The banalization of prose

This essay was first prepared as a talk for a gathering at the Santuari del Miracle.

Saturday at eight in the morning may not be the ideal hour to speak of evil. One usually hopes, at that time, for coffee, light, and perhaps a few forgiving illusions about the human condition. Instead, here we are, beginning the day with authoritarianism, war, climate catastrophe, and the general exhaustion of the planet. So let me at least say this: if anyone feels that this is asking a bit much of a Saturday morning, that feeling is itself evidence of sound moral judgment.

And yet it is hard to imagine a more unavoidable topic. We are living amid forms of evil that should not require much interpretive sophistication to recognize. The rise of authoritarian politics across many parts of the world, the normalization of war and displacement, the open contempt for vulnerable populations, the systematic destruction of ecological conditions for life, the strange persistence of overconsumption in a world that has already given us every warning it can—these are not hidden phenomena. They are visible, nameable, and insistent. They confront us at the level of institutions, policies, borders, markets, atmospheres, and screens. No longer marginal disturbances in an otherwise stable order, they are increasingly woven into the ordinary functioning of that order itself. We live, in other words, under conditions of constant emergency. But the deeper emergency is not that catastrophe keeps arriving, but that we have become so practiced at registering emergency that we do very little, almost nothing, to address it responsibly.

There is, then, an obviousness to evil in the present moment. Some things are simply cruel. Some arrangements are clearly destructive. Some forms of power degrade human beings while pretending to protect them. The ongoing war involving Iran has made that obscenity difficult to miss, as did Donald Trump’s threat that “a whole civilization will die tonight” if Iran did not make a deal—a sentence that should have met immediate and unanimous condemnation. Some habits of consumption depend on a disavowal so massive that it begins to look somewhat metaphysical: we consume as though the earth were inexhaustible, as though waste disappeared by magic, as though comfort had no world beneath it, no labor behind it, no future after it. Indeed, evil appears in brutal and familiar forms—domination, indifference, extraction, organized violence. It burns, floods, starves, expels, incarcerates, and abandons. It rarely lacks for evidence.

Now, this obviousness does not exhaust the question. Alongside the visible theater of burning cities and ecological depletion, another theater is consolidating at the level of code, chips, grids, models, and networks. One increasingly suspects that the coming cold war will be fought less through the spectacle of ideological opposition or a renovated arms race than through AI-mediated struggles over infrastructure itself. The United States and China may continue to posture as rivals in this domain, and surely they will; yet they may also be forced into moments of collaboration, because a generalized capacity for automated cyberattack would threaten both powers at once while lowering the threshold at which smaller actors can inflict enormous damage. Evil, in that register, appears as systemic exposure, permanent vulnerability, and the unsurprising democratization of sabotage.

Globalization once presented itself as the overcoming of geopolitics, as though markets, networks, and circulation would dissolve the older antagonisms of territory, empire, and rival blocs. In fact, the algorithm has made globalization far more operable, more granular, and therefore more geopolitically unstable. It coordinates supply chains, financial flows, satellite systems, ports, warehouses, platforms, attention, and extraction at a scale no imperial bureaucracy could ever have managed by hand. But precisely because these systems are so dense, so optimized, and so distributed, power no longer appears only in the old dramatic form of one sovereign facing another across a visible frontier. It becomes fractal: dispersed across cables, standards, choke points, chips, clouds, rare earth corridors, shipping lanes, payment systems, and privately owned infrastructures whose political force is immense yet often difficult to narrate. Geopolitics does not disappear under globalization; it fragments, multiplies, and retreats into technical architectures that are harder to see and therefore easier to naturalize. The old dialectic of clearly opposed powers remains, of course, but it no longer tells the whole story. Conflict now proliferates through networks whose local disruptions can trigger planetary effects, while responsibility becomes ever more difficult to trace.

So, what interests me this morning is not only evil in its most spectacular or legible forms. I am more interested in the quieter conditions under which evil becomes easier to produce, easier to normalize and to inhabit without ever quite naming it as such. I am thinking of the subtler degradations of judgment, attention, language, and relation that prepare the ground on which these other, more obvious evils, flourish. Evil is not the result of monstrous intentions alone. It is also produced by atmospheres, by habits, by administrative idioms, by repeated evasions, by the thinning of responsibility, by the replacement of judgment with procedure, by the gradual loss of contact with what one is saying and doing. If there is a terrifying lesson in recent history, it is that catastrophe often arrives fluent, procedural, optimized, and well formatted.

That, in part, is why I have turned more and more to wine and prose in recent years. And yes, to be clear, since we are speaking of evil at eight in the morning, this is partly a coping mechanism. I confess that geopolitical reality has done a great deal to improve my appreciation for fermentation. I have already fled two authoritarian countries and finally found refuge in wine-producing Catalonia. But the turn to wine and prose is not just a matter of survival, nor a compensatory gesture. It is also a philosophical bet. Both wine and prose remind us that reality is relational. We do not inhabit a clean subject-object world composed of detached minds confronting inert things. We live, rather, in an already existing web of relations: with others, with language, with labor, with memory, with landscapes, with territories, with climates, with inheritances we did not choose and responsibilities we cannot finally evade.

The more my life passes through writing and speaking, the more my philosophical and political attention turns to wine. The old world knew the intimacy between wine and truth quite well: oinos kai aletheia. What does wine disclose? What does it gather? What forms of memory, labor, and earth linger in a glass? In time, I came to see that words and wine are exposed to a similar diminishment. Both can be watered down, losing body and the lingering resonance that binds them to the earth and to others. Wine bears the marks of cultivation, labor, soil, weather, storage, delay, exchange, company, memory, and sharing. Words bear such marks as well. They arrive in address and response, in inheritance and risk, in misunderstanding, tone, judgment, and exposure. We speak to someone, from somewhere, under pressure or with pleasure, inside a web of relations that precedes us and will outlast us. Language is one of the places where a world—perhaps even the world itself—comes forth.

In his work on the five senses, French philosopher Michel Serres reminds us the obvious: our relation to the world is, first and foremost, tactile, gustatory, terrestrial. Taste matters because it binds body to world without the fantasy of mastery. One can taste a whole territory, a whole tradition, a family in wine; one receives the earth there in a form that lingers, differentiates, and demands attention. That is part of what makes wine philosophically instructive. It makes relation, and thus truth, immediately sensible.

Generative AI, by contrast, presents itself under the sign of frictionless abstraction. And yet there is no such thing as disembodied intelligence here. There is no AI without earth: without silicon, copper, gallium, germanium, rare earth elements, water, electricity, logistics, mining, refining, transport, and waste. The difference is that these terrestrial conditions do not return to us as savor. They are buried beneath interfaces and hidden in supply chains, rendered illegible by the fantasy of the cloud. Wine bears its earth in taste. AI consumes its earth while concealing it.

That is why I have grown increasingly uneasy about what generative AI is doing to prose. My worry does not begin with misinformation, job displacement, or the now tiresome dispute over whether machines can think—a dispute that has already produced more headlines than insight. I have deep sympathies for automation, in an unashamed Marxian sense. If technologies can release us from degrading routines and alienating bullshit jobs, if they can free up hours for conversation, study, idleness, love, music, play, wine, and poetry, I welcome that release. A life should contain more than administration, repetition, and managed exhaustion.

My concern lies elsewhere. Generative AI is standardizing prose at remarkable speed. It regularizes emphasis and fills language with prefabricated transitions, ready-made intensities, and an impeccable synthetic balance. The damage is easy to miss because the sentences remain serviceable.

Prose, I say, is not merely a vehicle for information. Like wine, it belongs to pleasure and form as well. It is part of the medium in which selves and worlds become shareable and enjoyable. Whenever language is reduced to instrument, the narrative and aesthetic construction of self and world flattens. Like wine, we are first earth, bound to finitude and return––dust to dust. Yet we are also what we can say, what we can answer for, the shape in which we come into view for others and for ourselves. Responsibility begins there: in answerability, in having to stand by one’s words and answer for their content and cadence.

Hannah Arendt’s work offers a precise grammar for the temporal structure of human life. Her distinctions among labor, work, action, and thinking help clarify what is now shifting under the pressure of generative AI.

For Arendt, labor belongs to necessity. It keeps biological life going. It feeds, cleans, repairs, tends, repeats. Its rhythm is cyclical because its products vanish into consumption and must be produced again. Bread is eaten. Dust returns. Bodies tire, hunger, fall ill, recover, and require care once more. Labor is indispensable, yet it leaves little behind. It does not settle into permanence. It circles back on itself and binds life to recurrence.

Work unfolds in another register. Work builds and fabricates a world. It gives form to things that last longer than the urgencies that produced them: a table, a house, a road, a book, a tool, an institution. Through work, human beings stabilize their surroundings and inhabit something more durable than need. Work resists pure consumability. It gives the world thickness, endurance, a certain power to remain. Work fights time.

Action takes place among others. It belongs to the space of plurality, where speech and deed disclose who someone is. Action initiates. It enters a web of relations already in motion and alters it in ways no one can fully control. For that reason, action is exposed to unpredictability from the start. It depends on appearance, on risk, on the courage to begin in the presence of others and without guarantees.

And then there is thinking. Thinking, for Arendt, is not a technical operation and not a storehouse of knowledge. It is a withdrawal, a suspension, a pause wrested from the pressures of immediacy. Thinking steps back from the rush of response without turning away from the world altogether. It interrupts the automatisms of conduct and opens an interval in which judgment can take shape. It is the human mind in action once released from the pressures of survival.

What unites these distinctions is a relation to time. Labor turns in cycles. Work gives duration. Action erupts in the interval between people and sets something in motion. Thinking arrests momentum and opens a reflective pause.

My claim is that generative AI carries the temporality of labor everywhere else. It brings the logic of reproducibility, replacement, and acceleration into regions of life once shaped, at least in part, by durability, initiative, reflection, contemplation, pleasure, and idleness. Language begins to move at the tempo of upkeep. It is produced, consumed, replaced, and produced again.

I am not suggesting that architecture will vanish, or land, or food, or crafted objects. The material world persists. Tables remain tables. Vineyards remain vineyards. The city stays there in stone, steel, glass, wood, trash, cockroaches, and traffic. Even the most enthusiastic prompt cannot yet produce a chair one can actually sit on. What changes is the discursive atmosphere in which these durable things appear and acquire intelligibility. Representations, commentaries, descriptions, explanations, captions, summaries, essays, emails, statements, reflections: all this can now be generated instantly, replaced without effort, and sent around before it has been tested, often before it has even been read, thought through, or given time. We have achieved, among other things, a civilization in which draftiness, error, experiment, the slow maturation of an idea, the struggle proper to creation, the playful and painful sketching of anything are increasingly treated as inefficiencies. It is not only that war is now often waged remotely, through technocratic and bureaucratic apparatuses far from the suffering of its victims. Something similar, at another register, is happening to our creative lives. The strain of finding the right word, the pressure of shaping the right thought, the pride, desire, delight, and hurt bound up with artistic, intellectual, experimental and imaginative struggle—all this begins to lose value before ready-made discourses and prefabricated “ideas.” What is severed in the process is effort as much as meaning itself: the last threads that bind expression to ordeal, patience, risk, and form. What remains is the gleaming product—the AI-generated song, the immaculate paragraph that the Australian musician Nick Cave described as “perfect in its cynicism, magnificent in its emptiness.”

Worldliness, as Arendt defines it, names the condition of sharing a world. It names the space in which things stand between us, gather us, separate us, orient us, and become available for speech and judgment. A world is not a collection of objects. It takes shape in the space where those things are named, described, remembered, discussed, contested, and interpreted. It depends on the narratives and distinctions through which plural beings come to hold something in common. Worldliness is that fabric of durability and appearance, object and meaning, thing and saying, through which human life becomes inhabitable together.

I fear generative AI makes that environment far more volatile. As it fills it with rapidly produced verbal matter, it lowers the real cost of saying something, thinking something, phrasing something, shaping something. It multiplies discourse while thinning ownership and, thus, responsibility. It accelerates the production of phrasing, commentary, and plays the game of interpretation surprisingly well. In doing so, it extends necessity into cognitive life. The demand once limited to producing food, services, or repetitive tasks now reaches into symbolic production itself: produce the post, the response, the paragraph, the explanation, the abstract, the grant statement, the literature review, the cover letter, the feedback, the lesson plan, the position statement, the poem—and keep producing.

At first glance this may not seem tragic. After all, automation relieves drudgery. No serious account of technology should romanticize exhaustion. Allow me to repeat myself: if a system can take over repetitive or metabolically draining tasks and create room for reflection, leisure, study, conversation, even contemplation, then let us enjoy it. I do not want to say that automation is evil, or that every use of AI is corrupt, or that scholars should refuse such tools on principle. My concern is narrower and, I hope, more serious. When acceleration becomes continuous, the interruption proper to thinking is crowded out. Seamless responsiveness begins to replace reflective pause. The system is always ready with another sentence, another reformulation, another summary, another beginning, another enlargement. Under those conditions, it becomes harder to sustain the delay in which judgment forms. This, I think, is where the question of evil enters.

Arendt taught us to attend to the conditions under which people cease to think. Her account of the banality of evil has been debated, criticized, defended, abused, and often oversimplified. I do not want to flatten it further. But one enduring insight remains: evil proceeds under conditions in which language, judgment, and responsibility become strangely disconnected. One speaks in formulas. One relies on clichés. One inherits ready-made phrases and processes that spare one the burden of examining what one is doing.

I am not saying that AI-generated prose or music is equivalent to evil. That would be absurd. What I am saying is that the banalization and automatization of prose belongs to the same moral atmosphere in which unexamined language becomes ordinary. It makes speech easier to produce and harder to own. It furnishes expressions without requiring inward labor. It encourages the circulation of phrases that bear the shape of thought without the cost of thought.

And this matters because, for Arendt, action discloses the “who” of a person. We reveal ourselves through speech and deed. A self is never “hidden inside” and later expressed. A self is made and disclosed in public, in relation, through forms of address, reply, promise, recollection, and responsibility.

If that is right, then prose is never a trivial matter. Prose belongs to the medium in which a person becomes a person. A world saturated with automated phrasing risks weakening that disclosure and flattening it. Narrative becomes easier to produce and harder to own. Self-presentation grows more fluent and less responsible. We still speak, though increasingly in sentences that no one has fully chosen, made, struggled with, polished, or embellished. And the loss is larger than style. We inherit fewer living phrases from other minds; we lean less often on tacit echoes of Scripture, the classics, or the books that first formed us as readers. What once came into prose as sedimented memory—as allusion, cadence, or half-conscious recollection—gives way to language that arrives ready-made and unplaced, detached from any durable scene of reading or tradition of thought. Alas, poor prose.

That is why I think the banalization of prose is also a danger for the narrative construction of the self. Our world is, we are, narrative through and through. We do not live on facts alone. We live by the stories through which events become intelligible, through which actions are linked, through which memory is organized, through which suffering and aspiration are given form. We are not just what we eat or wear, but mostly what we say we are, who we say we are. We become available to others in language, and we become available to ourselves through the retrospective and anticipatory work of narration. And when that narrative labor is outsourced to prefabricated rhetoric, something important is lost. One may still produce a coherent self-description, but coherence is not the same as singularity. Smoothness is not the same as truth. Fluency is not the same as responsibility.

This is especially troubling in academic life, which is my natural habitat.

Finding AI mannerisms on social media is one thing. Finding them in professional scholarship is sadder. One expects a certain amount of verbal litter online. In scholarship, one hopes for fewer fingerprints from the autocomplete of history. Scholarship should be one of the places where language bears the marks of discipline, judgment, risk, and revision. I do not mean ornament. I do not mean inflated seriousness. I mean prose that has been tested, pruned, resisted, and made answerable to a real mind. Scholarship matters because it is charged with something more than the delivery of information. It is one of the practices by which a culture teaches the next generation what thinking is: how to handle words with respect, how to inherit a world of meanings, how to examine that inheritance, and how to pass it on without either embalming it or betraying it. But scholarship is not alone here. Political life is public too, and for that very reason it depends on language that can bear judgment. A polity lives by the words through which it argues, remembers, promises, and holds itself to account. When those words are thinned into slogans, templates, and automated reassurance, both scholarship and politics are diminished together.

Here I have to make another personal confession. I firmly believe that living is editing. I believe that a good writer is, before anything else, a disciplined, stubborn, cold-blooded editor. Writing is not mainly the production of sentences. That part, as it turns out, is cheap. Writing is the labor of trimming, pruning, discarding, and continuing to discard until the page contains what belongs there and no more. Every serious writer knows the experience of cutting pages that once seemed promising but that finally had no right to remain.

Part of my own education consisted in unlearning rhetorical habits to which I had clung for too long. American academia helped me there. It taught me to distrust flourish when flourish was doing the argument’s work badly. I was once even more vulnerable to stylistic excess. Admittedly, I still admire that kind of prose—its density, its daring, its verbal torque. Yet admiration can harden into mannerism. At some point one has to decide whether style is serving thought or thought is being bent to serve style.

Our pressure now is no longer rhetorical inheritance, but algorithmic availability. One no longer needs to fall into mannerism through admiration. One can inherit it instantly from a machine trained on enormous quantities of already sedimented, grammatically and orthographically impeccable prose with no aesthetic aspirations at all. That is why editorial discipline matters more, not less. The task is not to renounce the tool in a posture of purity, but to refuse dilution.

And this brings me back to words and wine. Wine can be overhandled, flattened, sweetened, standardized, made frictionless, universally acceptable, easier to consume, and less worth remembering. Language can suffer the same fate. A world of endlessly serviceable prose may turn out to be a world with less savor, less contour, less relation, less beauty. One can drink it. One can pass it around. One may even prefer it for a while. Yet the flavor no longer lingers in the mouth—nor the mouth that drinks, nor the one that speaks.

I think Derrida, or at least one line associated with Derrida’s readers, helps name the deeper issue here: the formation of an egodicy, a justification of the self. Every serious intellectual life is engaged in some effort to make a self coherent, answerable, inhabitable. Philosophy is never impersonal. It is always, in some measure, a struggle over the form of one’s own existence. If that is true, then one has to ask: what kind of egodicy is being written when one’s prose is assembled from ready-made templates and machine-default turns of seriousness? What self is being justified there? What inward labor is visible in language that sounds as though it could have been written by anyone, for anyone, anywhere?

This question, which might seem adequate for the professional philosopher alone, really concerns anyone trying to become who she or he is. If we learned anything from Augustine’s Confessions, it is that every life that seeks coherence must, in some way, narrate itself, revise itself, answer for itself. We become intelligible by struggling to find words equal to our experience, our obligations, our losses, our loves, our failures. That struggle is not incidental. It is part of what gives a life its contour. When that labor is outsourced to prefabricated discourse, one may still produce a fluent self-description, but fluency is a poor substitute for selfhood. A person may come to sound coherent without having undergone the ordeal by which coherence is earned.

Evil flourishes where language has been severed from inward labor, where self-description grows smooth because self-examination has grown thin. One condition of evil is a willingness to settle into formulas never fully tested against conscience, suffering, or truth. The danger, once again, is ethical and therefore political, rather than stylistic. One becomes publicly intelligible while remaining inwardly unformed—fluent in borrowed language and therefore less capable of judgment, less capable of answerability, and less capable of resisting the moral ease through which evil so often advances.

For that reason, the question of prose reaches beyond literary taste or professional craft. It touches the basic work of becoming answerable for one’s existence. A self is not justified by smoothness. It is not redeemed by grammatical correctness, nor by the ability to produce an acceptable account on demand. What matters is whether one’s words bear the marks of having been lived with, tested, suffered through, revised, and made one’s own. The deepest danger in ready-made language is that it offers the form of self-justification while quietly emptying out the labor that might have made such justification true. And where that labor disappears, evil finds one of its most ordinary entrances.

Plato offers an image that says this better than I, or any AI, can. In the Seventh Letter, he speaks of a figure that is drawn and then rubbed out again. The Greek verb he uses there is exaleipho: to wipe away, erase, blot out. Strictly speaking, Plato uses it there to refer to an erasable image, not as a description of thought-processes. Even so, the image is exact for what I am trying to say. Thinking is not only the production of language. Thinking, as I suggested earlier, also erases, discriminates, discards, rejects, chooses. It wipes away what should not remain. It removes what obscures. It rubs out the clutter that gathers around the thing itself––around we ourselves.

That, to me, is what editing finally is. Not cosmetic, but ethical and intellectual through and through. It is a practice of relation to truth, relation to self, relation to others. It is one of the ways we resist dilution. Call it asceticism, if you will. If Arendt is right, thinking interrupts. If Plato is right, thought must also wipe away. And if I am right, if we are filling the world with automatic language, then our responsibility is not to keep speaking. It is to keep erasing, editing, wiping away, and even remaining silent. Responsibly.

So I would put the point this way: the danger of generative AI is not exhausted by falsehood or plagiarism. The deeper danger is that it extends the relentless, destructive logic of contemporary production into the very space where persons and worlds should become narratively intelligible. It tempts us with language that arrives before thinking. What Arendt called the banality of evil was not the claim that evil is ordinary in any morally deflating sense, but that it may proceed under conditions of thoughtlessness. The trouble is not so much wicked intention but the failure to stop and think.

I already said that, for Arendt, thinking is a withdrawal into the silent dialogue of the self with itself. Thinking creates an inner interval in which one examines appearances, tests words, and asks whether one can live with what one is saying and doing. This is not to say that thinking guarantees goodness, but its absence leaves one especially vulnerable to all the ready-made idioms and slogans by which responsibility is dulled. That is why Arendt saw in thoughtlessness one of the conditions under which evil becomes possible, and why the banalization of prose is not a minor concern. It belongs to a broader weakening of the conditions under which thinking, action, and responsibility can still appear. When language comes prepackaged, when expression precedes judgment, when the sentence is always already waiting for us, the pause in which conscience might form seems no longer necessary. One speaks and publishes before one has quite entered into relation with what one is saying. And that is why the matter is serious. The danger is not that prose becomes flatter––though bad prose makes the world uglier, and beauty is one of our most basic needs. The danger is that thoughtlessness acquires ever more elegant tools. Evil, after all, rarely announces itself in broken grammar.

And that, perhaps, is one of the subtler forms of evil-making today. Evil surely consists of bombs, prisons, camps, borders, torture, and the usual brutalities of men and women with power. But it also grows among us, relatively powerless citizens, wherever language becomes detached from judgment, wherever responsibility is thinned by fluency, wherever the appearance of thought begins to substitute for thought itself. The more seamless the sentence, the easier it may become to forget the burden of saying anything true. And this matters all the more in a geopolitical landscape like ours, where emergencies now arrive in clusters and at speed: the war involving Iran convulses oil routes and energy prices; Sudan continues to grind through hunger, displacement, and attacks on civilians; Nigeria absorbs the inflationary aftershocks of energy shock; Venezuela remains caught in the long afterlife of oil and foreign intervention; rare earths and critical minerals redraw the map of strategic dependency. We are surrounded by alarms. But part of the danger is that we have grown used to answering them badly: by registering, summarizing, circulating, and moving on; by outsourcing response itself to automated habits of language and online practices that simulate concern more easily than they sustain judgment. That incapacity to listen to emergency and answer it properly is not the whole of evil, but it is one of the conditions in which evil now thrives.

And yet I do not want to end in despair. Good wine and good words may still help save us—not because they teach us to look away, but because they return us to relation. Wine, left to itself, does nothing. But once tasted, it opens a world: flavor, memory, pleasure, earth, labor, history, weather, company. Shared among friends or family, it becomes a small school of attention, gratitude, and presence. Prose is not so different. A closed book does nothing on its own. But once I enter into relation with it, I am altered by it—and with me, however slightly, the world I inhabit. That is why I still believe an offline life can save us. Orderly idleness, creative otium, are forms of existence that resist permanent reaction and refuse the compulsion of immediate response. We need work, certainly. But we also need contemplation. Our Benedictine hosts knew as much. A life worth living cannot consist in output alone. It must also make room for silence, ripening, study, friendship, pleasure, pain, and the slow formation of judgment. Perhaps that is where resistance now begins: in recovering practices that teach us again how to attend, how to savor, how to think, and how to answer for one another without automation standing in for the work of the soul. Words, like wine, should not be watered down past recognition. Nor should life itself.