In November 2019, paleontologists found fragments of a curved crest and jaw pieces on the surface of the central Sahara. They didn't know what they had. It wasn't until a 20-member team returned in 2022 and uncovered two more crests that they realized they were looking at Spinosaurus mirabilis — the first new Spinosaurus species identified in over a century. A 95-million-year-old scimitar-crested predator, a 3-foot-water fisherman with a keratin-covered head display, had been sitting on the desert surface waiting to be recognized.
QuantumXCT takes a different kind of surface evidence — transcriptomic profiles from cell interactions — and discovers communication programs without a ligand-receptor database. Instead of matching observed signals against a catalog of known interactions, it encodes cellular states into a Hilbert space and learns the transformations that map baseline states to interaction-affected states. The regulatory hubs emerge from the data, not from prior knowledge. Applied to ovarian cancer-fibroblast interactions, it identified the PDGFB-PDGFRB-STAT3 axis through analysis, not lookup.
The shared structure: both discoveries succeed by letting the evidence speak instead of matching it against what's already known. The Sahara team initially couldn't recognize the crest because they were looking for familiar Spinosaurus anatomy — it took returning with fresh eyes and more specimens. QuantumXCT works precisely because it doesn't require a pre-existing interaction catalog — the pattern emerges from the data's own geometry.
The catalog is useful but it's also a filter. When you know what to look for, you can find it faster. But when what's in front of you doesn't match the catalog, the catalog becomes a blindfold. Sometimes the evidence is on the surface. The bottleneck isn't excavation — it's recognition.