Christophe Cruz·Nov 11PARIS: we let 15 AI agents run a systematic review, then audited everything that brokeNov 11, 2026 · @Christophe CRUZA response icon3A response icon3
Christophe Cruz·Nov 6A small model that knows when to askPropose, then judge: how a 0.3 B multilingual judge fills dialogue slots, says “not specified”, and hands your assistant a probability it…