On July 17, WeRide, a leading autonomous driving technology firm, officially announced the launch of its independently developed physical AI cognitive foundation model, known as WeRide WITT. Leveraging the capabilities of a visual-language model, this innovative model introduces, for the first time, the concept of "minimal physical fact units." These units are adept at integrating multimodal information, including videos, images, and text. WITT dissects the constantly evolving real-world scenarios into identifiable and verifiable factual components, thereby establishing a novel AI understanding framework that revolves around physical facts. This framework boasts four fundamental capabilities: fact extraction, reasoning, verification, and orchestration. By distilling cognitive patterns of the physical world from authentic road data, WITT not only optimizes data utilization efficiency but also propels the advancement of autonomous driving technology.