LLMs ate the internet. Robots cannot.
A world model doesn’t need another blog post. It needs the factory floor as it stands this morning. The loading bay that changed last week. The stairwell nobody photographed. The aisle that looks different from any satellite that can only see the roof.
That’s the bottleneck Jensen Huang flagged as Physical AI’s ChatGPT moment. Not compute. Ground-level 3D from real places, collected fresh, from the exact location that matters.
@vangrid_io turns the phone already in someone’s pocket into an on-demand capture node for that layer. A requester funds a bounty for one specific site. A contributor films a slow orbit. Faces and plates blurred on-device. Content hash, location, and timestamp fingerprinted. Batch root anchored on Base through EAS.
The output is a textured 3D mesh from ordinary video matching points found across frames, space reconstructed, colors and textures applied, exported as a GLB. No LiDAR rig. No specialist hardware.
$1,000 paint booth capture. $700 pier deck. $600 market square corner. $100 Belisha beacon. ~$325K settled total. 14 jobs open right now.
Amsterdam-based team. Robert Brighton from Microsoft HoloLens and Quantum Computing. Khalid Benrafik from
Booking.com, ING, and Mollie. $9M from HashKey, Animoca, Borderless,
Crypto.com Capital, Gate Labs, and Mapleblock. NVIDIA Inception.
You cannot scrape a loading bay. Someone has to walk it.