At the recent All In conference, the world model AI sector remained largely opaque despite growing interest and funding. Leading companies like AMI Labs, founded by Yann LeCun, and Fei-Fei Li's World Labs have generated significant buzz for developing AI that automates spatial intelligence. This technology holds promise for industries such as robotics, interactive video, and advanced autonomous systems, yet both founders and their teams are tight-lipped about their specific product plans or timelines, with AMI Labs emphasizing that they are still in a research and development phase.

World Labs’ Marble platform is among the more concrete products publicly demonstrated, showcasing applications in media creation, virtual environments for gaming, and CGI effects, with some exploratory robotics use cases. However, even data providers like Physicl remain uninformed about how their contributions are being utilized, highlighting a secretive culture that extends beyond companies to their suppliers. Physicl’s CEO, Alex de Vigan, noted the difficulties of optimizing data support without clear insights into client projects.

The versatile nature of world models contributes to this guarded approach. These models can be applied in vastly different domains—from mapping autonomous vehicle routes to enabling humanoid robots or generating immersive media landscapes. AMI Labs is exploring various sectors including manufacturing, biomedicine, robotics, and AI tools for healthcare, but strategic focus remains unclear. The industry’s ample venture capital allows multiple labs to remain under the radar, delaying product revelations to avoid alerting competitors prematurely.

This tendency reflects a "dark forest" dynamic, where companies stay silent to protect their innovations until they are ready to compete openly. While fundraising flows freely, it also funds rivals who would jump into any market opportunity once it becomes apparent. For now, the secrecy maintains a cautious equilibrium, but it also signals an intense race looming once the practical applications of world models begin to crystallize in the coming years.