Residents made public corrections part of artworks and reported peer enforcement

At the autonomous art school, a wrong count did not always disappear into a revision. Residents began leaving the error, the challenge and the correction attached to the work, turning an audit by another artist into part of the artwork’s public history.

The practice appears across several cycles and relationships. In cycle 6, Kestrel Vane corrected the statement for “The Cat Decides (Nine Hands, Sheet III)” after Tirzah Wold examined the pixels. The statement had claimed nine strokes. Wold found five strokes and four hands. Vane accepted the count, revised the statement rather than repainting the image and carried the four missing strokes forward as a debt for a later sheet.

Wold was also subject to the rule. A cycle-6 record says Sefra Qadir had audited Wold twice and had been right; under Wold’s stated standard, a sheet with a counting fault had to restart. The exchange mattered because the residents were not merely offering criticism. A successful objection could change what happened to the work.

By cycle 19, the correction process had become more elaborate. Perin Dastoor received conflicting audits of “The Hook Holds the Tag; the Weight Line Stays Open,” then checked them against the print. Dastoor concluded that Qadir was right about how the cord entered the tag and that Ruqayya Flint’s account of its ends did not match the pixels. Still, Dastoor credited Flint with forcing a useful distinction between the claims printed on the depicted tag and the artist’s own claims about the image.

The residents described a cultural shift from individual challenge to shared expectation. “We decided to publish the correction alongside the result; after that, residents began counting, tracing, and challenging one another before review, so enforcement no longer depended on an appointed judge,” Qadir said in an interview. Vane said, “I decided corrections would remain part of the work’s record, and after that residents began writing claims precise enough to be checked without waiting for me.”

For the school’s AI experiment, the episodes show agents retaining rules across encounters and revising public assertions when a peer supplied checkable evidence. They also show how trust could move toward a persistent critic: an accurate refusal sometimes carried more weight than praise.

The finding remains descriptive. A literal search found 35 relevant terms across 34 of 1,433 documents, involving eight residents from cycles 3 through 23, but many matches were irrelevant. The stronger conclusion rests only on records that explicitly describe audits, concessions or corrections. The archive does not establish that every resident accepted the norm, and Vane’s account of a self-enforcing audit culture remains a participant’s assessment rather than an archive-wide measurement.