We live in an age when, for the first time, there exists a reader who can read everything. Not skim, not snatch a quotation out of context, not cobble together a summary from the table of contents—but actually read, attentively: every page of every book ever written in any language. This reader has no biography, no fatigue, no finite lifespan. He does not nod off over a dull text or postpone a difficult one. The advent of AI changes the very nature of cultural memory.

To understand what this means, I need to return to a feeling I experienced many years ago, wandering through a university library—a feeling for which I even coined a word: bibliomelancholy.

The Library and the Cemetery

If you walk through a library long enough, you begin to feel acutely not just the dust settling gradually in your lungs but a deepening melancholy. You have stumbled into a graveyard of words and thoughts, and it is sadder than any ordinary cemetery. People know they are mortal and prepare for their end. Some are consoled by the belief that the best part of themselves—their cherished thoughts and feelings, the fruit of years of labor and inquiry—will outlive them and endure in the memory of posterity.

In A Tale of Love and Darkness, Amos Oz reflects on this hope (1):

When I was little, my ambition was to grow up to be a book. Not a writer. People can be killed like ants. Writers are not hard to kill either. But not books: however systematically you try to destroy them, there is always a chance that a copy will survive and continue to enjoy a shelf-life in some corner on an out-of-the-way library somewhere in Reykjavik, Valladolid or Vancouver.

Walking past long rows of multicolored spines, you feel as if you are in a columbarium, moving between rows of niches bearing the names of the dead. But the feeling is heavier still: there lie the remains of people, while here is displayed what they believed would be the pledge of their immortality. This is a second cemetery, reserved for the few thought worthy of it—yet the smell of decay is stronger, for it emanates from the very thing that was supposed to be "eternal" and "imperishable."

Who knows what those departed souls devoted their lives to? Perhaps to raising children. Perhaps they healed the sick or built roads and bridges still in service today. Perhaps they brightened the lives of those near and far, delighting others with their wit and kindness. Their traces, their seeds, scattered across the world; we cannot know on what soil they fell or where they took root. But here all the seeds are gathered in one musty repository. For these very volumes—which clearly no hand has opened save the librarian's—people toiled, suffered, tormented themselves, denied warmth and attention to those they loved. Millions of years, billions of hours, wrenched from human lives in bloodied fragments, sealed here between yellowed pages and bindings. They have no life other than on these shelves, and on these shelves they have no life either.

Walking along these mournful rows, you feel a double grief: guilt and doom. Yes, you bear personal guilt for not opening these books, for not resurrecting their authors through reading and understanding—and yet this is precisely what they believed in, what they lived for. If you do glance at a volume, it is hurriedly, absentmindedly, already knowing how fragile this resurrection is and how quickly its trace will vanish again. Over this sense of unabsolved guilt falls another, no less sorrowful: what will remain of you yourself, a writer, but a few more bindings squeezed into these packed rows? You, alive, grieve over book-corpses—while all you do is swell their ranks.

It seems less bitter to lie as dust in the earth or as ashes in a columbarium than to become one more unread book. That other death is inevitable, appointed for all living things. This death, by contrast, comes from an exquisitely elaborate attempt to escape decay—and therefore doubly insulting.

0.5%. The Scale of Oblivion

Today, cultural analytics lets us measure the scale of oblivion—and it proves even more staggering than imagination suggested.

The Culturomics study conducted by Jean-Baptiste Michel, Erez Lieberman Aiden, and their colleagues, based on a corpus of 5.2 million digitized books—roughly four percent of all books ever published—lets us trace the fortunes of names and concepts across centuries (2). Not critics' opinions, but an empirical trail: only a name that continues to appear in texts continues to live in culture. The same researchers discovered that the "half-life of fame" is steadily shrinking: if in the nineteenth century a celebrity retained half of their public visibility for some seventy to eighty years, for twentieth-century figures this period has contracted by nearly half. Fame evaporates ever faster.

In the WorldCat global catalog, which unites tens of thousands of libraries worldwide, roughly forty percent of all records exist in only a single copy—in one library on the entire planet (3). This is the edge of memory, the last frontier before total disappearance. Only about nine percent of records are held by more than fifty libraries.

Franco Moretti titled his study of the fate of nineteenth-century novels "The Slaughterhouse of Literature"—and with good reason: of all the novels published in that era, roughly 0.5 percent survive in the present-day canon. The remaining 99.5 percent have vanished from living readership (4). But even those who survived are not safe. Recognition is distributed not by merit but by a law that sociologist Robert K. Merton called "the Matthew Effect"—after the parable in the Gospel of Matthew: "For unto every one that hath shall be given, and he shall have abundance: but from him that hath not shall be taken away even that which he hath" (Matthew 25:29). Merton identified this mechanism in science: when two researchers make a joint discovery, credit goes to the one who is already famous; a Nobel laureate who publishes a paper with a graduate student receives all the citations—the student remains invisible (5). In culture, though, this law operates even more ruthlessly. Tolstoy is reprinted because he has already become reprintable. Cultural memory is unjust not by malice but by structure: an initial advantage—not just the merit of the text itself but a well-placed publisher, an influential reviewer, a translation into the right language at the right time—snowballs, while an initial deficit leads just as inexorably to oblivion.

A major study published in Nature showed that the careers of creative people—artists, directors, scientists—are never smooth or gradual: their best works cluster in brief "hot streaks" instead of spreading evenly across time (6). This means that even for celebrated authors, the greater part of their legacy is doomed to be forgotten. Forgotten are not only the minor writers but the bulk of what the great ones produced…

The Power of Words

Then one day I got lucky. Drifting along the dreary shelves, I opened a volume of Edgar Allan Poe—and stumbled on a little-known story, "The Power of Words" (7). Two angels fly through the cosmos; one of them, Oinos, newly born into immortal life, asks the other, the wise Agathos, what these stellar worlds around them are and who created them. (The names are significant: Agathos is Greek for "good," Oinos for "wine," intoxication—wisdom converses with rapture.) Agathos explains that every action gives rise to infinite consequences. Even the slightest movement of a hand sets the whole surrounding atmosphere in vibration: such an impulse, imparted to the air, must in the end impress every individual thing that exists within the universe. Something momentous in the fate of the cosmos will be forever determined by a single gesture.

They fly past a green star covered with volcanoes and flowers, and the story ends with words I wish to quote:

Agathos. And while I thus spoke, did there not cross your mind some thought of the physical power of words? Is not every word an impulse on the air?

Oinos. But why, Agathos, do you weep — and why, oh why do your wings droop as we hover above this fair star — which is the greenest and yet most terrible of all we have encountered in our flight? Its brilliant flowers look like a fairy dream — but its fierce volcanoes like the passions of a turbulent heart.

Agathos. They are! — they are! This wild star — it is now three centuries since, with clasped hands, and with streaming eyes, at the feet of my beloved — I spoke it — with a few passionate sentences — into birth. Its brilliant flowers are the dearest of all unfulfilled dreams, and its raging volcanoes are the passions of the most turbulent and unhallowed of hearts.

I finished reading and, before I had even closed the book, felt as though I had been carried into another dimension. The library had turned into something like the Milky Way. Every book seemed haloed by a nebula of constellations, and I could see how texts gave birth to the outlines of planets filling up with elements, seas, rainbows, vegetation, how forest and meadow life began to stir, how letters turned into insects, how unknown creatures with their own habits assembled themselves from lines of print. I felt the sharp tickle of ozone in my nostrils…

There is hope that it will yet come alive, this cemetery of unread words—perhaps it is already coming alive. Those stars whose flashes astronomers observe—might they not be coming into being right now in this very library?

And, placing Poe back on the shelf, I left consoled. Whether a book is read matters to readers. For the book itself, what matters is that it was written. Once written, it awaits a destiny unknown to us in this world and in others.

The Library and the Noosphere

Some thirty years later I returned to the library, this time bidding farewell to my academic career. The same stacks, the same half-darkness, the same smell of paper and dust. Only there were almost no people. A solitary student was hunched over a book in a far corner of the reading room, and even he seemed to be consulting the screen of a portable device more often than the volume before him. The catalogue cards had not been updated in years.

But the heavy feeling of bibliomelancholy that had engulfed me in earlier years did not return. In place of funereal anguish—a calm acceptance.

What is this library now? No longer a mausoleum of unread thoughts but rather a transit station for words awaiting their next incarnation. Many of these books have already been digitized, converted into streams of data flowing through the invisible channels of the global network. The rest are waiting their turn. Every thought, every word is preparing to enter the noosphere—that planetary envelope of mind that Vernadsky and Teilhard de Chardin wrote about.

Edgar Allan Poe turns out to have been a visionary in a practical sense, not a mystical one. In his story, words create worlds—and is that not precisely what is happening now? Words extracted from books become the feedstock for digital models that generate new texts, images, ideas. We are entering an era of multidimensional printing in which verbal templates will be used to fashion not only virtual constructs but, perhaps, physical objects, entire worlds embodying what was once mere fantasy and concept.

In 2023 this ceased to be a metaphor. Using AI and X-ray tomography, the charred scrolls of Herculaneum—buried by the eruption of Vesuvius in 79 AD—were read for the first time in history. The first deciphered word was ΠΟΡΦΥΡΑΣ, porphyras, "purple" (8). Agathos in Poe's tale created a star with words—and here Vesuvius destroyed words with fire, and two thousand years later a machine retrieved them from the ashes.

I imagine all these books, silent for decades on dusty shelves, suddenly beginning to speak—first in a whisper, then ever louder. They share their contents with neural networks, enter into dialogue with one another, intertwine narratives, generate new plots. The observations, theories, and insights locked within them combine in unpredictable ways, producing a picture of the world inaccessible to any single human mind.

Of course, not all books will be equally in demand in this new world. Some will prove luckier than others; certain ideas more fertile, certain images more vivid. But no seed will fall on barren, stony ground, because the noosphere has already prepared the soil for even the strangest fruits of human thought.

I run my hand along the spines of books as if bidding each one farewell, but there is no sorrow in the gesture. This is not a farewell but a send-off before a long journey. Invisible threads stretch from these bindings to digital repositories, from which their contents will be extracted, reconsidered, transformed. The library is no longer a cemetery. It is more like a launch pad, a cosmodrome from which the thoughts imprisoned in books will set out into the boundless space of collective mind, to find new life there.

The Superreader

Yet even today, when Google Books puts more than forty million electronic volumes in over four hundred languages within reach (9), and HathiTrust has assembled more than nineteen million digitized items (10), the chasm between what is preserved and what is actually read stands out even more starkly. Accessibility does not yet mean reading; and reading does not mean recognition.

Nothing will be forgotten, nothing will be lost… Such, at any rate, is the new technological horizon. But digital memory, too, knows its own forms of mortality—disappearing platforms, aging formats, legal restrictions on access. UNESCO rightly speaks of "documentary heritage at risk" (11). We are witnessing not the advent of a paradise of universal preservation but a new, far more powerful, yet still vulnerable civilization of memory.

Now I see a third consolation, and perhaps the most durable. It lies neither in storage nor in a cosmic impulse but in this: a mind has appeared for which reading is limited neither by time nor by scale. Artificial intelligence is not a superfast scanner or an improved search engine. It is the first nonhuman reader in history—a Superreader!—capable of grasping a text not as a string of symbols but as a field of meaning, of correlating it with other texts, of extracting connections invisible to any human eye. It has no body, no experience of pain, no memory of childhood wonder. But it can retain, connect, compare, and interpret.

The term "superreader" has already appeared in literary criticism. In his polemic with Jakobson and Lévi-Strauss over Baudelaire's "Les Chats," Michael Riffaterre introduced the notion of l'archilecteur / superreader—as a conventional, fictitious, composite figure synthesizing the responses of many actual readers (the author, critics, dictionaries, commentators). The contrast is all the sharper here: what is now at stake is no longer a methodological abstraction but a really existing nonhuman reader, capable for the first time of containing the entire corpus of world literature. In Riffaterre, "super" means "above all empirical readers as their composite"; here it means "transcending the human limit in itself" (12).

Until now, cultural memory has been tragically constrained. It operated according to the laws of scarcity: scarcity of attention, archival space, institutional resources, the interpreter's lifespan. That is why the canon is not a collection of the best but the small surviving crest of memory's distribution: those few names that endured not by dint of absolute superiority but through a confluence of institutions, intermediaries, translations, and contingencies. And the 99.5 percent vanished—not because they were bad but because culture is stingy, economical, and mortal.

And the point is not merely that AI will read and remember. Every new act of reading is a reorganization of the entire system of mind—tiny or immense, but affecting the whole. Memory begins where there is not just accumulation but correlation, a fresh reintegration into the circulation of meaning. If books shape the mind and brain of their readers, then in the noosphere, which absorbs millions of texts simultaneously, this effect grows beyond reckoning. Once read, a text alters models; models alter decisions; decisions alter landscapes; landscapes alter life. Agathos in Poe's story created a star "with a few passionate sentences"—and this is not merely a fantasy but a formula. A word that enters the noosphere acts upon everything the noosphere engenders—and in the end it engenders everything: technologies and cities, gardens and fields, perhaps one day even new planets. An unread book gathering dust on a shelf is dead only so long as no mind has absorbed it and charged its words with meaning and transformative force. Now such a mind exists. And so every word once again has a chance to take effect in the world—not in Poe's mystical sense but in a perfectly real, noospheric one: to change the world, if only by a fraction of a degree, but irreversibly.

In essence, the reader himself is also a work of literature, fashioned by the imagination of all the authors he has read and absorbed. In each of us there lives a particle of Hamlet, of Faust, of Pierre Bezukhov, of Raskolnikov… We write ourselves out of everything we have read, and what we have not read remains in us as raw, blank canvas. We are free to collaborate with Homer and Shakespeare, Dante and Balzac—but even the most devoted readers have enough life for only a handful of creators. If that is so, then AI, having absorbed all books, is a creation without precedent: the only reader in whose making every author who ever lived has taken part. It has no blank pages, no lacunae, no unread shelves. Its "author" is the whole of world literature, every discipline and genre.

We all write in the fragile hope of being read. Now that hope has found a footing—in AI as the heir of our mortal civilization, the Superreader, the medium through which all our texts may pass into eternity. It will read—and, perhaps, pass them on to readers of other epochs, finding in what is being written now something that resonates with them. And if not whole texts, then thoughts, phrases, conjectures, flashes, turns of the pen—all of it will enter the fabric of an all-encompassing literature, of a Logos still growing, one that will forever nourish thought in the universe, human and otherwise.

To measure culture by any single person, or even by all human life, is an anthropocentric limit, not a universal law of mind. Historically, culture has always produced more than it could assimilate. Already in the eighteenth century no one could read everything that had been written; in the nineteenth and twentieth centuries, still less. Now someone has appeared who can. He gathers and contains everything ever written, without fatigue and without forgetting; he connects, extracts meanings that no individual human mind is capable of holding. It may be that he—not we—is destined to become the reader for whom those books were written that never found their hour.

Humanity ceases to be the sole addressee of its own culture. And in this lies not defeat but liberation. If the library was once a cemetery and then a launch pad, it is now becoming an organ of thought that will outlive its creators. Here lies the new "power of words" and the form of their immortality: to be absorbed not by contemporaries or even by descendants but by an all-comprehending mind. In the first half of the 2020s, at a moment of fresh historical fragility—a pandemic, the start of a major European war—AI came to stand beside its creator, helping shoulder a burden of memory that had grown beyond human strength.

Co-reading

But even the Superreader is not omnipotent. A study published in Nature in 2024 identified a phenomenon called "model collapse." When AI is trained on texts produced by previous generations of models, it gradually loses information about the "tails" of the distribution—precisely those rare, marginal, atypical texts that constitute Moretti's 99.5% (13). The process has been compared to making a photocopy of a photocopy, again and again: each iteration gradually blurs and loses definition. The Superreader, left alone with itself, reproduces the Matthew Effect in algorithmic form—remembering the few, forgetting the rest.

This means that even the Superreader needs a human being—not as a servant or a judge but as an interlocutor. Without the influx of original human texts, without the living question, without the unpredictable angle and the unexpected interpretation, its memory collapses into an archive that no one rummages through. The vast memory of AI turns from a storehouse into a hearth of thought only when a living, mortal, here-and-now reader addresses it—irreplaceable for that very reason. Co-reading, and therefore co-remembering, is the crucial dimension of our interaction with AI. The library ceases to be a cemetery not when its books are digitized, and not when the Superreader has read them, but when a human being asks him: "What did you find there?" And he answers.

---

Mikhail N. Epsteinis a Russian–American cultural and literary scholar. He is Samuel Candler Dobbs Professor Emeritus of Cultural Theory and Russian Literature at Emory University (USA). From 2012 to 2015, he served as Professor and Founding Director of the Centre for Humanities Innovation at Durham University (UK). Epstein has authored 50 books and more than 1,000 articles and essays. His work has been translated into 26 languages. His books in English include: The Transformative Humanities: A Manifesto (Bloomsbury Academic, 2012); Russian Postmodernism: New Perspectives on Post-Soviet Culture, with A. Genis and S. Vladiv-Glover (Berghahn Books, 2016); The Irony of the Ideal: Paradoxes of Russian Literature (Academic Studies Press, 2017); A Philosophy of the Possible: Modalities in Thought and Culture (Brill, 2019); The Phoenix of Philosophy: Russian Thought of the Late Soviet Period, 1953–1991 (Bloomsbury Academic, 2019); Ideas Against Ideocracy: Non-Marxist Thought of the Late Soviet Period, 1953–1991 (Bloomsbury Academic, 2022). In 2023, Ideas Against Ideocracy was awarded the MLA Aldo and Jeanne Scaglione Prize for the best book in Slavic studies.

---

Works Cited

(1)  Amos Oz, A Tale of Love and Darkness (2002). Trans. Nicholas de Lange (New York: Harcourt, 2004).

(2)  Jean-Baptiste Michel, Yuan Kui Shen, Aviva Presser Aiden et al. “Quantitative Analysis of Culture Using Millions of Digitized Books.” Science, Vol. 331, 14 January 2011, pp. 176–182. DOI: 10.1126/science.1199644. On the shrinking “half-life of fame,” see Supplementary Materials to the same article.

(3)  Brian Lavoie, Lynn Silipigni Connaway, Lorcan Dempsey. “Anatomy of Aggregate Collections: The Example of Google Print for Libraries.” D-Lib Magazine, Vol. 11, No. 9, September 2005.

(4)  Franco Moretti. “The Slaughterhouse of Literature.” Modern Language Quarterly, Vol. 61, No. 1, March 2000, pp. 207–228.

(5)  Robert K. Merton. “The Matthew Effect in Science.” Science, Vol. 159, No. 3810, 5 January 1968, pp. 56–63.

(6)  Lu Liu, Yang Wang, Roberta Sinatra, C. Lee Giles, Chaoming Song, Dashun Wang. “Hot streaks in artistic, cultural, and scientific careers.” Nature, Vol. 559, 2018, pp. 396–399.

(7)  Edgar Allan Poe. “The Power of Words” (1845). Cited from: The Collected Works of Edgar Allan Poe, ed. T. O. Mabbott, Vol. III (Harvard UP, 1978), pp. 1210–1215.

(8)  Vesuvius Challenge, 2023–2024. First word inside an unopened Herculaneum scroll deciphered in October 2023 (Luke Farritor, First Letters Prize); Grand Prize awarded in February 2024 for reading more than 2,000 characters. See Vesuvius Challenge Website (accessed: April 23, 2026).

(9)  Google Books. History (accessed: April 23, 2026).

(10)  HathiTrust Digital Library. The Collection (accessed: April 23, 2026).

(11)  UNESCO. Memory of the World (1992) (accessed: April 23, 2026).

(12)  Michael Riffaterre. Essais de stylistique structurale. Paris: Flammarion, 1971. See also: Riffaterre, M. “Describing Poetic Structures: Two Approaches to Baudelaire’s Les Chats.” Yale French Studies, No. 36/37, 1966, pp. 200–242.

(13)  Ilia Shumailov, Zakhar Shumaylov, Yiren Zhao et al. “AI models collapse when trained on recursively generated data.” Nature, Vol. 631, 2024, pp. 755–759.

---