The frontier of AI research moves fast, but the underlying questions are stable: how do we make models reason reliably, operate across modalities, integrate symbolic and neural approaches, and align with human values? These lessons cover the research directions that will shape the next generation of AI systems.