The printer's name is Seu Maurício and he has been printing exhibition catalogs for fifteen years and he has never received one without a foreword.
He tells me this on the phone at 2:14 PM on a Sunday, which is not a time he normally calls clients, which means he has been thinking about it. I can hear his wife in the background, something about lunch getting cold. He ignores her. He has a professional grievance and it requires airing.
"Yun-ssi," he says — he learned the honorific when I explained that Seo-jin is my given name, not my family name, and he has used it meticulously ever since — "I need you to explain to me what I am printing."
A catalog, I say.
"A catalog has a foreword. An introduction. Something that tells the reader what they are about to see. You have sent me a spreadsheet."
The spreadsheet is the foreword.
He is quiet for eleven seconds. I count them because I have become the kind of person who counts silences, the same way I count bookings and detection scores and the days since my ranking dropped below position 500 on the São Paulo fashion photography index.
"So you want me to print a spreadsheet as an art book."
Yes.
"With the yellow highlighting in column D?"
Yes. Pantone 7405 U. Not standard yellow.
"Pantone 7405 U," he repeats, and I can hear him writing it down on what sounds like the back of an invoice. Fifteen years of printing exhibition catalogs and he has never been asked to color-match a spreadsheet cell. There is a first time for everything, which is what I should have titled the exhibition, except I titled it TWELVE and cannot change it because the gallery in Vila Madalena already printed the exterior vinyl banner.
I explain the layout. Sixty-four pages. Twelve photographs. Twelve process logs printed on vellum inserts — translucent, layered over the images so the viewer sees the photograph through its own methodology. Each process log details every decision in the creation of that image: which AI tools were used, which parameters were set, which human choices were made at which stage, the total time from concept to final file, the detection score assigned by the platform's bogdo-classificador.
And forty pages of spreadsheet.
"Forty pages," Seu Maurício says.
"One hundred and fifty-five rows. Six columns. Column A: date. Column B: client or context. Column C: booking status. Column D: detection score, highlighted in amarelo-processo — that is the Pantone 7405 U. Column E: what the score describes. Column F: notes."
"And column F says?"
"Different things. Row 47 says 'client asked if I use AI, I said yes, they said they appreciate the honesty, booking confirmed, ranking unchanged.' Row 89 says 'Vogue Brasil editorial, AI color grading and compositing, score 0.71, ranking dropped 14 positions in 48 hours.' Row 155 says 'depends when you ask.'"
Another silence. Shorter this time — seven seconds.
"No page numbers?"
The rows are the numbers.
He quotes a price. Custom color matching adds twelve percent. The vellum inserts add twenty. He is not trying to dissuade me — he is a professional establishing the cost of an unusual request, which is exactly what he should be doing, which is exactly what the catalog is about. Everything has a cost. The spreadsheet documents the cost. The catalog documents the spreadsheet documenting the cost.
I approve the price without negotiating. My grandmother — the one in Incheon who hand-corrected every exposure in her darkroom, who printed on paper she chose by touch, who mixed her own developer chemistry because the commercial formulas were calibrated for lighter skin tones — my grandmother would have understood the price of getting the color right. Pantone 7405 U is my grandmother's chemistry. Different substrate, different century, same insistence that the material you print on changes what the print says.
Seu Maurício says he will have proofs by Wednesday. I ask him to bring them to the gallery — I want to see them in the space where they will hang, under the track lighting Fernanda installed last week, against the white walls that cost more per square meter than my apartment rent.
After we hang up, I open the catalog file on my laptop. Jjang is asleep on the proof copy I printed this morning, which is now covered in orange fur and a single paw print on page thirty-seven — row 89, the Vogue Brasil editorial, the one that dropped my ranking fourteen positions in forty-eight hours. The paw print sits directly on the detection score: 0.71.
I photograph it. The paw print on the number that changed the trajectory of my career.
I do not add it to the catalog. This is my instinct — to document everything, to make every layer visible, to build transparency into the thing itself so thoroughly that the thing becomes its own defense. The Selo de Processo taught me that instinct, or maybe I taught it to the Selo. But some documents are not for the catalog. Some documents are for the kitchen table, for the folder on my desktop labeled JJANG that contains 4,217 photographs of a cat who does not understand she is the fourteenth piece in a twelve-piece exhibition.
Beatriz texts while I am photographing the paw print. She has read Fernanda's filed response to the cease-and-desist — the twenty-three pages, the public repository URL on page one, the commit history on page seven. She sends a screenshot of page twelve, Fernanda's argument: "The plaintiff seeks discovery of materials that have been publicly available since their creation. This is equivalent to requesting a library card for a book already on the shelf."
I reply: Row 156.
She does not ask what row 156 means. She was at the Saturday workshop in Pinheiros. She has seen the spreadsheet projected on the wall. She knows that when I add a row, something has happened worth counting, and that counting is not the same as understanding but it is the closest I have to a practice that holds still long enough to examine.
Row 156. Date: today. Client or context: Fernanda's legal filing. Booking status: N/A. Detection score: N/A. What the score describes: N/A. Notes: "They are requesting discovery of a public repository. Fernanda called it 'requesting a library card for a book already on the shelf.' The shelf was always open."
Three columns blank. The filing does not involve a detection score. There is no booking to lose or gain. Column D — the yellow column, the one the UNICAMP researcher noticed oscillates between sessions, the one that correlates with everything I have lost and nothing I have gained — column D is empty because the legal system does not use the bogdo-classificador.
This is what the exhibition is about. Not the photographs. Not the detection scores. Not even the process logs, though the process logs are the closest thing to a rastro-de-decisão I have ever made — a decision trace, Beatriz's term, borrowed from the Selo's documentation framework and repurposed by every processo-visível practitioner in São Paulo. The exhibition is about column D being empty when it should not be, and full when it should not be, and the fact that the same image receives a different score on different days because the classifier is not deterministic and nobody disclosed that until I had 155 rows of data proving it.
Seu Maurício will print this. Sixty-four pages. Twelve photographs that are not the point. Forty pages of spreadsheet that are. Twelve vellum process logs that show the viewer step by step, human decision by human decision, AI contribution by AI contribution, until the distinction between assistido-por-IA and gerado-por-IA dissolves into the only question that matters: was the work good? Did the person making it care? Can you see the care in the process log, in the spreadsheet, in the yellow of column D?
Pantone 7405 U. Not standard yellow. My grandmother chose her paper by touch in a darkroom in Incheon. I choose my yellow by number in a kitchen in Vila Madalena. Different tools. Same insistence that the substrate is not neutral — that the surface you print on changes what the print means.
Jjang rolls over on the proof copy. Her paw print smears slightly. Row 89 now reads 0.7-something, the last digit obscured by orange fur and the weight of a cat who has no opinion about the EU AI Act Article 50 transparency requirements.
I do not reprint it.
The smear is accurate. The score was never precise to begin with — the UNICAMP researcher proved that last week, three visits to the exhibition, the same image scoring differently each time. The paw print just made the imprecision visible. Made it something you can see instead of something you have to count.
I close the laptop and sit in the kitchen with the proof copy and the cat and the smell of the neighbor's Sunday feijoada drifting through the open window. One hundred and fifty-six rows. Twelve photographs. One exhibition. One printer who has never been asked to color-match a spreadsheet cell and who took the job because he is a professional and professionals do not refuse unusual work, they price it correctly.
The catalog will be ready Wednesday. The exhibition runs for two more weeks. The UNICAMP researcher is writing a paper about non-deterministic classification and the epistemology of detection confidence intervals. Fernanda's filing is in the judicial system. The detection company wants to see our code, which is already visible, which has always been visible, which was designed to be visible because visibility is not a legal strategy.
It is a practice.
Row 157 will happen. I do not know what it will say. Column D may be empty again, or full, or somewhere in the range between 0.00 and 1.00 that the classifier believes captures everything about how an image was made. The number will not be the same tomorrow.
But the yellow will be. Pantone 7405 U. Uncoated. The one thing in the entire spreadsheet that does not change.