Volumes I–III of the institute's error index: entries spanning 2029–2032, sorted not by severity but by latency — how many months an error took to become useful. The median is nineteen. The distribution has a long right tail of mistakes that are still waiting.
Status · PublishedThis note is in the ledger. The full transcript is not yet cleared.
Editorial status · Abstract and research roots published. Full body forthcoming.
Research roots · incident review; error taxonomy; learning systems.
Three years of errors, indexed by how long they took to become useful. Reviewer 2's verdict: Accept as archive, not as argument.
← Back to the ledger