Crowdsourced mapping has long faced a tension between utility and privacy. The more detailed and useful the footage, the higher the chance it contains identifiable information about people or vehicles. Centralized approaches often upload raw video and blur later, creating both privacy risk and storage liability. @vangrid_io performs the privacy filtering at the edge. Faces and vehicle number plates are obscured on the device during capture and compute. Unblurred frames never leave the phone, are never uploaded to any server, and cannot be recovered later. What exits the device is already privacy-filtered video together with a cryptographic fingerprint of the content. This architectural choice is what makes a global, permissionless sensor swarm viable at scale. Contributors can participate without exposing bystanders. Buyers receive usable spatial data without inheriting the compliance burden of raw identifiable footage. Edge-computed privacy is not an afterthought or a marketing claim. It is the reason the network can expand through ordinary phones while remaining defensible from both a privacy and a data-governance perspective. The design inverts the usual cloud-first model.