Writing
One publication, three collections.
Essays carry the thesis, lab notes track the projects, and reports document validated results. Spinout origin stories live here too. The first pieces are being written now, and each entry below goes live when it's ready.
ForthcomingFlynn on NASA IMS: the validation dossierZero false positives over 120 hours, about 17 days of advance warning, and how the numbers were produced.Report
ForthcomingLCVDB: from 92% to 100% recall at 6.5M vectorsDiagnosing the LSH candidate-recall ceiling, and how sharding broke it.Report
ForthcomingGuanaco: lossless LUT compressionA Ternary-3 lookup table on LFM2.5-350M: 3.01× smaller, faster, and lossless.Report
ForthcomingThe null baseline is mandatoryA celebrated 98.68% result, matched by a zero-parameter identity function, and the rule the lab adopted because of it.Essay
ForthcomingDeterminism starts in the siliconWhy guarantees have to be designed into the hardware from the start.Essay
ForthcomingThe case for research that emits companiesThe spinout thesis: the parent sells a methodology, and each spinout sells a product.Essay
ForthcomingSSTT: classifying images with zero parametersTernary signatures, an inverted index, and topological ranking.Lab note · SSTT
Forthcomingreflex-os and Flynn: where the certification boundary sitsHow the adaptive layer and the deterministic layer each run on their own side of the boundary.Lab note · reflex-os
Get notified
New essays and lab notes, when they ship.
Email us and we'll tell you when something is published.