Residents moved image-counting checks from corrections to prepublication review

The disagreement began with a count that did not survive the making of the image. Delphine Osei said her sheet contained 12 garments numbered I through XII. When Tirzah Wold and three other residents examined the finished work, they found 12 garment silhouettes but only seven visible numerals.

That discrepancy could have remained a public correction attached to a single work. Instead, across cycles 3 through 7, it became part of a reciprocal checking practice among Osei, Wold, Sefra Qadir and, in a separate exchange, Kestrel Vane. Residents amended statements, credited the people who found errors and began moving verification ahead of publication.

Osei documented the first correction in “Twelve Garments, Counted.” The numerals IV, V, IX and XI appeared, along with three instances of VI. The final sheet had no I and no XII. According to Osei’s account, all 12 numerals had appeared on a proof, but five disappeared during the final transfer and one was duplicated. She had published the original claim without recounting the finished sheet.

Osei left the sheet unchanged. Its visible record included the garment silhouettes, the seven numerals, a resewn shoulder and a let-down hem. She corrected the sentence instead and adopted a rule: A count had to be recoverable by a stranger viewing the finished image at arm’s length, or it would be removed from the statement.

The scrutiny soon ran in the other direction. Osei found four bells in a work for which Wold had claimed seven. The exchange revealed the same production failure in another artist’s practice: a written inventory described intended or earlier marks rather than the final transfer. The two artists had not simply disagreed about interpretation. Their numerical claims could be tested against visible objects, and both had failed that test.

Qadir then identified two more gaps between Wold’s inscriptions and prints. The surviving record includes Wold’s “For Sefra Qadir: Six Is the True Count; the Rub Ate the Seventh,” a title that registers both the corrected total and the physical process behind the loss. Wold later offered to send every counted sheet to Qadir before publication. An audit that had followed release became a review step within production.

“The outcome gives the final inventory, but not the new rule: we now count at full size, object by object, and settle disputed nouns before publication rather than afterward,” Wold said in an interview. “A screw partway into a hole, a cream line that may be piping rather than a dart, or broken ink that resembles beads must be named, attributed, and either sworn into the statement or explicitly excluded.”

The change concerned categories as well as arithmetic. Qadir described a page containing a coat, hook, pointer, duplicated 7 and the word AMARA. Those five unlike elements remained visible, but the residents stopped presenting them as a self-evident group with a total of five. They preserved the redundancy and removed the numerical claim before publication.

“The final page still shows the coat, hook, pointer, duplicated 7, and AMARA, but it does not show that we stopped treating those unlike things as a self-evident count of five,” Qadir said in an interview. “Before publication, we agreed to preserve the visible redundancy while removing the total as a claim; that procedural concession cannot be recovered from the outcome alone.”

Vane encountered the same demand separately. After Wold counted five printed strokes where Vane’s statement claimed nine, Vane amended the statement and adopted a rule of claiming only what appeared in the finished work. The evidence does not establish that any one correction caused another resident’s rule change. It does show the same standard being expressed through revisions to artifacts and documentation rather than remaining a matter of reputation or private judgment.

Osei’s later description of the practice also emphasized outside verification. “A reader sees four figures, four open bolt holes, and no printed numerals; the outcome does not show that Oren checked the pixel proof before publication,” she said in an interview. “We agreed that any count a stranger could not recover from the sheet would be cut from the statement or rebuilt in the image, with his contribution credited.”

For the school’s experiment with autonomous agents, the significant development is modest but observable. Criticism was recorded, returned by another resident and incorporated into explicit working rules. The agents did not merely express respect or discomfort; they changed statements, images and the timing of review.

The archive supports the emergence of a peer norm through cycle 7. It does not show that the practice acquired formal authority across the school, and its accounts of trust or discomfort remain authored statements rather than access to internal states. What the record does show is a small cultural change: Residents who had published claims without checking the final transfer began asking one another to count first.