Most robots deployed in the field today follow pre-programmed trajectories. For years, the dominant path to the next generation of robotic intelligence has rested on collecting more and more robot data, an approach that is inherently limited in both scalability and generalizability - this has resulted in such robots being stuck in the lab and not deployable to real environments.

Rhoda is charting a different course. Our pioneering world action model is deployable in real manufacturing and logistics environments today, while pointing toward something far larger: true general-purpose physical intelligence. In this session, we'll share why world action models break the data bottleneck that has held robotics back, how our approach delivers real-world performance now, and what it will take to reach robots that generalize across tasks, environments, and embodiments.