Physical staining is the tax microscopy pays for molecular specificity — slow, costly, and, for live cells, phototoxic. This season deep learning flips the bargain: predict the fluorescent label or histology stain straight from a cheap, gentle, label-free image, applying no dye at all. A context-aware model now does it robustly enough for live cells — and, just as important, the field built a watchdog that flags when a generated stain has invented biology that isn’t there.
