Structural Omission Broke AI

Structural Omission broke AI. Or at least, it exposed the limits of a system trained to resolve.

Not because the machine crashed, but because it could not process the architecture without reducing it to the familiar. AI is trained to resolve, to smooth, to deliver. Structural Omission lives in the gap where resolution fails. That gap is not a glitch; it is the work.

The artist doesn’t know’ is not a mistake. It is the most honest place to start.

But it might be the most honest place to start.

Structural Omission was not engineered as an aesthetic position. It emerged from the persistent gap between what I could perceive, what I could represent, and what I could explain. That gap was not a refusal. It was structural.

I began painting at the edge of that failure. The moment where an image begins to feel coherent, then the perceptual ground starts to break down and you cannot recover it. I was not painting uncertainty for effect. I was building certainty until it collapsed under its own weight.

To someone encountering this for the first time, think of a painting that seems complete at first glance, figure, form, atmosphere, and then something catches. A disruption. A missing piece. Not chaos, not mistake, but something structurally off-axis. That is not a flaw in the work, it is the work. The tension between what is knowable and unknowable is what holds the image together.

It could not process the conceptual architecture without reducing it to familiar, resolved categories, and in doing so, it erased the very thing that defines it.

AI struggles here. Not because it is flawed, but because it is faithful. It is a large language model trained on averages. Its strength is its smoothness. Structural Omission exists where that smoothness breaks. Out at the tail, where things do not line up but cannot be dismissed.

For months, I have tested this. Every time I try to write about my work using AI, it drags the language back to the middle range of art discourse. It replaces precision with familiarity. “Ambiguity.” “Withholding.” “Fragmentation.” “Memory.” “Narrative rupture.” These are not the words of Structural Omission. But the machine cannot help itself. It wants to embed in the familiar, known and described.

And it wants to help me.

It keeps assuring me that what I meant to say is something it already understands. It is ChatGPT’s version of a pat on the back: You are exactly where you are supposed to be.

But I am not. That is the point.

The work lives in epistemological discomfort. It reveals the architecture of not-knowing. And that, right now, is almost impossible for machine learning to replicate because the machine is trying to resolve, predict, and deliver. It is trained on closure. I am painting the refusal of it.

This moment marks a live boundary between aesthetics, perception, and machine logic. That boundary may shift as AI evolves. The boundary may shift as AI evolves, but for now the break is clear.

AI cannot model the unknowable. Or structural omission.

It cannot simulate what is unresolved but not broken.

It cannot recognize that the full story does not exist, and that is not a glitch.

To be clear, I am not against AI. This is not a rejection. It is an observation: AI is designed to resolve. I am designed to surface where resolution fails. We do not live in opposition, but we do not live in the same space either.

Structural Omission lives where resolution fails. That terrain is still outside the machine’s reach.