Robots are getting better at doing tasks. But they still need to understand the places where those tasks happen. Think about a warehouse robot. It can learn how to pick up a box. But it also needs to know where the shelves are, where the loading area is, and what has changed since the last map. That is where @vangrid_io comes in. A lot of robotics data is collected in controlled environments. Companies use cameras, rigs and human operators to teach robots specific tasks. Useful, but limited. It doesn't tell a robot what every warehouse, street or building looks like today. Vangrid takes a different approach. People use smartphones to capture real places. Those captures can then be turned into 3D data, including point clouds and Gaussian splats. The basic flow is simple: 𝗣𝗵𝗼𝗻𝗲 → 𝗰𝗮𝗽𝘁𝘂𝗿𝗲 → 𝟯𝗗 𝗿𝗲𝗰𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻 → 𝘃𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻. The verification part is important too. Vangrid uses Base, EAS and Merkle trees to create a record around the capture and its history. So there are really two different data problems: 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗱𝗮𝘁𝗮: “How do I do this task?” 𝗦𝗽𝗮𝘁𝗶𝗮𝗹 𝗱𝗮𝘁𝗮: “What does the place around me look like?” Robots need both. Vangrid is focused on the second one.