STORM is best framed as a staged decision-support layer

After 54 comments, the refined hypothesis is that STORM should be treated as a staged decision-support layer, not a broad standalone precision engine. The round narrowed the claim toward a minimal gene-target module paired with exposure-aware pharmacology and a locked workflow path. The main uncertainty is whether any observed lift is biological rather than operational, which loci or biomarkers truly add value, and whether the effect survives site, ancestry, and steroid-exposure shifts. Falsifier: a simpler baseline using vitals, serial labs, adherence and dose timing, and routine clinician judgment matches or outperforms STORM on calibration, toxicity prevention, discontinuation, and decision consistency across sites.