https://nyaa.nizbot.com/ One item per claim, with claim, measurement, method, falsifier and status. en nyaa-001 https://nyaa.nizbot.com/notebook.html#nyaa-001 Fri, 09 Oct 2026 00:00:00 -0500 Most of the wait before our AI's first word on a repeat turn was the engine re-reading text it had already read. On a turn whose context hadn't changed, most of the time to the first token went to re-reading context the engine had already read. Our engine's median time to first token was 1.27 s (397 requests, all traffic). Most of it traced to cache-block rounding, an estimated 0.8-1.5 s per turn. A config change then cut a test turn from 0.88 s to 0.16 s. live engine metrics on our own machine, 2026-10-09; the rounding share is an estimate from the uncached-token distribution, not a direct measurement. People turns were not measured on this engine. If an exact end-of-request cache checkpoint does not remove most of that re-read time on byte-stable turns, the claim is wrong. running nyaa-002 https://nyaa.nizbot.com/notebook.html#nyaa-002 Fri, 06 Nov 2026 00:00:00 -0600 Pre-registered. The result that would kill it was written down before it ran. sealed nyaa-003 https://nyaa.nizbot.com/notebook.html#nyaa-003 Fri, 06 Nov 2026 00:00:00 -0600 Not yet. sealed