Generalist AI builds embodied foundation models — AI "brains" that let robots understand and act in the physical world across many different hardware bodies (arms, humanoids, mobile platforms). Rather than programming a robot for one fixed task, Generalist trains large-scale models on massive real-world multimodal datasets (video, proprioception, action traces, language) so a single model can generalize to new tasks, environments, and robot types. Its GEN-0 model (Nov 2025) demonstrated scaling laws in robotics; GEN-1 (Apr 2026) claimed 99% reliability on dexterous manipulation tasks — 3× faster than prior state-of-the-art. The company sells or licenses its model stack to robot OEMs and enterprise deployment partners (factories, warehouses, labs). Founded in 2024 by Pete Florence (ex-DeepMind, PaLM-E / RT-2 lead), Chief Scientist Andy Zeng, and CTO Andrew Barry (ex-Boston Dynamics), with a team drawn from OpenAI, DeepMind, and Boston Dynamics.