World Models Need a World

More than $2 billion has flowed into world models in just four months, yet every one of those models lives in a datacenter. None can stand in your room, anchored to your space, in real time, on the glasses this industry spent fifteen years building. Who builds the layer between intelligence and presence? And why is XR quietly holding the strongest cards?

In this session, Luis Oscar Ramirez, founder and CEO of Mawari, draws on nine years of building real-time 3D streaming infrastructure from Tokyo, including winning a telecom giant's 2019 challenge to stream an AI-powered digital human to smartglasses before "embodied AI" had a name, and filing patents that same year on analyzing how people interact with live 3D content. He maps the world-model landscape from the XR side: what today's models can and cannot do, why delivery to real spaces remains unclaimed, and why the spatial interaction data XR generates is the one asset every AI lab needs but cannot collect.

Attendees will leave with a clear framework for where XR fits in the AI era, the case for why the next 36 months decide who owns this layer, and an invitation to help build the community-powered infrastructure that makes world models real.