Google's Population Dynamics Foundation Model (PDFM) is transforming the field of spatial analytics by learning from vast, multimodal datasets to represent how people interact with places over time. Unlike traditional models that are built for specific tasks using limited data, PDFM is trained on massive, diverse datasets to learn a deep, general-purpose understanding of human behavior in space. This shift enables the creation of rich embeddings that can be reused across many different applications, from disaster response to infrastructure planning. CARTO has explored the application of PDFM by predicting its own flagship spatial feature, the Human Activity Index, and found strong correlations between the model's outputs and expert-engineered indicators, demonstrating its real-world relevance. To further advance these models, CARTO is working on advancing them in areas such as moving beyond administrative boundaries, smart data selection, fine-tuning for local use cases, transparent and explainable AI, and coupling with agentic workflows.