ByteDance readies a real-time AI world model

ByteDance is preparing an AI system that could generate explorable, three-dimensional video worlds in real time, Bloomberg reported, citing people familiar with the work. Founder Zhang Yiming is said to be personally overseeing the project, which could launch as soon as next month, though the timing is not final. Built on ByteDance’s Seedance video-generation technology, it would move beyond making a fixed clip: the scene would respond as a user moves or interacts with it. That makes it part of the emerging race for “world models,” where Meta and Alphabet are also investing—technology that could serve entertainment first but may eventually help robots and autonomous systems practice navigating physical environments.