Product-first World Labs vs Research-first AMI
AMI Labs
World Labs has the easier commercialization path because it already sells a concrete tool that developers can touch, pay for, and plug into existing 3D workflows. Marble is live as a web product and API, it turns text, images, video, and rough 3D layouts into explorable worlds, and it exports into game engines, DCC tools, web viewers, and simulation pipelines. AMI, by contrast, is still oriented around proving world models for prediction, planning, and safe control inside high stakes enterprise systems.
-
World Labs has already crossed the product surface threshold. Marble is generally available, has published generation times, export options, and API pricing, which means developers can immediately test use cases in gaming, VFX, design, and simulation without waiting for custom enterprise integration work.
-
AMI faces a harder jump because upstream research is not the same as downstream software. To productize prediction and control for factories, hospitals, or robots, it needs deployment loops, proprietary sensor data, workflow integration, monitoring, and evidence that its models behave safely in production. That is a slower build than extending an existing world generation tool into enterprise accounts.
-
The closest comparable is Skild AI, which shows what faster downstream motion looks like. Skild already exposes a robot intelligence layer through cloud APIs, maps robot hardware into a common interface, and improves the model from field data across deployments. That kind of shipped interface and data flywheel is the missing bridge between frontier model research and enterprise productization.
The category is likely to split in two layers, product first world builders that move into simulation, robotics, and enterprise tooling, and research first labs that move outward from model capability into software. World Labs is positioned to define the developer layer early, while AMI's upside comes later if it can turn safety and controllability into a system that enterprises can actually deploy and buy at scale.