Researchers used continuous glucose monitoring and multimodal data from over 3,000 people to reveal hidden patterns in glucose spikes across normoglycemia, prediabetes, and type 2 diabetes. Their AI-driven model exposed high individual variability and offers a personalized path for early detection and prevention of diabetes.