Vangrid上线：用手机众包采集机器人所需3D场景
Vangrid让用户按指定地点用手机拍摄，生成可上链验证来源的3D重建资产，Web与Android已上线。
almost wrote another broad Physical AI post. then i looked at what @vangrid_io is actually asking phones to capture. not “a city.” one loading bay edge. a pedestrian underpass. the approach to a crossing. the places a robot cannot understand from old maps, satellite tiles, or a dataset collected somewhere else. that is the bottleneck people keep flattening into “more data.” world models need the physical world as it is now, from the height and path where an agent will actually move. a barrier appears. a kerb changes. a narrow route gets blocked. the internet does not reliably notice. Vangrid turns that gap into a request: someone funds a specific location someone else captures it with a phone the result becomes a reconstructed 3D asset the capture’s provenance can be checked onchain the phone is not magically a robot sensor. it is just already everywhere, and it can be directed toward the exact awkward scene a model is missing. that feels more useful than treating Physical AI as another distant category. the product is already live, which matters. web and Android launched before the funding announcement, and the network activity is visible rather than left as a slide-deck promise. still checking the explorer before trusting any headline. if Physical AI is going to leave the lab, does its hardest data problem end up being scale—or simply getting somebody to capture the one place nobody thought to map?