Om AI's VLX series models, the world's first edge-native streaming multimodal models for physical AI, evolved from a multimodal model that in 2023 unexpectedly understood surveillance footage better than a small model trained on surveillance data for years, a breakthrough that strengthened CEO and Chief Scientist Dr. Zhao Tiancheng's conviction that multimodal training enables generalization in the physical open world. By the summer of 2026, VLX introduced the novel architecture of "edge-native streaming multimodal," achieving for the first time on the edge a complete physical AI闭环: continuous perception, precise localization, and action decision-making. While the AI industry remains fragmented and undecided on dominant conceptual pathways like VLA and world models, Om AI Lianhui's VLX series models have completed the commercial闭环 from simulation experiments to industrial application, endowing physical terminals such as robots, drones, wearables, security cameras, and AI PCs with autonomous perception and cognitive decision-making. Dr. Zhao, who graduated with his Ph.D. in 2019 from Carnegie Mellon University’s Language Technologies Institute, chose entrepreneurship over academia or a major tech company.