The bench

Numbered instruments. Each one is a live model you can run yourself, with known answers where they exist.

AA-2026-01	Remaining useful life	RMSE 15.85 / limit 16.14	run it ↗ write-up
AA-2026-02	Return risk	ROC-AUC 0.992 / lift 8.7×	run it ↗ · write-up
AA-2026-03	Standards question	50/50 factual / 29/33 router	run it ↗ write-up
AA-2026-04	Unit health	497 units / 2.88M readings	run it ↗ write-up
AA-2026-05	Work order reader	F1 94.4% / 1.1% of params trained	run it ↗ write-up
AA-2026-06	Failure check	recall 91.2% / threshold 0.775	run it ↗ write-up