After the Original

Art in the Age of the Compiler

Most new art begins as a failed copy. An inspired teenage band hears a record and tries to reproduce it in a garage. They do not own the right guitars. Their amplifier is too small, the drummer cannot quite play the rhythm, and the singer mishears a lyric. They probably lack talent. The room compresses the sound in a peculiar way. What emerges is recognizably descended from the original, but it is not the original. It may be worse. It almost always is. Occasionally, it is much better.

This is not an exception in the history of art. It may be the principal mechanism. A musician imitates a musician who imitated another musician. A painter copies the master but has different pigments, a different hand, and a different patron. An auteur attempts Hollywood with three lamps, six actors, and no money. A provincial misunderstanding becomes a movement. A technical limitation becomes an aesthetic. Who can forget the delicious wonder of B-grade science fiction, or the so bad it is good of For Y’ur Height Only? An error acquires its fans.

Originality is often imitation passing through inadequate instruments. We need a flattering creation myth: the tortured artist stands alone before the void and summons something unprecedented from the interior furnace of the self. Culture, however, is closer to a game of telephone played with minds, bodies, technologies, and materials. The message is repeatedly transmitted, distorted, and made new. Remix. Repair. Replace. Memes, minds, and evolution.

Art advances because the copy is imperfect.

This makes the arrival of generative machines less alien than it first appears. They are prodigious imitators, certainly, but so are we. Mr Ripley would have a multitude of new careers in the age of the LLM. The question is not whether a machine has seen other art before making art. Of course it has. So has every artist who has ever walked into a museum, listened to a record, read a novel, or watched another person move. Has art ever really been created de novo?

The more interesting question is what happens when imitation becomes programmable: when an artistic idea can be separated from any single artifact or performance and expressed as a system capable of producing many of them. This is where the Art Compiler enters.

The artwork is not the artifact

The Art Compiler begins from a simple rejection: the output of an artistic AI should not merely be an artifact that a machine subsequently copies. The fundamental object should be the fabrication score, a sequence of physical operations through which an artwork comes into existence. The AI composes; printers, plotters, paint, materials, and people perform; cameras observe; and the system responds with the next movement. This is a far more consequential idea than art generated by AI.

The astute among you will be exclaiming, “You have just transposed the musical process.” Wait. We will get there.

A generated artifact is an endpoint. A fabrication score is a beginning.

The score might specify a field of lines, a set of printable volumes, a palette, an order of operations, some regions of strict machine precision, and others reserved for human interpretation. It may contain tolerances, permissions, randomness, registration marks, drying times, and instructions about which accidents—happy accidents!—should be preserved. Its intermediate representation can describe not only paths and surfaces, but also materials, human gestures, acceptable variation, and provenance. Maybe someday the robots will learn those gestures too. It is not as though Damien Hirst paints all his own spots.

The same score could be performed by a pen plotter on paper, by a 3D printer producing a relief, by a person with an airbrush, by a robot dragging charcoal across plaster, or by all of them in sequence. It could be recompiled for a wall, a book, a piece of translucent film, or a building. The artwork would no longer be identical to any one of its physical manifestations. It would be something closer to a musical composition.

You cannot systematize art into a process of production, I hear you say. You cannot tell it: allow noise here, but only within these parameters. You can. You might even argue that life and human development are specified in much the same fashion. But with art, you will have the score. You can change it. How do humans edit themselves—pick a different mate? I am sure you could put the original on a blockchain, although I hesitate to say that too loudly in case the crypto bros appear with plans for life-size Bored Apes.

A Bach score is not exhausted by any single performance of it. The notes persist, but the piece changes with the room, the instrument, the conductor, the tempo, the musicians, and the assumptions of the period about how Bach should sound. A performance can be technically impeccable and spiritually dead: another critic-perfect Bach. Another can violate convention and reveal a work nobody knew was there.

Why should physical art not acquire the same structure? A sculpture could be a composition capable of being performed in wood, resin, clay, light, or steel rather than one mute object. A drawing could be a score whose meaning includes the order, pressure, and speed of its marks rather than a fixed arrangement of ink. The implementation of a work need not be treated as mere manufacture. In fact, the machinery used in manufacture will become part of it.

That slight servo wear on the x-axis, putting a consistent wobble into layer 37? Perhaps that is the mark of the original. The Stradivarius of 3D printers. Humans cannot avoid ranking things. We will find a way to have an original.

In such a world, there will be great composers of art and great performers of it. Some people will create astonishing aesthetic systems without ever becoming especially skilled fabricators. Others will become virtuoso implementers, masters of how a score behaves when translated into ink, stone, movement, biological growth, or industrial machinery. A third group will be conductors, assembling humans and machines into temporary orchestras.

One artist may compose a family of possible works, while another may produce the definitive interpretation. The implementation will not be a debased copy of the idea. It will be an argument about the idea.

The original was always overcrowded

This troubles our already fragile notion of authorship, but authorship was never as clean as the gallery label pretended. A great painting may contain the labor of assistants—I see you, Andy—pigment makers, patrons, models, restorers, and the accumulated visual grammar of centuries. The old masters had secret recipes for blue and for the “good red,” which was not a reference to the evening wine. Perhaps a new generation of AI millionaires will revive the occasional commission.

A record contains engineers, session musicians, instrument designers, manufacturers, producers, and the acoustics of a particular room. A film is made by hundreds of people and then attached, by convention and marketing, to one director. You know, the idea for that scene really came from… The solitary author was partly a convenient compression algorithm.

The machine makes that compression harder to maintain because it exposes the layers. Who made a work produced by an Art Compiler? Was it the person who wrote the initial concept, the model that proposed the composition, the engineer who designed the generative system, the person who chose among its candidates, the maker who performed the airbrush gestures, the machine whose tolerances created an unplanned texture, or the critic system—I still do not care what you think—that saw the result and altered the next layer?

The honest answer may be all of them, but not equally and not in the same way. Authorship does not disappear. It becomes stratified. There is authorship of the world, authorship of the rules, authorship of the score, authorship of the interpretation, and authorship of the accident. It is the world’s most complex credit-assignment problem, now with even bigger egos. What is the artistic equivalent of git blame?

This is where the postmodernists will attempt to shuffle back into the room and announce that they predicted everything. They did not. They saw quotation, copies of copies, unstable meanings, cultural fragments, and the theatrical construction of the author. Fair enough; I suppose they had to do something in the postwar years. Yet they mostly imagined reproduction as a crisis of authenticity and treated the copy as evidence that the original had become hollow. The result was often an art of inverted commas: clever, knowing, sterile, and endlessly pleased with itself for noticing that culture refers to culture—then getting meta about commentary on culture being culture. Have you ever read an author you wanted to throw a drink at?

They announced the death of the author and then spent half a century writing densely authored footnotes about the corpse. The joke is that they convinced everyone to read and debate those footnotes as though they were quantum chromodynamics. They saw the museum gift shop. They did not see the machine shop. They probably mocked the machinists as blue-collar too.

The postmodern copy was usually inert. It quoted, displaced, and mocked. The computational copy is reproductive. It can mutate, respond, remember, recombine, and acquire a body. It can be recompiled into a different material, performed by a different agent, and altered by what happened during its previous execution.

Warhol reproduced the soup can. The Art Compiler can create a lineage of soup cans that breed. Burroughs cut up the page. Now the scissors can read.

The earlier avant-garde discovered operations such as sampling, cutting, repetition, and appropriation, but had to perform them with the limited machinery of its period. Cutting up text was done with blades. The serial image required a Factory. The remix depended on records, tape, film, and human patience. We are entering a world in which the operation itself can become intelligent. This is not merely more postmodernism. It is an entire change of state.

When every object is unique

The economics of this world will be peculiar. Industrial production made identical things abundant and unique things expensive. The original always had something the copy did not. Digital production then made identical copies essentially free. Generative production now promises to make difference itself cheap.

A system can produce a million images, each unlike the others. A printer can vary every object. A novel can be rewritten for every reader. Do not get any ideas, Amazon—but did Hemingway A/B test his novels? A piece of music can adjust itself to every room. A physical artwork can be compiled from the same score but incorporate local materials, the gestures of its owner, and the accidents of its fabrication. Everything can be unique, and everything can also, in principle, be duplicated.

This sounds like a contradiction only because we have conflated several kinds of uniqueness. A work may be formally reproducible but historically unique. The same score can be performed again, but the first performance occurred in one place, with particular people, under particular circumstances. That time will never come again.

A physical work may be one of a kind while belonging to an infinitely reproducible family. A digital object may be copied exactly, while the path through which a person encountered and altered it remains singular. You heard that band for the first time, and then you bought the CD, found the vinyl, or added it to a playlist. The copy did not reproduce the encounter.

When uniqueness becomes abundant, uniqueness will cease to be sufficient evidence of value. A billion snowflakes are not a billion masterpieces. Value will migrate elsewhere: to provenance, attention, commitment, judgment, context, and consequence. Humans have to rank things and assign credit. We really are the judgy ape: Homo iudicans meets Homo ludens.

Value will lie in the system that generated the object, in the lineage of decisions behind it, and in the community that treats it as meaningful. It may lie in the right to continue the work, to perform it, recompile it, or add a new branch to its family. Illegal remixing or unauthorized performance of art? Bring it on.

The scarce object may no longer be the artwork. It may be a place in the artwork’s history.

This will be uncomfortable for an art market that has often used physical scarcity as a substitute for aesthetic argument. Number the copies, destroy the plates, issue a certificate, manufacture a little shortage, and allow the price to perform the criticism. When every collector can own a unique version, however, the old scarcity trick becomes less convincing. The serious question will be causal rather than numerical: Who created the system from which this work became possible?

Originality will not mean producing an object with no ancestors. Everything has a lineage; pretending otherwise was always fantasy. It may mean inventing a new reproductive grammar, a system capable of producing a family of works that could not previously have existed. The original artist becomes the founder of a species.

Are we all artists now?

The inevitable—tired, so very tired—slogan will be that everyone is now an artist. It is both true and useless. Everyone with a phone is already a photographer, in the same sense that everyone with a throat is a singer. Access to production matters. It broadens participation, discovers excluded talent, and permits forms of expression that would otherwise remain unrealized. Making an output, however, is not the same as making a work.

The machine abolishes many technical excuses, but it does not abolish judgment. Taste! I hear you cry. Quick, someone find Wilde’s The Critic as Artist. Abundance makes judgment more important. Who has not scrolled through Netflix, adding things to a list, only to realize it is time for bed?

When it is difficult to produce one artifact, production itself appears impressive. When it is easy to produce ten thousand, the artistic act moves into choosing, constraining, connecting, and refusing. The artist of abundance is partly an editor of possibility and a curator of attention.

Artists decide what the system may vary and what it must preserve. They establish which accidents—happy or otherwise—count as information. They recognize when a beautiful result is merely decorative and when an ugly result has opened a door. They give a sequence of objects enough internal pressure to become a body of work rather than a directory full of files. They argue and perform a dialogue with society, independent and uncaring, with the occasional: You really think so?

The great artist may be the person who can say no to 999,999 compelling possibilities and explain, at least to themselves, why the remaining one matters. This is not a retreat to some mystical cult of taste. Taste itself can be examined, trained, modeled, challenged, and perhaps even learned. The existence of machines with astonishing fluency will make the absence of purpose more obvious, not less.

A machine can generate surprise, but it cannot guarantee stakes. A machine can generate novelty, but novelty alone cannot decide what deserves to persist. That decision belongs to a larger system containing artists, audiences, institutions, materials, histories, and consequences. The artist is not necessarily the sovereign of this system, but neither are they merely the operator who types the prompt.

The artist is the person, group, or institution—formed by machines and humans—that makes the work answerable to something.

Art after Playtime

The future artwork may not resemble an object at all. It may resemble Jacques Tati’s Playtime, a vast field of simultaneous action in which no single figure permanently owns the viewer’s attention. A joke happens at the edge of the frame while another story unfolds in the reflection of a glass door. Minor characters become protagonists for a moment. The architecture performs alongside the actors. Viewers discover different films inside the same film depending on where they look.

There is no excuse now for not knowing such classics. It is all there. Do not rewatch the latest Venn diagram of your interests on Netflix.

That may be a better model for computational art than the endlessly generated picture. Imagine an artwork containing a million characters, subplots, objects, and evolving relationships. No viewer experiences the whole thing. A figure who is central in one person’s encounter passes unnoticed through another’s. Stories intersect, separate, and leave physical traces. A printed object in one home alters a narrative elsewhere. A melody heard in one room appears years later as the structural rhythm of a building. The system remembers its previous performances and uses them as material. Maybe the whole system is the artwork.

The artwork becomes a world.

Its composer does not manually specify every event, just as a city’s designer does not choreograph every conversation that will occur within it. The composer establishes laws, affordances, tensions, motifs, and channels through which events can affect one another. The audience does not merely observe the work. It takes a path through it.

Every encounter is unique, yet the underlying world remains reproducible. Different parts can be printed, staged, heard, inhabited, or performed. One participant may experience it as a story, another as a piece of music, and another as a physical sculpture generated from events elsewhere in the system. This is the Art Compiler expanded to the scale of culture.

Instead of compiling a score into a plotted line or printed relief, it compiles an artistic world into whatever form a particular moment demands. The work may become a film for one person, a room for another, a conversation for a third, and a sequence of painted instructions for a fourth. There need not be a privileged center, and there may never be a final version: a continual yet separable sculpture.

The masterpiece of the future could run for decades. It might accumulate characters, performances, interpretations, and physical artifacts. Some branches will die. Others will develop schools of interpreters. Particular performances will become canonical, not because they are the only authentic version, but because someone conducted the possibilities with such force that everyone afterward was challenged by it. This would be a strange combination of cathedral, game, opera, city, and ecosystem. It would also be exhausting.

We are not going to suffer from a shortage of art. We will risk drowning in it. Infinite generative culture can become an immense background sludge: aesthetically competent, perfectly tailored, and instantly forgotten. Every wall could be decorated, every silence scored, and every private preference reflected back at us until taste becomes a padded cell. Every wall in every home gets its perfectly unique IKEA print. Not bad, not good—just unique and okay.

The answer cannot be to retreat to handmade scarcity and pretend the machines never arrived. I see you, “No AI” artists. I love the badges. So good to know badge-making never dies. We must demand more from the systems we make. We should build machines that do not merely satisfy preference but complicate it, scores that reserve space for human interpretation, works that become richer when translated badly, and compilers that treat matter not as an obedient output device but as a collaborator with memory, resistance, and surprise.

The future artist will not necessarily make every mark. They will decide which marks must remain impossible to predict.

The artist as Homo aleator: entropy’s editor. Ludens plays, iudicans ranks, and aleator refuses to let the score be written in advance.

A new kind of original

The old original was imagined as an object without precedent. The new original may be a system with descendants. Its achievement will not be that it cannot be copied. Its achievement will be that copying it creates more difference. It will survive translation into other hands, machines, and materials without becoming identical to itself. It will possess a recognizable identity but no definitive body. It may be the ultimate expression of a memetic organism.

There will be Bachs of such systems: people who establish structures so fertile that centuries of performers can continue finding new works inside them. There will be great implementers whose physical interpretations eclipse the initial score—the Capability Browns of art installation. There will be conductors of humans and machines, virtuoso fabricators, material hackers, editors of sprawling artificial worlds, and artists whose primary medium is the allocation of agency.

There will also be millions of us cutting, quoting, prompting, and compiling: Burroughs with animate scissors, Warhol with a Factory that never sleeps. Most of it will be rubbish. That is not a failure. Most paintings, songs, poems, photographs, and conversations have always been mediocre and dull—the small talk of creativity. Culture does not become worthwhile by preventing bad work. Nothing good has ever come from trying to ban the avant-garde. Culture becomes worthwhile by creating the conditions in which improbable good work can appear.

Art will remain what it has always been: imitation under pressure, passed through the wrong instrument, altered by the world, and occasionally transformed into something nobody intended. Only now the instrument can imitate us back.

That is the wager of the Art Compiler: not to automate the finished object, but to compose the conditions from which unpredictable objects, performances, and worlds can emerge. The score gives the work an identity. Every performance gives it a life.

The original of the coming age will not be the object no one can copy. It will be the world worth entering, the score worth performing, and the idea strong enough to survive its escape.