THE REVIEW WE COULD NOT GET
No archivist has reviewed this. So we went and read what archivists published.
A practitioner review is the right way to evaluate a tool like this, and there was no way to obtain one before the deadline. The next most honest thing is to take demands that domain experts have already put on the record, in a peer-reviewed study of AI cataloguing, in the FIAF manual, in the Library of Congress's own account of how it identifies unidentified film, and in the structured analytic technique the intelligence community uses for identification under uncertainty, quote them verbatim, and answer each one.
Four of the fourteen are marked unanswered or only partly answered, including the missing review itself. Nobody quoted here has seen this system, and nobody quoted here endorses it. The register is served from /v1/practice and every structural claim on it is checked by tests.
Summary
The register
Loading the register…
How these sources were found
Using this product's own Parallel Search surface. Two of the primary sources, the Library of Congress press releases and the ArchiveGPT paper, were unreachable to an ordinary scripted fetch, and Parallel retrieved them. The research pipeline this project ships was the pipeline that assembled the evidence base for this page.
Quotations are verbatim. Where a source could not be retrieved and read directly, it is not cited here at all.