
Is AI Peak Reification?
Benedikt Heudorfer
November 19, 2026
“The position that an epoch occupies in the historical process can be determined more strikingly from an analysis of its inconspicuous surface-level expressions than from that epoch's judgments about itself. Since these judgments are expressions of the tendencies of a particular era, they do not offer conclusive testimony about its overall constitution. The surface-level expressions, however, by virtue of their unconscious nature, provide unmediated access to the fundamental substance of the state of things. Conversely, knowledge of this state of things depends on the interpretation of these surface-level expressions. The fundamental substance of an epoch and its unheeded impulses illuminate each other reciprocally.”
Siegfried Kracauer – The Mass Ornament
FEVER DREAMS
‘We made the rocks think,’ is what they scream, hysterically, while venture capitalists throw massive amounts of money at AI companies on the phantasmagorical promise to replace all human labor. The class of workers who build the AI systems — programmers — are the first in line to be replaced. And further down the road, everyone’s job is at stake. There is panic at the prospect of being replaced by something trans-human. Gone are the days of Utopian Star Trekkism. The internet is filled with wild fever dreams about the future that is AI. What is our purpose in this new world, run by AI? Will we survive as a species, once we handed over all our skills to AI, rendering us useless, ready to be annihilated by sentient AI agencies? Of course, this will all happen sooner than we think. People just don’t realize how soon and how bad it will be. After all, humans really are bad at interpolating, and the curve really does go up, up, up – exponentially!
But is all of this reality? Or is it fantasies? An ideology of capitalism that necessarily projects outwards what ultimately are reflections about ourselves which we can’t consciously realize?
Chris Cutrone once drew an interesting analogy with extraterrestrials. Sentient beings may or may not exist outside our solar systems, but either way, people theorize endlessly about eventualities and possibilities: How will they turn out to be, once we meet them? Will they be hostile or friendly? Will we meet our destiny or our doom? These reflections are independent from any reality of extraterrestrials. We literally know nothing about them – if they exist at all. But, in debating our imagery of extraterrestrials, we reflect society, its value(s), and how it will be judged by someone on the outside. We debate our own judgement day. In earlier days, these reflections about society would have been based on discussions about God (and for many they still are). Today, it’s AI which we use to debate our reckoning.
Much like extraterrestrials, images about AI are not something outside of us, they are about us. The anthropomorphism of AI says as much. We could talk about parameter estimation of regression models, or we could talk about machines ‘learning.’ We could talk about a manifold of basic, granular equations from linear algebra interacting to approximate nonlinear functions of arbitrary complexity, or we could talk about ‘neural networks.’ We could talk about large language models (LLMs) or we could talk about ‘artificial intelligence.’ AI developers could train the models on any other query-response scheme, instead unanimously opt to fine-tune LLMs to mimic a human assistant. Anthropomorphism is the attribution of human qualities, emotions, intentions, and behaviors to non-human beings, objects, or abstract concepts. It is baked deep into the discourses on AI, and even model design itself.
This is no coincidence. AI comes upon us at a specific time and in a specific manner. Never before in human history has the body of knowledge been bigger, never has it been more fragmented. For quite some time now, it has been impossible for a single individual to penetrate the full breadth of human knowledge. You can only master so many scientific domains until the individual fragments of apparent knowledge start to contradict each other, rendering true understanding impossible. And right at this point in time, we place our faith in the ‘deep learning’ of an ‘artificial intelligence’ outside of us, outside of our control, to reconcile humanity with the knowledge it has produced. This represents the necessary unconscious recognition of our own inability to learn, and to progress our collective intelligence and consciousness in a way not alien to us. AI seems to possess all the power we produce but feel unable to take for ourselves — and must thus be placed outside of us. It is the intention of this piece to make these thoughts accessible.
THE APPROPRIATION OF CREATIVE POWER
Karl Marx’s critique and Adam Smith’s moral philosophy are both based on the observation that participation in modern bourgeois society is based on labor. The idea that everybody who works should be allowed to participate in society shattered the old societies based on feudal benefices and made way for the bourgeois revolution. However, the resulting society, built on this idea, implied that people who can’t or don’t work are not allowed to participate in society. Labor under the conditions of industrial capitalism ultimately created a crisis of this bourgeois ideal. Labor became a commodity alongside other goods and is exchanged on competitive markets. To contribute and participate in society through work is no longer a choice based on individual capabilities, but a question of demand for labor. Insofar people can’t or don’t work but are allowed to participate, it is only by the choice of those who work and share their resources (if available) through the welfare state or charitable institutions.
But this does not fully account for the problem of capitalism that Marx scrutinized. As Martin Nicolaus points out in his essay The Unknown Marx, the key aspect about labor as a commodity in modern society is not the exchange of labor for money itself. Capitalism is not capitalism because injustices happen by underpaying staff. The point is that in purchasing “labour as value-creating activity,” capital “purchases control over this creative power,” as Nicolaus puts it, and allows it to produce surplus value in the downstream production of goods. That is the true appropriation, the capitalist exploitation in a Marxian sense: the appropriation of the workers’ creative power, who is forced to sell it due to the conditions of the society he lives in. But as Marx emphasizes in a part of the Grundrisse known as the fragment on machines, the effects of this appropriation really only fall into place in the materialization of labors’ creative power in the form of machines:
“As long as the means of labour remains a means of labour in the proper sense of the term, such as it is directly, historically, adopted by capital and included in its realization process, it undergoes a merely formal modification [...] But, once adopted into the production process of capital, the means of labor passes through different metamorphoses, whose culmination is the machine, or rather, an automatic system of machinery [...], set in motion by an automaton, a moving power that moves itself; this automaton consisting of numerous mechanical and intellectual organs, so that the workers themselves are cast merely as its conscious linkages.”
In the process, it is not capital but the workers themselves who want to save time, and voluntarily seek to transform their work. They automate tasks to make them easier, and thereby improve the machinery. But the better the machinery becomes in exercising labor’s creative power, the more it renders the workers unimportant and ultimately replaceable. In the end, all you need is someone who pulls the machines’ levers, and not a series of skilled craftsmen. As a result, work becomes less valuable on the market, at the expense of the increased creative power of the machinery. That is the antagonism defining capitalism as a crisis: capital is appropriating labor with the purpose of replacing labor, calling into question the worker's participation in society, contradicting society’s claim to be built on those who work — on labor.
AI models reveal themselves as the most recent, crystallized form of this process. They are developed and run by tech companies, fueled by (venture) capital, based on a very specific promise — to replace labor. They are improved continuously, not only by AI architects and tens of thousands of clickworkers who are getting paid to fine-tune the models, but also by the workers who are using the models to ease their own work and to increase their productivity, to stay afloat in their ongoing competition with other workers. The prompts and responses of these users are channeled back via reinforcement learning to further fine-tune the model. It is the workers who improve the machinery. Anthropic’s Claude is the best model for coding because it is the model most used by coders to write code.
But not only the creative power purchased off the people employed to build and improve the model (click workers, AI architects etc.) is appropriated. AI is also kick-started by extracting (‘learning’) the essence of creative power previously hidden in products of previously applied labor (‘data’) that are freely available (or stolen) on the Internet. And then companies and workers pay AI companies for the very usage of the models that further appropriate their creative power. In other words, the improvement of the machinery is monetized systematically, and workers appear as consumers of their own appropriation. This raises the question: does this qualify as a qualitatively new dimension of the capitalist mode of production?
PROLETARIANIZATION OF WHITE COLLAR WORK
The stated goal of AI development is Artificial General Intelligence (AGI). Casting all mystic discussions about the essence of intelligence aside, the pragmatic definition of AGI in the OpenAI Charter is most telling: AGI is an AI that can technically do most of the work that humans can do. ‘Technically’ because deploying AGI will still come at a cost. Whether it will actually replace human work is dependent on its price and the degree of devaluation of human labor, as Chris Cutrone summed up in Robots and Sweatshops. But labor devaluation does not only come in the form of declining wages and lost jobs, it also comes in the form of an increase in productivity. People work harder (provide more output for the same amount of work) for less money because the value of their work is decreasing.
In one word, the aim of AGI is to proletarianize white-collar labor. [1] The part of the workforce that thought they were protected from what happened to manual labor in the previous two centuries, is seeing their work devalued, because intellectual and cognitive tasks are now exercised by machinery. Their jobs might not disappear, but anyone will be able to do them without highly specialized skills, which previously enabled demanding high wages and good working conditions. As expensive skills are outsourced to the machinery, white-collar workers, too, become mere appendages of the machines, pulling their levers. We are all becoming prompt monkeys. Gone are the days where middle class kids were able to convince themselves that they are special, and that there is a safe path ahead. Once again we might see that “all that is solid melts into air, all that is holy is profaned, and man is at last compelled to face with sober senses his real conditions of life,” as Marx wrote in the Communist Manifesto almost 180 years ago. None of this is new, it’s just capitalism all over again.
REIFICATION
There is a necessity to the fact that people refuse to realize the “real conditions of life”, but immerse themselves in AI fantasies and anthropomorphisms.
As Lukács recapitulated in Reification and the Consciousness of the Proletariat, in capitalism, the commodity form became the universal category of social existence. It permeates all aspects of life, reshaping society according to its principles. In tendency, all social relations between people assume the form of commodity relations: relations between things. Workers (people) aren’t full masters of their product, but are merely part of an automated, mechanised system beyond their control, conditioning their actions as if from outside. Their position becomes passive and contemplative. They handle merely fragmented production subprocesses whose connections are determined through calculation and rationalization dictated by the needs of capital. Their social bonds are broken up, isolating and atomizing workers, replacing immediate connection in a previously integrally social labor process with a labor process that is almost exclusively mediated by the fragmented production dictated by the commodity form.
Likewise, when using AI, we are simultaneously becoming masters of, as well as powerless against the production process. We can unleash (approximations of) all human knowledge that ever existed, but it is impossible to grasp the full production process that the output of the model assumes – we type a prompt and gaze at the response, but there is no way in telling exactly how it came to be. As Georg Lukács put it in Reification and the Consciousness of the Proletariat: “[N]either objectively nor in his relation to his work does man appear as the authentic master of the process […]. He finds it already pre-existing and self-sufficient, it functions independently of him and he has to conform to its laws whether he likes it or not.“ In becoming prompt monkeys, we become “less and less active and more and more contemplative” with respect to the true production process. And so, as developers adopt AI in coding, they increasingly don't understand the code anymore because they didn't write it. They de-skill themselves, becoming increasingly incapable of handling and maintaining codebases, aggravating their dependency on AI for coding. They lose their grasp on the code (and its quality) because of the sheer volume being generated, and become distant from the codebase as a result of it. Relying on AI coding agents creates apathy towards the work and the software itself.
Work is increasingly applied at the interface between us and the model, or between model and model, removing us from the process of production in the knowledge economy, veiling its societal character. Knowledge production through AI turns workers into what Lukács describes for the conditions of industrial capitalism as “isolated abstract atoms whose work no longer brings them together directly and organically; it becomes mediated to an increasing extent exclusively by the abstract laws of the mechanism which imprisons them.” But this isolation and atomization of the individual is only apparent. It is mirroring the fact that “for the first time in history [...] the whole of society is subjected, or tends to be subjected, to a unified economic process, [...] determined by unified laws.” Individuals appear isolated, but are in fact more entangled in society than ever before. And this illusion of atomization is necessary, because “the immediate, practical as well as intellectual confrontation of the individual with society [...] could only take place in the form of rational and isolated acts of exchange between isolated commodity owners.” In other words, we are inclined to misinterpret AI because it reflects our immediate experience interacting with it as something external removed from us. Thus, we don’t see it for what it really is: a social process. We only see the machinery as the thing looming over us, defining the seemingly natural course of events that dictates our action and our reality: a super-human intelligence transcending us.
This process of alienation is what Lukács calls reification, and what Marx refers to as commodity fetishism in Das Kapital:
“A commodity is therefore a mysterious thing, simply because in it the social character of men’s labour appears to them as an objective character stamped upon the product of that labour; because the relation of the producers to the sum total of their own labour is presented to them as a social relation, existing not between themselves, but between the products of their labour. This is the reason why the products of labour become commodities, social things whose qualities are at the same time perceptible and imperceptible by the senses. [...] It is only a definite social relation between men that assumes, in their eyes, the fantastic form of a relation between things.”
In other words, products as commodities appear to interact with each other as if they were social beings. And it is important to remember that in our society, everything is a commodity, including AI. So when AI increasingly interacts with AI without direct human intermediaries, they are reified forms of social relations interacting seemingly outside of us, whereas in reality, they are exclusively defined and driven by human (bourgeois) social relations. LLMs constantly need data (commodities) produced by humans for humans, both during training as well as during their usage (reinforcement learning), to arrive at their established performance. AI is not something outside of us, AI is us: Materialized bourgeois social relations structurally mediated by the commodity form. Only in realizing this universal connectedness in the commodity form can the entirety of social existence be understood in its unadulterated essence. And it is within this context that the nature of the necessity of AI phantasmagoria becomes clear as well: it is reified consciousness. As Theodor W. Adorno put it in his Marginalia to Theory and Praxis, “individuals cope with their own disintegration, with their private paranoia, by integrating themselves into the collective delusion, the collective paranoia.”
THE CRISIS OF REASON
‘Artificial intelligence’ is a tautology because intelligence is by definition artificial, abstract: it is social. There are smart individuals, but knowledge and intelligence manifest themselves in society. In the 12,000 years since the neolithic revolution, human genetics has changed remarkably little. But in the same time span, we went from domesticating plants and livestock to neural networks and flights to Mars. Evolution does not explain what is grounded in an aggregate intelligence that accumulates historically in the social fabric.
In Jean-Jacques Rousseau’s On the Social Contract, he locates a general will in-between people when you “take away from these [individual] wills the pluses and minuses that cancel one another, and the general will remains as the sum of the differences.” Adam Smith attests in The Wealth of Nations an invisible hand when, “by pursuing his own interest, [the individual] frequently promotes that of the society more effectually than when he really intends to promote it.” Social intelligence, Rousseau’s general will and Smith’s invisible hand all point in the same direction: Reason is located in society, not the individual.
The society Rousseau and Smith talk about is the modern bourgeois society we still live in. However, bourgeois society changed profoundly since Rousseau’s and Smith’s time. With the onset of the Industrial Revolution, it entered the crisis described by Marx. And with it, Reason has regressed. The general will of the people is disregarded. The collective actions of the individuals on the market do not benefit all. And our knowledge lies scattered before us, unintelligible to ourselves, consolidated only by a form of collective social intelligence placed outside our control, preventing us, as society, from assuming sovereignty over it.
Under the condition of industrial production, bourgeois society in its totality is organized according to the commodity form, and the fragmented production process gives birth to reification. This comprises not just the relations of industrial production, but all creative output of humans, including that of our mind – Marx’ notion of production is total. Science and the knowledge economy in general today exercise only what Horkheimer calls instrumental reason, and not Reason. The dogma of calculability and rationalization dictated by capital produces a compulsion to fixate phenomena as what G.W.F. Hegel calls “terms of the understanding” (Verstandesbestimmungen). But fixed moments of immediate reality can only be properly understood when grasped dialectically, allowing us to sublate the illusion (Schein), revealing actuality.
In other words: facts and laws violate reality. But given its mode of production, science today can only fixate truths in the form of facts and laws in mathematized, calculable systems. This is causal to the fragmentation of scientific knowledge that renders reconciliation of those fixated truths of different scientific systems impossible. As Lukács puts it, science,
“the more highly developed it becomes and the more scientific, the more it will become a formally closed system of partial laws. It will then find that the world lying beyond its confines, and in particular [...] its own concrete underlying reality lies, methodologically and in principle, beyond its grasp. [...] It is the very success with which [science] is totally rationalised and transformed into an abstract and mathematically orientated system of formal ’laws’ that creates the methodological barrier to understanding [...] the qualitative existence of the ’things.’”
In this context, the ‘reified maturity’ of a knowledge domain directly translates into its predictability. Machine learning, deep learning, and LLMs all rely on data, which is quantified reality. These models then try to approximate reality by optimizing for a minimal mathematical distance between its prediction and the data. The better a data domain is ‘well behaved’, as the terminology goes, the better a model can learn. In other words, the better a certain domain has been fragmented, quantified, mathematized – reified – the better a neural network can approximate reality. But only as expressed by this reified systematization, not reality itself! That’s why system prompts as well as skill and rule files etc. need to be stitched on top of a necessarily incomplete model barely working as such.
Accordingly, AI models are doing specifically well in domains that are already completely navigable with formal logic, such as coding, engineering, or mathematics. For the same reason, they are consistently wrong in inherently indeterminable domains like text, where hallucinations are prevalent. The reason they replace human writing anyhow lies in the degree to which the ability of the readers regressed to spot the reasonable. All the models need to do is to sound convincing, which is precisely what they are trained for, as OpenAI openly admitted in Why Language Models Hallucinate. We can’t tell the difference anymore, not because the models are starting to think like a true human, but because we stopped to do so. As David Faes once said, mechanisms called AI can approximate our knowledge and our thinking because we have reduced our reasoning to something mechanistic. We even conceptualize brains as biochemical machines, as few-shot learning machines as Yann LeCun likes to put it, thinking that by doing so we have understood something about human intelligence. All of this is reification working in both ways: We can design machines that can replace us only because we have become things replaceable by machines.
BARBARISM OR SOCIALISM
The feeling of ‘being replaced’ already encapsulates the essence of what our society – bourgeois society – is based on. When a human says that ‘we will all be replaced by AI,’ then his unconscious identification of ‘being human’ with the reified forms of work and thinking he exercises is total. The fear that AI could destroy society is only possible with the unconscious acceptance that bourgeois society, and participation in this society, is exclusively built on labor – and that this society is in crisis: capital seeking to replace labor, bourgeois society undermining its own mode of existence; the capitalist contradiction. In our disavowal of this true condition of society, it is pushed into the unconscious. Thus, phantasmagorical phantasies haunt us. The fantastic promise of AGI is nothing short of the total overcoming of the dependence of capital on human labor. In that, it is a true but unconscious reflection of the actual. Because the overcoming of the dependency of capital on human labor by capital would in fact be the destruction of bourgeois society: consummated barbarism.
On the other side, AI promises access to the full scale of collective intelligence. We only have to utter the words – type the prompt – and the machine does everything we want it to do. If not now, then surely one day, right? This is the Utopian promise of the ability to harness and control universal social production freely. But all major AI companies only survive on venture capital for the specific promise to replace labor, with no sign of net profit in sight. As these companies race to gain market shares, customers don’t have to pay the full price yet for using the powers of AI. What we get today is ten times cheaper than the market would actually command. So, at this unique point in time, we are granted a partial glimpse at what could be possible if social production would in fact not be constrained by the capitalist contradiction – if the creative power of society would not be fully consolidated in capital and subjected to the commodity form, but freely available.
But this underlying potential can not be realized in capitalism. Venture capital demands AI companies to break even eventually. As our dependency on the models grows, they become more expensive for the user. This is already happening. But the users are workers, and the company that employs them gauges their model usage for cost-effectiveness in the production process of the commodity they offer. Only models whose usage is reasonably justified will be bought in. This forces the companies providing the models to further develop them according to these market demands. In other words, the form of potency that AI will represent, as well as how these models are used, will ultimately be fully subject to market dynamics acting on it as a commodity. There will be no free availability of the true potential of AI within capitalism. The realization of the full potential of social production within capitalism is only possible in socialism.
EPILOGUE: THE ABSENCE OF THE HISTORICAL SUBJECT
Like this, or somewhat like this, might the issue of AI be grasped from a Marxist point of view. To many, it might sound like an adequate inquiry into present day reality. In truth, there is a blind spot that consumes any soundness the arguments might appear to have. The Marxist perspective presupposes the proletariat as the social subject acting in history. But historically, this subject as a consciously organized force faded into nonexistence after the failure of the socialist revolution in the early 20th century. Honest study of the history of the Left reveals its failure in full clarity: there once was a proletarian movement for socialism, but it is no more. Of course, large parts of today’s marginalized Left insist – consciously or unconsciously – on the contrary. But their refusal to accept past failures just underscores the scale of the unbearable tragedy of the death of the Left that constitutes the main problem of politics today.
All of the thoughts and concepts brought together and arranged in this text stem from different moments in the history of this dead movement. Relying on these past thought fragments offers a treacherous path. As Chris Cutrone wrote in The Death of the Millennial Left, if we only rehearse unprocessed, shattered fragments of failed generations, we might never grasp what is unique and specific about the possibilities of the present moment. To grasp what is new and different, to learn from past and present, would require a judging subject that interacts with the object that is society. So, without an acting movement for socialism, what sense does it make to invoke all these Marxist thoughts and thinkers to frame the phenomenon of AI in this way?
Walter Benjamin makes this point explicit. Fundamentally, the thoughts invoked in this text are disparate, fragmentary. Arranging “those elements of the past which come together in the constellation of a single moment” can only become the cause for anything momentous if they are arranged to establish “a conception of the present as the ‘time of the now’ which is shot through with chips of Messianic time,” as Benjamin puts it in his essay On the Concept of History. Messianic is Benjamin’s term for the intervention into a “single moment” with the intent to change the course of history: revolution. Whether or not the specific arrangement of thoughts in this text allows any recognition of the true nature of our historical moment is impossible to tell without the subject that was supposed to execute this messianic act. In the words of Benjamin, only “the struggling, oppressed class itself is the depository of historical knowledge.” Historical insight exists only for it, and only for it in the historical moment. However, “the immeasurable pressure of domination has so fragmented the masses that it has even dissipated the negative unity of being oppressed that forged them into a class in the nineteenth century,” as Adorno put it in Reflections on Class Theory.
The conundrum is that, although society is as proletarianized as never before, there is no proletarian class as a historical subject anymore. We can only assume mentally the existence of this subject in the historical moment, arrange historical fragments for it, and perform speculative thoughts based on that – as done in this text – but that does not make the constellation of these thoughts reality. It’s just an analysis at that point, which may match reality, but just as likely it might not. Maybe we don’t even live in capitalism anymore and none of these thoughts hold up. After all, the essence of what Marx called ‘capitalism’ is the emergence of classes for itself, i.e. a self-conscious subject tasking itself with its own sublation. Without that, there’s only the class in itself, i.e. oppressed masses as a mere sociological category, which wouldn’t be capitalism in the Marxian sense. Under these conditions, even if this kind of analysis allows us to ‘understand’ that AI is reification representing the crisis of reason in society, collectively we would still continue to place AI outside of us, making AI increasingly material, in other words objective — An Alien Mind, as Jakub Pachocki from OpenAI puts it in his eponymous essay. Without Praxis, i.e. an existing societal subject capable and willing to change said reality, there is no point in ‘understanding’ any ‘real conditions of life’ in capitalism, because then these are not relevant for the ongoing manifestation of social reality, which is practical.
Given these conditions, AI might just take the place of humans because we allow it. Maybe machines that perform work truly autonomously might really change the character of labor in bourgeois society. Maybe they could in fact represent a kind of runaway creative power detached from labor performed by human subjects of society. Maybe leading to a cataclysm on the order of “the dissolution of all fixed personal (historic) relations of dependence in production” (Marx in Grundrisse) that marked the onset of capitalism. Maybe ultimately leading us into the dissolution of bourgeois society as well as its crisis, capitalism, into something different altogether? Into the abolition of humans in itself, made redundant as archaic carriers of an ancient version of creative power named labor? Condemned to vegetate in pods as a source of energy (or intelligence) for the Matrix?
With the historical collapse of the organized proletarian movement for socialism, we lost the ability to distinguish reality from fantasy. The verdict that AI is a reified phantasmagoria might simply be based on the bare insistence on it being true. Much like people insist that there must be a reason that people just keep on lifting Marx from his grave, both to aid them as well as to kill him all over again. Society surely must still be tasked with solving its contradictions, right? In reality, we simply can not know. Rightfully so, it drives us mad.
[1] This paragraph, including the point on the proletarianization of white collar labor, draws heavily on the thoughts laid out in Rodion Bruckner’s The Beginning of the Atomic Age, which is well worth reading in full.












.png)





.png)
.png)



