After a 100-million-yuan seed-plus-angel round, Awomo founder Yu Kaicheng makes his first public case for an implicit, concept-space world model that beats the data-driven scaling law he once helped write. Yu first discussed world models in 2023, when the field was seen as distant. Three years later, with Fei-Fei Li, Yann LeCun and new entrants raising billions, world models have become the physical-AI narrative. Yu, a world champion on ECCV's autonomous-driving track and number one on the Fail2Drive benchmark, rejected the technique that won those titles. His path includes Alibaba's DAMO Academy, where he joined as an Alibaba Star after his PhD on AutoML; his BEVFusion work became a NeurIPS 2022 highlight. He moved to Westlake University in 2023 to set up AutoLab, building generative world models for autonomous driving. Deployment with a top-tier Chinese automaker led him to reason that a model generating realistic data already partially understands the world, making pixel rendering an expensive detour. In late 2024 the team pivoted to an implicit route, building representations of physical concepts in abstract mathematical latent space and decomposing scenes into composable concepts. Internal experiments in 2025 showed notable gains in success rate on complex tasks over imitation-learning and reinforcement-learning baselines, with lower prediction cost. In January 2026 Yu spun the research into Awomo, based in Hangzhou