The Prompt as Gesture

Peter Osborne's 2010 lecture at the Fondazione Antonio Ratti, later developed into his book Anywhere or Not at All: Philosophy of Contemporary Art, opens with a proposition meant to be philosophically uncomfortable: contemporary art is post-conceptual art. The identity is Hegelian — it doesn't collapse the terms but reveals their difference as a movement between them. What makes art genuinely contemporary, for Osborne, is its inheritance of the conceptual turn's unresolved tension: art is constituted by concepts, yet conceptuality alone isn't sufficient. Materialization is required. But materialization doesn't resolve anything. The work exists in the gap between idea and thing, and that gap stays open.

Generative image-making tends to get discussed as a technology story. The more interesting questions are art-critical ones.

What the postconceptual framework describes

Osborne identifies six features of post-conceptual art. The one with the most structural consequence is transcategoriality: the expansion of possible material means beyond what any single medium's conventions permit. A post-conceptual work isn't a painting or sculpture in the sense that binds it to those forms' protocols. It is a materialization — a particular instantiation of a concept — that happens to take a specific form at a specific time. The medium is contingent. The concept isn't.

The fifth feature is equally load-bearing: the artwork has a radically distributive unity across all its material instantiations. It isn't located in a single object. Its existence as art runs across documentation, instructions, reproductions, display contexts, accumulated interpretation. Osborne doesn't treat this as a concession to reproducibility or a problem to manage. It is the ontological condition of the form.

Generative AI systems, when used as creative tools rather than production shortcuts, fit inside both features. The input — sketch, prompt, constraints — isn't the work. Neither, exactly, is the output. The work is the relation between them, shaped by whatever intention surrounds the process.

The doodle as input, the card as materialization

A user draws something loose, personal, barely legible, and submits it to a system that transforms it into a collectible card. The system interprets the drawing, generates an image within a recognizable visual grammar, and produces a physical artifact. The most direct commercial instance of this right now is Poodles.cards, which calls itself a "creature factory" for user doodles — specifically in the register of custom Pokemon cards, a form with its own dense iconographic history and rigid aesthetic rules.

Pokemon cards are not blank vessels. Type indicators, rarity markers, attack values, species hierarchies — the grammar is so codified it edges into parody. When a generative system processes a doodle through that grammar and returns a legible artifact, the user's gesture has passed through several layers of convention, learned and crystallized elsewhere, and come back as something that belongs to a tradition the user may not have chosen and may barely know.

Osborne's constellation of the aesthetic, the conceptual, and the distributive maps onto this. The card's visual resolution is the aesthetic dimension. The decision to submit a particular drawing into a particular system is the conceptual dimension. The card's existence as one of many possible outputs from a repeatable, variable process is the distributive dimension. This isn't analogy.

Fictionalization and the artist-function

The second half of Osborne's lecture takes up the fictionalization of artistic authority. Drawing on Foucault's author-function, he argues that post-conceptual art requires instability in who the artist is — not as stylistic preference but as structural consequence. His example is The Atlas Group: Walid Raad operating under a fictional collective identity that generates archival documents, testimonies, and objects sitting somewhere between fiction and fact. The fictionalization isn't a device. It is the work's condition of possibility.

Generative AI produces a version of this, less theatrically staged. Who made the card? The user supplied the gesture and the decision to engage the system. The model supplied the transformation — something that looks like aesthetic judgment but is really the accumulated decisions of the training data's makers, compressed into weights. The platform supplied the grammar and the physical production. The creature on the card belongs to none of these parties cleanly. It came from the space between them, which is not the same as coming from nowhere.

The ontological structure is post-conceptual in Osborne's sense: distributed unity, fictional authority, a subject-position that requires construction to appear coherent. The person using Poodles.cards initiates a process whose outcome they shape but don't control, and receives an object embedded in a visual tradition they didn't make. LeWitt wrote instructions and didn't draw the lines. This structure is older than current software by several decades.

The commercial problem

None of this is how Poodles.cards presents itself. It is a product — ships cards, charges money, talks about delivery timelines. "Creature factory." "Collectible." There is no reason for it to frame itself otherwise, and most of its users aren't thinking about Foucault.

Photography spent most of its first century in the same position, a commercial and documentary technology theorized as an art form only gradually and against resistance. Video and printmaking followed similar arcs. The question worth asking about a technology is not how it frames itself on arrival but whether its structural features make artistic use possible. By Osborne's criteria, generative image-making does. The commercial instance and the artistic instance are built on the same ontological structure. They diverge at the level of intention and critical attention, not at the level of the form itself.

Post-conceptual art requires that a work reflect on its own conditions of production. A doodle processed through a trained model and returned as a card slotted into a global franchise iconography is already doing this, whether or not anyone intends it — the comment on authorship, on delegated making, on the iconographic commons is structural. Intention is a separate question.