World models to replace LLMs as next frontier

World models are the "next transformers" moment, moving AI from sophisticated pattern matching to a fundamental understanding of physical reality. Foundation models like NVIDIA’s Cosmos and World Labs’ Marble are already bridging the gap between digital simulation and physical AI. Unlike LLMs, world models build internal representations of the world, enabling long-term planning and reasoning about physical constraints. The current heavy lean toward robotics (Physical AI) leaves a massive "blue ocean" for applying world models to abstract systems like business logistics and financial markets. Key industry figures like Yann LeCun and Jensen Huang view world models as the necessary step to overcome the "reasoning ceiling" of current language models.