Spatial intelligence · Human behavioral data · AI culture
Real Spaces, Real People, Real Behaviour.
What is VLGE
vlge /ˈvɪlɪdʒ/ “village”
Through a growing universe of 3D and 3DGS (Gaussian Splatting) worlds, VLGE captures how real people create, explore, interact, and make decisions, establishing a human-grounded behavioral foundation for AI.
That foundation becomes the anchor for scale. LLM-based agents are grounded in observed real-world human behavior, enabling them to generate human-believable interactions, decisions, and trajectories at massive scale without losing connection to the people they are designed to model.
As the network grows, every human interaction strengthens the behavioral ground truth, every grounded agent expands its scale, and every world contributes new environments, behaviors, and scenarios. The result is a compounding data engine that produces both high-quality training data and benchmark-ready datasets with quantitative metrics and qualitative behavioral streams, giving research labs a common foundation for training, evaluating, and comparing embodied and spatial AI systems.
Our mission is to ensure people remain the architects of their own minds by making real human behavior the anchor of AI training, simulation, and evaluation.
Explorable 3D environments of any space, a museum, a street, a store, a tower, a game, that real people walk, build, and play inside, online.
Every session becomes spatial & behavioral video, audio, text and meta data of where people go, what they touch, how they dwell. Generating the proprietary intelligence that teaches machines how the physical world works.
Scale human-believable data using LLM-based agents capable of creating a world and playing within it by perceiving, reasoning, taking actions, and making changes to the world.
Grounding the agents to real-human ground-truth data in the real-world to bridge the sim-to-real gap for embodied AI. The human-grounded data is benchmarked so AI labs can evaluate and improve their agentic systems for free.
captured with consent · GDPR-compliant · and counting
As featured in