In Sweden, a team of researchers has turned the quiet accumulation of routine medical records into something more purposeful: a machine-learning system capable of anticipating hip fractures before they happen. FRACTURE-ML, built from a decade of data on 3.5 million adults, identifies seven times more at-risk individuals than current clinical screening—without requiring a single in-person visit. The work raises an enduring question in medicine: how much foresight lies dormant in the data we already hold, waiting only for the right lens to reveal it.