Manifesto for Artistic Documentation as a Resurrection of the Archives of Possibility
I return to my archives. Not to contemplate them as one would contemplate ruins—a nostalgia I refuse—but to twist them, to fold them, to bring them to the point where material conditions betrayed the project. Years of constitutive gap between what I wanted and what my means allowed. This gap was not a failure. It was the ordinary condition of a practice working with finitude, bearing its marks: overexposed production stills, low-resolution screenshots, storyboards never built, sequences filmed with the wrong lens because the right one was too expensive. Indexical images in the strictest sense—imprints of what took place—but imprints of a reality that was not yet what it wanted to be.
I return, then, to these traces. Not to archive them further. But to regenerate them and finally allow them to happen.
An actor in a hallway. The light is poor, the framing shaky, the focus approximate. I feed the network this image and a hundred others like it. I formulate a few words to guide the drift. And what returns is not a restoration of what I had filmed. It is something I never filmed, yet I recognize it immediately: it is what the scene wanted to be. The light becomes just, the framing breathes, the focus holds. None of this took place. And yet, it is more faithful to the intention (but behind the intention lies the reality of the possible) than the trace ever was.
This moment—this precise moment where one recognizes in a generated image something that never existed but was expected—is not new in art history. It has been sought, approached, bypassed, and simulated in many ways for over a century. What is new is that, for the first time, a technique allows it to be accomplished from its own indexical archives, without falsification, without hoax, without speculative restoration. What I am attempting to describe here is why this moment constitutes an ontological deepening of a question that art has never ceased to ask.
In 2010, Artie Vierkant published The Image Object Post-Internet. The text is short, dense, constructed as both a technical and theoretical manifesto. Its central thesis: in the contemporary world of digital networks, the distinction between original and reproduction has become operationally void. A work now exists simultaneously in multiple states—high-resolution file, print, screen display, screenshot, redisplay on another screen, JPEG compression, republication on a third-party platform—and none of these states is more original than the others. Each display modifies the work: colors vary according to screen color profiles, dimensions vary according to resolutions, metadata accumulates or is lost. These modifications are not degradations of a stable original. They are the very life of the work within its real conditions of existence.
What Vierkant describes is not merely a technical observation. It is an ontology. The post-internet image-object has no essence separable from its conditions of distribution. It is what it does within networks, the traces it leaves, the modifications it undergoes and produces, the contexts it traverses. There is no “being” of the image that precedes its appearances: there are only appearances, each as real as the next.
This ontology had a precise critical dimension in its historical moment. It responded to an art economy based on scarcity, on the idea that a work is valuable for its uniqueness, for the physical presence of an authenticated original, for the signature that guarantees contact with the artist. The post-internet image-object dynamited this economy not through a declared political thesis, but through a description of what was actually happening: works circulated, modified, and proliferated in networks independently of any economy of scarcity, and this circulation was the real mode of existence of the contemporary image.
It is necessary to emphasize what this thought presupposed without saying it, because it is precisely this presupposition that the regeneration of archives fractures.
Vierkant presupposed that there were images to distribute. Images that had been produced, that existed in an initial state, even if that state was worth no more than subsequent ones. The image-object is born somewhere, at some point, in an act of production. It then enters circulation and transformation. Vierkant’s question is entirely localized in this after: what happens when the image circulates? He does not ask the question of the before: of what has not yet taken place, of what should have been produced and was not.
More precisely: Vierkant presupposes that the act of production is accomplished. The image has been made. It exists. It enters the flows. What the flows do to it—that is the question. But he does not ask the question of the image that was not made. Of the image that wanted to be made and that material conditions prevented from existing. Of the image that remained in the archives in an insufficient form—a trace of what should have been produced, not an accomplished production.
The image-object is a thought of the image’s present, of its life in networks, of its contemporaneity with itself in its multiple states. My current practice is a thought of the image’s unaccomplished past. It is not the same question. It is not the same ontological regime.
But to grasp the exact gap between these two regimes, one must first retrace the long history of attempts through which art has sought to accomplish what reality could not or would not be. This history is older and deeper than is generally believed. Regeneration is not its rupture: it is its realization.
There is a question that art has asked for at least a century with increasing persistence: what is to be done with the unrealized possible? With the work that could have existed and did not? With the archive that carries within it something it does not show? What is the artistic utopia? The historical answers to this question form a genealogy that I want to trace here, not to position myself in a lineage, but to show what each answer failed to accomplish, and why it is precisely this lack that regeneration comes to fill.
The first answer is that of the assumed fake, the aesthetic hoax. Joan Fontcuberta is the most elaborate example. With Fauna (1985) and Sputnik (1997), he manufactures complete and indiscernible archives—scientific files, photographs, records—of animal species and a Soviet cosmonaut that never existed. The archive is perfect. It produces a total reality effect. But Fontcuberta always maintains, deferred and shifted, the possibility of revelation. There is a moment where the hoax declares itself. The suspension of the boundary between true and false is temporary, aesthetically controlled, intended to produce a critical effect on our regimes of belief. What Fontcuberta does is show that the fabricated archive can be indiscernible from the real archive, drawing an epistemological lesson on the fragility of our relationship with documentary truth. This is not what I am doing. My regenerated archives are not fakes. They do not seek to be taken for real documents. They do not prepare a revelation. They are afactual in the sense that they have suspended the question itself, not as a rhetorical strategy, but as the ontological condition of what they are.
The second answer is that of the fabricated emotional archive. Christian Boltanski is the central figure here. In his Archives, his Inventories, his installations of blurry photographs and handwritten records, he constructs memories of people who perhaps never existed, or about whom we know nothing, or whose traces have been deliberately made illegible. What is unsettling in Boltanski, and what brings him closest to what I am trying to describe, is that the memory effect does not depend on the reality of what is archived. The trace produces the feeling of the trace even when it traces nothing verifiable. He shows that the archive functions by its form before it functions by its content, that the memorial emotion is triggered by formal signals independently of their referent. This is a thought very close to afactuality. But Boltanski does not start from his own archives. He manufactures generic, universal archives based on the formal codes of collective memory. What he produces is a phenomenology of the archive, not a regeneration of his own traces. The difference is decisive: in my gesture, the archives that return to me transformed are mine. They carry a singular, irreplaceable memory that cannot be substituted by generic archives. What the network accomplishes is the precise intention of these specific archives, not the general form of the archive.
The third answer is that of chronological falsification. Giorgio de Chirico, after 1920, spent decades remaking his own metaphysical canvases from the 1910-1917 period, backdating them. He claimed these refabrications were “truer” than the originals because he now mastered the technique better. This is a claimed loyalty to intention against the event—exactly the formula I use to describe my own gesture. De Chirico claimed to do what reality had prevented him from doing the first time: to produce the paintings he wanted, not the paintings he was able to make with the means and mastery of his time. But this gesture was corrupted by two things. First, economic interest: he remade his works because they sold, not because an unaccomplished intention called him to do so. Second, falsification: he backdated, he lied about the origin, he sought to pass his refabrications off as originals. This is not a suspension of the question of truth; it is a manipulation of truth within the very regime he claimed to transcend. My gesture does not backdate. It does not falsify. It does not seek to pass what is produced off as what took place. The regenerated images stylistically bear the mark of their own present. What they accomplish is not a lie about the past, but a correction of the past from the present—a correction that does not hide the fact that it is a correction.
The fourth answer is that of adversarial counter-factuality. Duchamp, Rauschenberg erasing a De Kooning drawing: gestures that turn the archives of art against themselves, making them say something other than what they intended to say. These gestures were legible, declared, and bore the mark of a sovereign subject who had chosen to intervene. The distance between the trace and the manipulation remained visible. One knew there had been an artistic decision. What these gestures did not do was remain loyal to the original intention. They deliberately betrayed it. Adversarial counter-factuality is a critical thought of the archive: it seizes it to subvert it. What I do is fundamentally different in its orientation: I do not turn my archives against themselves. I accomplish them. I am loyal to them, even in their betrayal by factual circumstances.
These four answers all have in common that they approached the question of the unrealized possible without being able to resolve it entirely. Fontcuberta made it an epistemological lesson. Boltanski made it an emotional phenomenology. De Chirico made it a mercantile falsification. Duchamp and Rauschenberg made it a critique of the institution. Each stopped where their technique and problematic limited them. What unites them is a shared intuition: that the archive does not only preserve what took place, but the ghosts of what could not take place. That each trace carries within it the silhouette of what it did not become. That the unrealized possible is a reality in its own right, not a simple absence. What AI allows for the first time is the actualization of these ghosts from the archives themselves—not by speculation, not by falsification, not by the manufacture of generic archives, but by the statistical induction of relationships between the elements of the archives. This is why the gesture is new. Not because it breaks with this history, but because it accomplishes it where it had always stopped.
To name what regeneration produces, we must now specify what it is not in the exact terms this history suggests. It is not a restoration. Restoration assumes a degraded original state that one attempts to recover. It is loyal to the event. What I do is not loyal to the event. What took place during those shoots—the poor lighting, the bad framing, the insufficient means—was already a degradation of the project, not an original state to be respected. It is not a hoax à la Fontcuberta. The hoax maintains the distinction between true and false the better to denounce it. What I produce does not prepare a revelation. The suspension of the boundary is not temporary. It is not a generic emotional archive à la Boltanski. The archives that return to me transformed are mine. They carry a memory that nothing can substitute. It is not an adversarial counter-factuality à la Duchamp. I do not turn my archives against themselves. I am faithful to them.
What it is, is what I have named instauration: not a return to an original state, but the production of missing elements that never existed but which the archives called for. When the Lumière brothers’ film is processed by a neural network to reach 4K definition at 60 frames per second, it is not a restoration because no original exists at that definition. It is an instauration: elements are generated that were never captured, and the result is unsettling precisely because it no longer has the realism of 1895; it has the grain of a 1971 video. Instauration produces an anachronistic realism. It is not a lie about the past: it is the realization of a possible that the past contained without being able to accomplish.
To map the possible relationships between a generated image and the reality it alters, I have proposed four modalities of counter-factuality ordered by their increasing distance from what took place.
Completion: one enriches, repairs, or adds what is missing to an existing document. The network infers from the latent space what the image did not show. The notion of the archive is troubled because what is added did not pre-exist anywhere; it is statistically induced, belonging to a world of images absorbed by the network and redistributed into the document’s void. This is the modality closest to instauration.
The Alternative: images that resemble without reproducing, that remind without copying. Credible versions of what could have happened. The recursion of the possible produces a strange situation: one manufactures evidence of what did not take place, exhibits for events that never happened, counter-factual facts that trouble identification procedures.
Metamorphosis: categories mix, realms interpenetrate, the separations that define things undo themselves in the generalized analogy of the latent space. This is what I have called decategorization—an involution in the very formation of percepts, where the mineral becomes organic, the landscape becomes body, history becomes fiction.
And Afactuality. This is the deepest modality, and it is what defines what the regeneration of archives produces at its core. Afactuality is not the negation of factuality. These images are indeed factual: they exist as pixels, as files, as material artifacts. Afactuality is an indifference to the distinction between what is and what is not. A suspension. The afactual image does not ask the question of its truth. It has dissolved the question. Boltanski approaches this when he shows that the memory effect functions independently of the referent. But Boltanski does it consciously, as a declared artistic strategy. The afactuality of my regenerated archives is not a strategy: it is the result of a technical process that produces images that never had to decide whether they were true or false.
This is where the fracture with Vierkant deepens until it becomes abyssal. Let us return precisely to what Vierkant says and does not say. The Image Object Post-Internet is a text that theorizes the life of images in circulation, their transformation by display conditions, their existence in multiple equivalent states, and the impossibility of designating one state as more authentic than others. This theorization is just and rigorous in its domain. But it presupposes a regime of truth that it does not question: the regime in which one can ask if a certain version is closer to the original than another. It answers: the question is ill-posed, there is no original. But it maintains the framework of the question—the framework in which images circulate, transform, and can be compared with one another. Vierkant dissolves the original into circulation. This is a dissolution that operates inside the regime of truth; it says the question of the original is ill-posed, that the concept is insufficient. But it leaves something essential intact: the idea that there was an event of production, something that took place and is now circulating. The image-object is always the trace of an act of production, even if that trace is no longer “originary.” There is a moment zero, even if that moment is worth no more than subsequent moments.
Afactuality goes further and in a different direction. It does not displace the regime of truth by saying the question of the original is ill-posed. It suspends the regime of truth by producing images that never had a moment zero from which to compare. The afactual image is not a version of an image that would have first existed elsewhere. It is the primary actualization of a possible that had never been actualized. Between it and what took place, there is no relationship of derivation, reproduction, or transformation in circulation. There is a relationship of obligation: the obligation of the intention toward itself, the debt of art toward what it wanted to be. This relationship of obligation is what Vierkant could not see from his theoretical position. He looked at images in their states of circulation; he observed what happened to them after their production. He did not look at what precedes production: the tension between what wanted to be done and what was able to be done—the gap that economic conditions carve between intention and archive. This gap does not exist in Vierkant’s world because in that world, the image has been produced. It is there. It circulates. The question is what it becomes. In my world, some images are not there. They wanted to be there and they were not. These are not absences; they are deferred presences, possibles in waiting. And it is precisely this delay that regeneration comes to fill.
To understand why this filling is ontologically new even in the context of the long history I have traced, one must descend into what AI produces as real structure.
In a neural network, reality is twofold: it is the dataset—the images that fed the training—and it is the latent space that this training constituted. The latent space is not a warehouse of images. It is a space of probabilistic relationships between vectorized elements, reduced to series of coordinates in a space of thousands of dimensions. A face, a word, a texture, a light: all are points in this space whose distances define resemblances and analogies. What is generated is not extracted from this space as one extracts an object from a drawer. It is an actualization, a point of crystallization in a continuous flow of possible transformations, an event that emerges from the structure without being contained in it as a preformed image, because the constitutive resemblance of motifs in a latent space is not identical to the elements we distinguish in an image.
This space is non-totalizable. No finite set of generated images exhausts it. There is always a reserve of as-yet-unactualized possibles, an infinity of configurations waiting for their prompt. This is an objectively mediated contingency: actualizations are not determined by the network’s structures, but they are not random either. They depend on choices of exploration, the words chosen, the direction given to the drift, because the system of attention reorganizes the set of vectors. They are contingent relative to these choices, but objective in their structure.
When I feed the network only my own archives, I construct a singular latent space. Not the general latent space of a model trained on billions of images, but a constrained space, oriented and haunted by the relationships between my own data. The images that emerge from it do not circulate in the same flows as Vierkant’s image-objects. They are not versions of an image that would have first existed in an initial state. They are the primary actualizations of a possible that belongs only to this singular space. This is where the comparison with Vierkant becomes most instructive. For Vierkant, the space of circulation is common to all images; it is the space of digital networks, platforms, and flows. All images circulate there according to the same logics of modification and equivalence. The latent space of my archives is the opposite: it is as singular as possible, as un-common as possible. It is constituted only by the relationships between my own data. What emerges from it cannot “circulate” in Vierkant’s sense—not because it is a unique work in the traditional sense, but because its singularity comes not from physical uniqueness, but from statistical singularity: this latent space belongs only to these archives, and the actualizations emerging from it could have emerged from no other space.
The latent space fundamentally changes the stakes: the unrealized possible is not filled by the artist but revealed by the statistical structure of the archives themselves. What emerges is not what I “wanted to put in”; it is what the archives contained without being able to show. It is not my interpretation of my past intentions. It is the implicit structure of my archives actualizing itself.
This is where the Meillassouxian dimension of contingency becomes relevant, not as an ornament but as a precision. The contingency of the latent space—the fact that one can never say that this image rather than another was necessary, that the actualization is real but was not inevitable—is not algorithmic randomness. It is an objectively mediated contingency: relative to the structures of the network, conditioned by the dataset, but never reduced to necessity. And this contingency has a property that previous practices could not produce: what emerges was not predictable, but it was contained in the archives. The artist recognizes it without having foreseen it. There is a truth that precedes the consciousness one has of it, and it is the very truth of the possible against the indexicality of the real.
This regime of images—which derive from no trace but are more faithful to the intention than the traces were—is what I have named disrealism. Not fiction, which invents freely. Not photorealism, which reproduces with precision. A mode situated at the frontier between the known and the unknown, between what was and what never existed, producing an uncanny familiarity: one recognizes something that never took place.
Disrealism is post-post-indexical. Vierkant operated in a post-indexical regime: he took note of the failure of the indexical sign in digital circulation and theorized what remains when the imprint is no longer a guarantor of anything. But his world was still populated by images that had been made, which carried—even if dissolved in the flows—the mark of an event of production. Disrealism radicalizes this degradation to its point of inversion: it produces images that have no trace of an event of production, that were imprints of nothing, and yet which carry a memory more precise and more true than any photograph.
This is no longer the same crisis of representation. The post-internet crisis was a crisis of the original: what remains when everything is equivalent in circulation? The disrealist crisis is a crisis of the trace: what remains of the trace when that which has no trace is truer than that which does? The first dissolves the aura in the Benjaminian sense; it shows that technical reproduction is not the degradation of an original but its normal mode of existence. The second dissolves the very necessity of contact, of the imprint, of physical contiguity. It shows that fidelity to intention does not need to have “taken place” to be real.
Vierkant wrote, in a formula that summarizes his entire position: the post-internet image-object “exists as both presentation and representation.” Presentation is the concrete display in a given context; representation is the persistence of the image beyond that display. This simultaneous double existence was what defined the image-object: neither pure physical presence nor pure abstract reproduction, but both at once in every state. It was an ontology of simultaneity. The disrealist image is neither presentation nor representation in Vierkant’s sense. It is not the presentation of one state among other equivalent states. It is not the representation of an original that would have preceded its reproductions. It is the advenance (the arrival/happening) of a possible that never had a prior state. An ontology no longer of simultaneity, but of primary advenance.
What this implies for the relationship to one’s own past has an economic and political dimension I do not wish to evade. There is an art economy that determines what can be produced. Production budgets define what enters the real, what remains in the project, what dies in the archives. This economy is not neutral: it selects, filters, and orients according to logics that have nothing to do with artistic intentions. What it leaves in the archives is not the failure of art; it is art amputated of its means. The indexical traces I regenerate are the scars of this amputation.
De Chirico, in his corruption, at least understood one thing correctly: that his late refabrications were the response to an economic injustice. His early canvases had been sold cheaply, misunderstood, and dispersed before he could control them. He sought to repair this injustice through falsification. The method was bad. The intuition was right: the art market economy confiscates the artist’s possible; it steals the work before it can be finished.
Regeneration is the gesture by which I twist economic reality so that it finally joins art. Not through falsification. Not through mercantile reclamation. But through the deferred accomplishment of what the economy had prevented. And this accomplishment is possible only now, only through this technique. It is not a revenge. It is a correction that time and technique have made possible.
In working this way, I am in a troubling relationship with my own past. Not because I falsify it. But because I discover retrospectively that the archive was a call. It called, since the beginning, for its own regeneration. It contained, encoded in its very limits, the silhouette of what it could not yet show. Every poorly lit image carried within it the lighting it wanted. Every shaky framing carried the framing it sought. These intentions were not ideas in my head: they were materially inscribed in the defects of the trace, in the gap between what had been captured and what should have been.
This material inscription of intention in the defect is what fundamentally distinguishes the gesture of regeneration from all previous forms of counter-factuality. Fontcuberta read the formal codes of the archive to reproduce them. Boltanski read the emotional effects of the archive to simulate them. De Chirico read his own past intentions to remake them consciously. In all these cases, the reading was conducted by the artist—conscious, interpretative. The network reads differently. It does not read the meaning of the archive; it reads its statistical structure. It encodes the relationships between pixels, textures, orientations, recurrences. And in these relationships, what was intention finds itself encoded without having been spoken. The latent space of my archives contains a memory of the project that I could not have formulated myself. What the network actualizes is this implicit memory, this underlying structure that was there from the beginning in the archives without ever being able to show itself.
AI does not give me a new power over my archives. It reveals to me the power that the archives had always exercised over me—this unfulfilled vow that every indexical image carried without being able to formulate it. And by revealing this power, it retrospectively changes the status of the archives: they are no longer tombs of what took place. They are organisms in waiting. Calls without answers that were waiting for their technique.
This relationship between me, my archives, and the latent space is not one of mastery. I do not determine what images emerge. I do not know exactly what I want before I see it. It is not the network producing autonomously; it produces nothing without the constraint of my data and the orientation of my choices. What produces the images is the contingent relationship between this data, this architecture, and this exploration. A co-constitution whose meaning emerges in the aftermath, when I look at what has come to be and recognize, with a shock, sometimes, what I wanted without knowing it.
This is perhaps the most fundamental difference from Vierkant and the entire genealogy I have traced. In all previous cases—Fontcuberta, Boltanski, De Chirico, Duchamp—there was a subject who preceded the work, who willed it, who decided upon it, even if that subject did not entirely control what emerged from it. Regeneration produces a subject that comes after the work. Who recognizes themselves in what has emerged without having foreseen it. This is not the dissolution of the subject in the flows in the post-internet manner. It is the birth of the subject in the aftermath of what the relationship between their archives and the network has produced. This reversal of the temporality of the subject relative to the work has no exact precedent in this history. It is what regeneration adds to the question art has asked for a century about the unrealized possible.
It is no longer the image as an object in circulation. It is no longer the document. It is no longer fiction. It is no longer the hoax. It is no longer falsification. It is the image as a realized haunting, a ghost taking body, a possibility actualizing itself in the aftermath, an intention finally finding its matter, an archive retrospectively accomplishing what it wanted to be. A haunting different from the one Derrida described: not a revenant (one who returns), but an advenant (one who arrives). Not what returns from what took place, but what arrives for the first time from what wanted to take place.
Vierkant dissolves the original in circulation and poses: all versions are equal. Fontcuberta dissolves the distinction between true and false and poses: belief is a construction. Boltanski dissolves the referent and poses: memorial emotion precedes its object. De Chirico dissolves the boundary between the work and its refabrication and poses: intention is worth more than the event, even when corrupted. Duchamp dissolves the boundary between art and non-art and poses: it is the gesture that makes the work.
Regeneration dissolves the boundary between the trace and the possible, between what took place and what wanted to take place, between the archive and the work, between reality and its unfulfilled power, between what the economy permitted and what art demanded. And by dissolving this boundary, it does not pose a thesis. It opens a question that its predecessors could not yet formulate: no longer “what is true?”, nor “what is original?”, nor “what is art?”, but “what was owed?” What did the archive owe to the intention it carried without being able to accomplish it?
This question of debt—the debt of reality toward art, the debt of the event toward intention—is perhaps the deepest political question that regeneration poses. And it can only be asked now, through this technique, from this moment where the latent space of my own archives reveals to me, in the aftermath, what I wanted without having known it.