Select All Images That Make Your Bones Shrivel is the third zine released under the Institute of Machine Unlearning.

It emerged from a series of public experiments titled CAPTCHURED: Training Sets for Refusal, held in Lisbon and Barcelona. In these workshops, participants collectively examined one of the most ordinary interfaces in contemporary computing: CAPTCHA. The small grid that asks us to prove we are human by identifying traffic lights, stairs, bicycles. A task so routine it barely registers.

The resulting grids are ambiguous, sometimes absurd, often impossible to solve. They resist the clean logic of recognition that standard CAPTCHA systems rely on. Instead of helping machines see more clearly, they complicate the act of seeing itself.

Read together, the images form a speculative dataset of perception. Not a record of what a system should recognize, but a record of what people notice, linger on, and treat as meaningful or strange. Humans and non-humans, animals, pigeons, dogs, trash, textures, small awkward moments. Some images are easily legible. Others resist naming.

What started as a workshop has since expanded into a broader research project. The participant-made CAPTCHAs have been presented as video projections, installations, and are being transformed into an interactive game that renders verification unwinnable. The familiar interface becomes a site of critical play.

This zine gathers the CAPTCHAs produced across these sessions. Each page is both a standalone work and a trace of an attention. Together they form a dataset shaped not by optimization, but by coincidence, humour, bias, hesitation, and refusal.

The Institute of Machine Unlearning uses zines like this one as a method. They hold traces of ephemeral encounters, circulate ideas beyond academic contexts, and turn workshops into objects that can travel, be held, misread, and revisited.

////////

Select All Images That Make Your Bones Shrivel is the result of CAPTCHURED: Training Sets for Refusal, a series of public workshops facilitated by Dominika Čupková in Lisbon, Portugal, and Barcelona, Spain.

This zine was compiled and edited by Dominika Čupková, with support from Ema Diehelová, and published in February 2026 by the Institute of Machine Unlearning under the CC BY-SA 4.0 license.

This publication would not have been possible without the generous contributions of all participants who consented to share their work. Special thanks to Elisava School of Design and Engineering, Mais Uno +1 project space, and WellRead for hosting the sessions.

////////

Released in three different cover versions. The content remains the same, but the exterior shifts, echoing the project’s interest in surface, recognition, and first impressions. Same bones, different skin.