Physical AI needs more than powerful models. It needs reliable data from the real world. @vangrid_io is building a network that turns smartphone captures of physical locations into 3D reconstructions for robotics, AI agents, and world models. The part I find interesting is how Vangrid approaches data collection. Rather than relying entirely on dedicated mapping equipment, the network allows people to capture real-world spaces using their smartphones. Businesses and developers can request specific locations through bounties, creating a direct link between the demand for spatial data and the people who collect it. The network also introduces an agent-driven workflow. AI agents can request real-world captures through x402, connecting digital requests with physical work. Payments are handled in USDC, with Base and Arc supporting the settlement process. This creates a different way to think about spatial data: not just as something collected by large companies, but as a resource that can be contributed and purchased through an open network. As robotics and autonomous systems continue to develop, accurate 3D data of real environments will be increasingly important. @vangrid_io is working on the infrastructure that could help make that data more accessible.