General Intuition, a startup backed by Jeff Bezos, says video game data could help train artificial general intelligence (AGI) systems that better understand how objects and agents move through space and time. The company’s premise is that current large language models—such as ChatGPT and Claude—perform well with text but may be less effective at learning the underlying dynamics of the physical or simulated world. General Intuition argues that game environments and their resulting data provide a structured way to learn those spatial and temporal patterns, which it frames as important for producing intelligence that generalizes beyond language. Across the coverage, the focus remains on the startup’s core bet: using gaming-generated data to address limitations in existing AI training approaches. The articles describe General Intuition’s approach at a high level but do not provide detailed technical specifications or performance results in the excerpts provided.