The lab at Tallinn University of Technology is small—four rooms, nine researchers, and a budget that the team lead, Dr. Kaia Tamm, describes as "optimistic." On the wall is a printout of the Concern's hardware specification, every page annotated in three colors of ink. Page 91 has been photocopied and pinned separately, circled in red. Page 91 is the problem.

Page 91 defines the minimum sensor sampling rate for inter-finger spatial relationships: the speed at which the gloves measure the distance between fingers. The spec requires 800 Hz. For most manual tasks—cooking, surgery, craftsmanship, piano tuning—800 Hz is more than sufficient. Human fingers rarely move faster than 400 Hz in any productive activity.

Sign language operates differently.

"In Estonian Sign Language, there are twelve phonemes that require finger transitions faster than 600 Hz," Dr. Tamm explains. "In American Sign Language, there are eight. In Japanese Sign Language, there are twenty-three. At 800 Hz, the gloves capture these transitions—but barely. The Nyquist margin is too thin. You get aliasing. The captured signs look correct to a human reviewer but contain subtle timing errors that an AI training on the data would learn as truth."

In other words: the Concern's gloves can capture sign language. But they capture it wrong. Not obviously wrong. Insidiously wrong—wrong in a way that would teach machines to sign with an accent nobody can hear but every deaf person can feel.

Dr. Tamm's team discovered this eight weeks into their certification attempt. They had built twelve pairs of modified Gen 2 gloves with the inter-finger sampling rate boosted to 2,400 Hz—three times the spec minimum. The sign language data captured at 2,400 Hz was, according to their analysis, "categorically different" from the same signs captured at 800 Hz.

"We showed the 800 Hz captures to three deaf linguists," Dr. Tamm said. "All three said something felt wrong but couldn't identify what. We showed them the 2,400 Hz captures. All three said, 'That's correct.' The difference is invisible to hearing people. It is not invisible to the community that actually uses the language."

The finding sent a tremor through the Concern's spec committee. If the glove spec was insufficient for sign language, what other domains had similar blind spots? Were there crafts, skills, or forms of expression that the current spec captured incorrectly without anyone noticing?

☜cert_authority convened an emergency review. The conclusion, published five days later, was unusually candid:

"The specification was written by people who hear. It was tested by people who hear. It was optimized for tasks performed by people who hear. This is not malice. It is the ordinary failure of a perspective that doesn't know what it doesn't know. We are grateful to the Tallinn lab for showing us what we couldn't see."

The spec committee proposed an amendment: domain-specific sampling rate profiles. Instead of a single minimum, the spec would define baseline rates for broad categories—manual labor, fine craft, medical, musical—and allow certified producers to implement higher rates for specific use cases. Sign language would require 2,400 Hz minimum. Other domains would be added as blind spots were identified.

The amendment passed unanimously.

Dr. Tamm's lab is now working on the first certified sign language gloves. Their ☜handle—☜silent_hands—has become a rallying point for deaf contributors who had been hesitant to join the Registry.

"We're not asking the Concern to change for us," Dr. Tamm said. "We're asking the spec to include us. There's a difference. The spec is a living document. Living things grow."

The Tallinn lab's certification is expected next quarter. Pre-orders from deaf schools in fourteen countries have already exceeded their first-year production capacity.

Page 91 has been rewritten. It is now three pages long.

☜