Researchers trained a humanoid robot to play tennis using only 5 hours of motion capture data, achieving a 90% success rate hitting balls traveling over 15 m/s (34 mph) and sustaining multi-shot rallies with human players. The LATENT project demonstrates how minimal training data can enable complex physical tasks in robotics. This represents a significant breakthrough in data-efficient robotics training that could accelerate AI adoption in sports, entertainment, and physical labor applications.