Critical Essay · Bitcoin Ordinals · Generative Art
On Programmed Oblivion, Neural Compression, and the Permanence of Disappearance
There exists a category of artworks that do not ask to be looked at but to be lived with — works whose full meaning only unfolds across time, works that demand the patience of a witness rather than the attention of a spectator. Memory Loss by EP belongs unambiguously to this category, and in doing so it occupies a singular position within the emerging landscape of on-chain generative art: it is not a collection of images but a collection of processes, each one a slow and irreversible dissolution that will outlast its creator.
The project begins with an act of radical intimacy. personal photographs — private memories selected for their biographical weight — are subjected to a programmed degradation that will unfold over one hundred years, measured not in calendar time but in Bitcoin block cycles. Each of the 250 unique inscriptions is a distinct interpretation of one of these memories, generated deterministically at the moment of inscription and left to erode according to rules encoded permanently in the blockchain itself. There is no update, no correction, no salvation.
What results is a paradox so precise it feels inevitable: works that are technically immortal and visually dying; images that are immutable in their mutation; testimony to events that are historically true and perceptually becoming inaccessible. Memory Loss does not represent the corruption of memory — it enacts it.
Memory Loss situates itself within a nascent artistic movement termed Compressionism — a framework that treats the computational processes of data compression not as technical necessities but as expressive languages capable of carrying philosophical and psychological meaning. At its core, Compressionism posits that the operations performed on digital information — pixelation, dithering, desaturation, quantization, lossy encoding — are not merely formal operations but structural metaphors for the way biological memory works.
The alignment is not superficial. Neuroscience has established that human memory is fundamentally reconstructive rather than archival: we do not store experiences like files on a hard drive and retrieve them intact. Instead, each act of recall is an act of recomputation — the brain reassembles a version of an event from incomplete fragments, introducing distortions, filling gaps with inference, and overwriting the previous reconstruction with the new one. Memory, in this sense, is a lossy compression algorithm that runs every time we remember.
Digital compression processes are not metaphors borrowed from neural deterioration. They are, structurally, the same operation performed on different substrates.
Compressionism takes this equivalence seriously as an artistic premise.Memory Loss foregrounds the process itself. The artifact is not the image as it initially appears; it is the trajectory of its degradation. The work exists most fully not at genesis but across time, in the accumulated distance between what the image was and what it is becoming.
In Memory Loss, this framework is deployed with exceptional rigor. The twenty-one dithering algorithms employed are not interchangeable stylistic filters — each one encodes a different failure mode of memory. Floyd-Steinberg diffusion scatters information laterally, like the way peripheral details of an event bleed into adjacent recollections. Bayer matrix patterns impose a rigid geometric structure on visual noise, analogous to the mind's tendency to impose narrative order on fragmentary experience. Cross-hatch and grid algorithms break continuous surfaces into discrete units, reflecting the compartmentalization that occurs when memory is retrieved in isolation from its original emotional context. Each algorithm is, in the language of Compressionism, a neurological process — a specific way of remembering badly.
The conceptual architecture of Memory Loss rests on a fundamental observation: the same events, remembered at different moments, produces radically different images. Collective memory is a fiction assembled from irreconcilable individual ones. What we call "shared history" is an agreement to act as though our respective corruptions align.
The project materializes this observation through its generative system. Personal photographs — specific, biographical, irreplaceable — are each subjected to a unique combination of:
Generative Variables
Each of the 250 resulting works is a different version of the same set of memories . The dithering algorithm determines how information is reorganized under pressure. The color palette determines the emotional register through which the past is being accessed. The color comparison method determines the perceptual framework through which similarity and difference are evaluated in the act of recall. Together, they constitute an individual subjectivity.
This is not a metaphor in the decorative sense. The 250 inscriptions are not 250 representations of forgetting — they are 250 instances of it, each one distinct, each one running on its own deterministic trajectory, each one moving toward a different shade of dissolution.
The thirty-five chromatic systems in Memory Loss are organized into six rarity tiers, drawing from historical art movements — .
Palette Rarity Hierarchy
Every aspect of Memory Loss that appears subjective — which algorithm processes a given image, which palette filters its colors, how far its degradation has progressed — is in fact fully deterministic. There is no randomness in the conventional sense. The apparent multiplicity of the collection is the product of a single generative formula applied to different seeds.
Deterministic Seed Construction
Each inscription's unique visual identity derives from a PRNG seed built as:
seed = inscription_id + "04022000"
where 04/02/2000 is the artist EP's date of birth. The inscription ID — the unique identifier assigned by the Ordinal protocol at the moment of inscription — provides the irreducible specificity of each work. The birthdate provides a biographical constant: the shared origin point of all the memories being processed.
This construction ensures that every viewer, on any device, at any moment, sees exactly the same image. The work is not a rendering — it is a calculation, repeatable and verifiable to any party with access to the inscription ID.
The degradation of each image follows a strict sequential pipeline in which the order of operations is not interchangeable but constitutive of the result. The stages are:
1. Pixelation. The source image is first reduced to a grid of enlarged pixels, the size of which increases with each degradation cycle. At genesis, pixel size begins at 3×3 pixels — barely perceptible, a subtle coarsening of the image surface. At the end of one hundred years, pixel size reaches 32×32 pixels: the original photograph has become a mosaic of color blocks in which no individual feature remains legible. The image is still there. The memory it encodes has become inaccessible.
2. Brightness and Contrast Adjustment. Before dithering is applied, each pixelated image is processed according to its assigned brightness and contrast parameters, derived deterministically from its seed. This step ensures that the palette quantization in the subsequent stage operates on a consistent tonal range.
3. Dithering. The palette-quantized image is then subjected to the dithering algorithm assigned to that inscription. The algorithm determines how the error introduced by color quantization — the difference between the original color and the nearest palette color — is distributed across neighboring pixels. Different algorithms produce fundamentally different patterns of this error distribution, resulting in radically different visual textures even when applied to identical source images with identical palettes.
Temporal progression in Memory Loss is measured exclusively in Bitcoin blocks. One degradation cycle corresponds to 52,560 blocks — approximately one calendar year, given Bitcoin's target block time of ten minutes. Over one hundred cycles, the project spans 5,256,000 blocks.
Each degradation cycle advances the pixelation parameter by one step and applies a corresponding deepening of chromatic desaturation. Colors progressively converge toward one of four terminal palettes — each representing a distinct stage of cognitive loss, from early distortion to terminal erasure. The final palette state is not grey silence but a specific chromatic condition: a color of forgetting, distinct for each inscription, derived from the same seed that determined everything else.
The genesis block height — the Bitcoin block at which each inscription was recorded — is locked permanently into the calculation. This means that two inscriptions created at different moments will be at different stages of degradation when viewed simultaneously, even if their other traits are similar. The blockchain timestamp is not metadata; it is a structural parameter of the work.
The technical substrate of Memory Loss is the Bitcoin Ordinals protocol, which enables the inscription of arbitrary data onto individual satoshis — the smallest unit of Bitcoin — creating digital artifacts that inherit the immutability and decentralization of the Bitcoin network itself. Within this protocol, Memory Loss employs two structural mechanisms — recursion and the Parent-Child inscription system — that are not merely technical optimizations but conceptual elements that deepen the project's meaning.
Human memory is recursive by nature. When we recall an event, we do not access a stored recording of the event itself; we access the memory of the last time we recalled it. Every act of remembering is an act of reconstruction that overwrites its own source. We remember to remember, and in doing so we rewrite what we remember. The past is not retrieved but recomputed.
The code of Memory Loss replicates this structure technically. Each inscription does not self-contain all necessary visual data. Instead, it references external resources — typographic files, foundational visual components — through Ordinals recursion endpoints. The inscription calls outward to the protocol to assemble itself, just as a memory calls outward to other memories to reconstruct an event. The technical recursion is not a workaround; it is an architectural decision made in full awareness of its metaphorical resonance.
Additionally, the code retrieves live Bitcoin blockchain data at render time — current block height, network statistics — to calculate the precise degradation state of the image in the moment of viewing. The work exists in a state of continuous present: it is always now, always computing its own distance from its origin.
The most structurally innovative — and philosophically charged — element of Memory Loss is its treatment of the collector's role through the Ordinals Parent-Child inscription system.
At any moment during the one-hundred-year duration of the project, a viewer may press the P key while viewing an inscription to export a high-fidelity SVG file representing the current state of the degrading image. This file can then be re-inscribed on the Bitcoin blockchain as a Child of the original inscription, creating a permanent, verifiable Parent-Child relationship: two inscriptions linked on-chain, one dying, one preserving a moment of the other's life.
The child inscription does not arrest the parent's decay. It witnesses it — and in witnessing, changes the probability of what comes next.
The protection mechanism operates through a probabilistic system that is itself deterministic once the input parameters are known. Each Parent inscription queries the Ordinals protocol in real time for the count of its registered Children. This count — retrieved through the recursive endpoint /r/children/ — determines the probability that the inscription will skip its next degradation cycle:
Protection Probability by Child Count
The protection probability never reaches 100%. No accumulation of Children can arrest the degradation entirely. The system encodes an ontological position: care can slow loss, but it cannot prevent it. The maximum protection of 10% per cycle means that even the most actively preserved inscription will, in expectation, complete its degradation within the hundred-year span.
This mechanism transforms the act of collecting into the act of witnessing. The collector who creates a Child inscription does not own a static object — they enter into a relationship with a process. They become a participant in the duration of the work. The SVG they inscribe is a timestamp, a testimony: I was here, and this is what I saw.
The resulting on-chain structure — Parent inscriptions surrounded by constellations of Child inscriptions, each Child a moment of arrested decay — constitutes a collaborative archive that no single actor controls. It is a social memory system built on the most censorship-resistant infrastructure in existence.
The central tension of Memory Loss — the paradox that gives the project its philosophical weight — lies in the contradiction between the absolute permanence of its substrate and the programmed impermanence of its content. Each inscription is immutably encoded on the Bitcoin blockchain: it cannot be altered, deleted, or censored as long as the network operates. And yet the image it contains is designed to become, over the course of one hundred years, visually indecipherable.
The code will persist. The calculation will remain correct. The blockchain will faithfully record every parameter of the degradation formula. And the formula will faithfully produce, cycle after cycle, an image that looks increasingly unlike the memory it encodes. The truth of the past moment — the actual event that the photograph records — is immutable in its historical factuality. But access to that truth, through the image, is being systematically foreclosed.
This is precisely the condition of memory. We do not doubt that our past happened. We doubt our ability to reach it accurately. The events are fixed; our perception of them is not. Memory Loss encodes this epistemological problem into blockchain infrastructure, making it not a statement about memory but a working model of it — a mechanism that produces, in real time, the experience of knowing that something is true while being unable to see it clearly.
There is no redemption arc in this structure. The project does not hold out the possibility that memory might improve, that degradation might reverse, that the image might crystallize back into clarity. The algorithm moves in one direction. The blockchain moves in one direction. Time moves in one direction. The future is predictable, and it ends, always, in dissolution.
Memory Loss is not a meditation on death. It is a demonstration of it — patient, precise, and indifferent to whether anyone is watching.
Situating Memory Loss within art history requires identifying the traditions it inherits and the point at which it departs from them. The project draws from at least four distinct lineages: process art and duration-based practices; conceptual art's privileging of idea over object; net art and its interest in network infrastructure as medium; and the emerging field of on-chain generative art that has developed alongside the Ordinals protocol.
From Compressionism — the movement that positions digital compression algorithms as the primary expressive language of neurological deterioration — Memory Loss inherits its central equation: that pixelation, dithering, and desaturation are not aesthetic choices but structural replicas of how biological memory fails. Where Compressionism as a broader movement explores compression aesthetics across many contexts, Memory Loss commits to its most radical proposition: that a compressed image and a degrading memory are not analogous but identical in structure, differing only in substrate. The artwork does not illustrate this claim — it enacts it.
From process art — the tradition running from Hans Haacke's biological systems through the durational performances of the 1960s and 1970s — Memory Loss inherits its insistence on time as a constitutive dimension of the work rather than a context in which the work sits. The work does not exist at a moment; it exists across moments, and its identity at any given moment is defined by its distance from its origin and its distance from its terminus.
From conceptual art, the project inherits its structural precision and its willingness to subordinate visual appeal to conceptual necessity. The dithered, pixelated images of Memory Loss are not beautiful in any conventional sense — at advanced stages of degradation they are, deliberately, barely images at all. The work asks to be understood before it is appreciated, and positions understanding as the more durable form of appreciation.
From on-chain generative art, Memory Loss takes its technical substrate and its understanding of inscription permanence as a medium-specific condition. But it departs sharply from the dominant conventions of that field — the celebration of generative beauty, the emphasis on visual richness at genesis, the implicit assumption that an inscription is most fully itself when it is freshest. Memory Loss inverts this assumption: the work becomes most fully itself only through loss.
For institutions engaged with questions of digital preservation, blockchain culture, and the intersection of technology and mortality, Memory Loss offers an unusually coherent object of study. It is simultaneously a technical artifact, a conceptual proposition, a biographical document, and a live experiment in the sociology of care — the question of who, if anyone, will act to slow the loss of another's memories, and why.
What Memory Loss ultimately proposes is a new category of artwork — one for which existing critical language is only partially adequate. It is not a static object that can be catalogued and stored. It is not a performance that ends when the artist leaves the room. It is not a generative system that produces infinite novelty. It is a finite process with a known beginning and a predictable end, operating on a substrate that will persist after both the artist and any living witness have died.
The Bitcoin blockchain will record this process faithfully and permanently. It will preserve the code that calculated the degradation, the block height at which each inscription was created, the exact parameters of every palette and every algorithm. It will preserve everything necessary to understand how the disappearance worked. What it cannot preserve — what no technology can preserve — is the memory itself. That was always going to be lost.
Memory Loss is therefore not a failure of preservation. It is a monument to the impossibility of it — inscribed permanently, for anyone who arrives later to read.