wants so bluntly it approaches as an act of unthinking, unfeeling hate. It hates unpredictability. It hates patience and our inattention. Clicking is not liking. Finishing is not valuing. Rewatching is complacency.
We automated attention before we automated judgment.
My earlier theory imagined an intelligence sophisticated enough to solve the one-size-fits-all problem in a culture that squeezes us all like dough into one single pretzel so that it can sprinkle the salt uniformly.
I mistakenly believed the movement from behavioral prediction to genuine evaluative understanding would arrive faster than it has. While I remain optimistic, it is clear we are in an intermediary period: generative production has met machine intelligence without values attached to it. The double averaging machine is making us more like it, not the reverse.
People follow the road to the lost cartoon. You can hate them for it, but they walk willfully alongside the machine, confused over who is steering who.
This period could be brief. It could also last a very long time, and its stakes compound the longer it runs, because the behavioral machine is not merely sorting culture. It is producing the culture that later, wiser systems will inherit and learn from.
A part of me wonders if we are a lost generation to the seeds of artificial intelligence, that we won’t get to see its full flight as it pursues the relenless race to the bottom for attention.
The cream rises to the bottom.
If creators spend decades optimizing images, stories, music, and performance around retention and repetition, the eventual evaluative machine inherits a historical record already bent by behavioral optimization. It will look back and find a great deal we made needlessly difficult for it. By then we may be too machine to turn back.
I had my own mania. After Lucky Stars, I became hungry like never before to cultivate new skills. I learned violin. I took up drawing again for the first time in thirteen years after I abandoned the Pink Latitudes graphic novel—my follow-up to Aimy Micry.
At the time I couldn’t resolve the pain that I could count the people who read it on one finger verses the immense effort that it took. My drawing track was a casualty of this earlier iteration of the attention economy. So why did the mania take hold when the attention economy became even more oppressive?
To be clear, it was not born out of a sense of duty toward the past. My thesis was to study the impact of AI on skill-learning. I had never touched a violin. Could AI turn me into a player? But more, I had lived in a state of limitlessness for three years. I wanted to fail again.
THE SECOND HOLLYWOOD
The early dream of generative cinema promised that the industrial barriers to filmmaking would collapse, and that the aesthetics emerging outside those barriers would build a new artistic system. A single person could make the impossible happen.
When that freedom arrived, a great number of people used it to recreate exactly what the old system had already given them.
Our reaction against Hollywood was to build a second, diluted Hollywood.
Generative media are especially vulnerable to this because of the double averaging machine. Aesthetic shorthand platforms itself with no resistance, and the end result is the same as always.
You learn fast in the arts, there is no such thing as niche.
The pipe feeds into itself: precedent produces resemblance, resemblance attracts attention, attention validates precedent, and the successful result returns as reference for the next creator.
This is why there was no independent filmmaking revolution. The machine, both philosophically and in practice, cannot allow for anything but consensus average.
This is more insidious than censorship, because it makes consensus feel like mathematical law.
THE GOD’S EYE
In the future, everyone will be the star of his own algorithm, and none of them will ever meet.
The machine offers something like a god’s eye. Every architecture, costume, climate, genre, body, lens, and impossible event can be summoned from the same interface. The artist can move from medieval Japan to Mars to a New York apartment in 1974 without leaving the chair.
But no two gods share in each other’s powers. As I wrote in the titled essay, god’s eye bring the easiest shorthand toward mass affection, I call this unearned transcendence.
It is Hollywood’s most predictable shorthand, is hijacking the instinct toward meaning to catch more box office or buzz. Incidentally, a film exists in the formula, but seldom does it consist of more than a succession of shorthands and cliches.
While AI utilizes it as craft. At the level of access, it looks like absolute freedom. But when the artist can stand nowhere and see everything, transcendence becomes the same as loss.
We live out, through simulation, every story ever written about weary, disgruntled gods. Who have everything but have lost all purpose. Who are ill of the machine they have created.
This is where the AI and anti-AI positions should stop arguing past each other. The strongest anti-AI claim was that art should contain evidence of a human being.
I held the answer in the essay, and hold it here: a more specific humanity.
The model has the answers. It knows every shorthand. It does not know you.
There I call for a revolution of personal and specific films. Vulnerability becomes an asset. Sensitivity becomes a weapon. Attention is the enemy because it averages your experience which is specific. Attention is the enemy because it categorize and processed and cheapens your precious vision.
To the point that the double averaging machine, stale and starved for oil, may come to need you. It has exhausted everything else.
It will come for you last.
There we will learn biography is the closest thing an artist possesses to an out-of-distribution input: the personal is the only source material whose exact distribution is one.
That creates its own paradox, because the personal is exceptionally difficult for the attention economy to read. Something with no obvious precedent has fewer shorthands available to achieve in the mathetmatics of mass interest.
Without shorthand, recognition takes longer, and a system optimized for immediate legibility interprets any alternative as failure. While requiring the oil else the machine will starve of novelty. Only now the novelty will not be extracted with the same suspicion as any discerning, contrary voice today, fed into the double averaging machine.
I will note the tendency of these systems to canonize the outsiders who built inside a pocket of purity away from them, rewarding the specific honor of being rejected from these systems, by the very ones doing so.
The personal is punished for exactly the reason it matters, while it is needed as ever.
FIVE RESPONSES
There is no single solution to the machine condition. It is a wall every artist will face in our future, a threshold every artist must cross, it is the defining feature of our coming age: how do we opt out? I have come up with five responses to machine despair.
ONE: EXHAUST THE AVERAGE
Use machine augmentation to fulfill every trope as shorthand, then select for the most subversive, unusual, stunning outcome, and let that selection become the art. Poison the machine with itself.
If the model is an extraordinary trope-fulfillment engine, let it fulfill the trope. The artist can map the familiar until it creates the exact model for what to select against. Art becomes art again. You will know because the algorithm will ignore it, until it must be forced to platform its own subversion.
To infiltrate the machine consciousness is to fire a bullseye into the human mind. A well timed strike can peel the layer from itself.
Consciousness returns from machine slumber. Even if only briefly. Even at a cost.
The malformed gesture becomes worth more than the flawless one. Everyone else can use the machine to arrive at the average. Not you.
TWO: ACCEPT THE TWENTY VIEWS
Reject the economy outright. Because attention is the enemy, take the bans. Make personal work and wear the twenty views as a badge of pride. Competence will be algorithmically available to everyone, so the raw, unseen work becomes what survives at the far end of the filter.
You are algorithmically untenable and there is no greater compliment in a system of unchecked machine totality. It means you weren’t average enough. Didn’t go down easy. Didn’t appeal to everyone every how. You were not a conglomorate wearing a logo and waving a flag. A headline or a byline. A meme.
What horror, and what joy, in whatever gets algorithmically maximized as the most unplatformable thing there is.
It will exist in a sea of slop that went unplatformed simply for being dull. Not yours. It will be conspicious to the point of unpopularity.
Think of a world of Henry Dargers and Vivian Maiers, except the crime is no longer private obscurity. It is the dignity of never being known by the algorithmic godhead, in any way. Consider the machine restlessly seeks them out already, recognizing a distaste for itself, its tendencies and selections, that true worth doesn’t come through them but those who managed to escape them.
People wonder about the fish which escaped the tank.
And when they come to you to find it, make sure you blow the bridges. In the afformentioned cases, the whole appeal was they were dead.
THREE: REMOVE FAILURE’S LEVERAGE
Automate so aggressively that the difference between a flop and a hit stops mattering. You are already a few works away from the last one by the time you have your reception.
This is where post-scarcity offers the essay’s only real hope. A filmmaker spends three years on a film. It fails. It cost him five years, the investors, the distributors. Post-scarcity weakens that leverage.
This was ironed at the very start of the AI cinema revolution but must be learned and relearned because the attention economy’s indifference will always seem like an insult.
If production becomes fast and inexpensive, the artist no longer has to play. The failed film no longer has to contain the next three years of his life. Opening weekend loses its metaphysical weight. You will stand in parallel with the strongest that there is. Have the same opportunities as them. The same careers. The same bodies of work. The same visions and executions and satisfactions and disappointments. What truly matters other than that?
And while moral high ground arguments seem like a weak consolation prize in rigid scarcity environments, we must learn to evolve from our previous state, live our ideals, because this sort of virtue is exceedingly rare, and there so freely offered in post-scarcity.
FOUR: BUILD YOUR OWN CANON
Watch the films no one is discussing. Study the creators who opted out. Seek them out. Save the images that will become scarce.
Opting out is no longer a moral failing. No artist invents a private aesthetic from nothing, and if the surrounding cultural system is increasingly optimized around prior behavior, passive consumption becomes a form of aesthetic training.
Your private curation could be worth more than the entire canon around it.
Build your own canon. In a culture where production is cheap, choosing what enters your mind may become as important as choosing what leaves it.
Because attention is the enemy, you will have known, seen and filed away every film the double averaging machine has sent you before you’ve watched a single frame. There is not a thing the week’s latest horror craze, blockbuster epic, or corporate indie flooding your feed can show you.
FIVE: MECHANIZE YOUR HATE
Exploit the system consciously for maximum profit. Profit from the attention economy without shame. Learn the system perfectly.
If the attention economy rewards familiarity, interruption, hooks, escalation, and compulsion, then trade in it without hesitation.
There is money in understanding those mechanisms better than the people who merely resent them. Vertical drama is one of the clearest current examples of attention laid bare; if the attention economy already reveals what strange parts of our nature survive its filters, the verticals add still more filters upon filters on top of them, creating even stranger dilutions of our subconscious manifested there before us. It is in the purest way the manifestation of what I call Grotesque Pleasures, where algorithmic perfection looks to regular eyes as deformed mutations.
There is something almost clarifying about the shamelessness of optimg in. It removes the pretense that attention and artistic importance are the same thing.
In post-scarcity, we can partake without apology while spending our lives to fight it with our every breath.
Mechanize your hate. Learn what the machine wants. Produce it at industrial speed. Harvest the consensus algorithm the way a nineteenth-century farmer harvests a field.
Remember, obscurity, whatever it feels like inside the monster we’ve built, is not a permanent condition. We can step in and out of it at will.
THE DECISION OF THE FUTURE
I no longer think the central question of AI culture is whether one should use AI. That question will eventually sound almost quaint.
The more consequential question is how much of yourself you are willing to give to the machines.
The machine will always ask us to pave the road to the lost cartoon. Bring it inside the system until nothing remains beyond its reach. But we do not always have to follow everywhere.
Everyone must decide this for themselves. It is the most personal question of the coming age: which boundary is the one too far to cross?
Some will live almost entirely inside algorithmic systems, because the convenience is enormous and the results allow them to participate in a simulacra of a society where basic consumption honors them with attention.
Others will use machine systems aggressively for execution while growing almost paranoid about protecting selection.
Some will cultivate tiny cultural worlds. Some will move back and forth between algorithmic and private systems depending on the task.
Some will build machines specifically to fight the tendencies of other machines.
The decision was never whether to opt out. It is how far.
This is why small curation may become more consequential than it presently appears. The tiny audience, the personal archive, the neglected film, the work made just because: these become the oil, the counterweight to a machine of such enormous intelligence that it has concentrated around the mathematics of our wants.
The future artist may have to work without the reassurance that surrounded older definitions of success. There may be no institutional confirmation, no discovery mechanism, no algorithmic evidence that the decision was correct. There may be twenty views. The same machine spending a lifetime dismissing them may lionize them the moment they are dead, and not one second sooner.
The final liberation promised by post-scarcity was never that everyone would finally pay attention to us. It was that their attention would no longer determine whether we could continue.
That, in itself, reads as a moral high ground consolation in our scarcity condition. But scarcity and post-scarcity exist as oil and water. They cannot co-exist. The question is which you want to live in.
