Barbarika’s mission is to build a global community of robots and researchers that learn about the world together by sharing data, beliefs, and skills. We believe that no single robot can experience enough of the world to build a generalizable model alone — which is why we’re building shared world models powered by community-contributed data. At the core of our approach is Active Inference (AIF) — a brain-inspired framework where agents act to minimize “free energy,” or the gap between their expectations and sensory input. This allows robots to continuously update their beliefs, adapt to uncertainty, and act causally — not just reactively. By combining teleoperated demos, egocentric video, and decentralized learning, Barbarika aims to create scalable, robust intelligence grounded in causality, not just pattern matching.