Robots learn alone.
Humans never did.
Sixth Sense Labs is building the refinement layer for embodied AI. A network where real people capture real-world experience, and every robot gets to learn from all of it.
Begin scroll
Partners
Act I · The problem
Every robot today
learns in a silo.
That's the bug.
Act II · 01 / Capture
From the human perspective.
Smart glasses record first-person view of real tasks. What the human sees, what their hands do, how they move. No staged demos. No simulations. Just lived experience.
Act II · 02 / Ground
Anchored in real space.
LiDAR-equipped phones capture the room in 3D from the outside, synchronized to the millisecond. Every action gets spatial ground truth, every gesture gets coordinates.
Act II · 03 / Refine
Shared across every robot.
The fused dataset feeds a public refinement layer. One any robotics team can train on, contribute to, and benefit from. Collective experience, finally compounding.
Act II of IVCapture
Act III · The math
Built for the scale of real intelligence.
Multi-perspective capture, sub-millisecond synchronization, and outputs that drop into every major robotics pipeline.
Perspectives fused
2×
First-person and exocentric views fused into a single spatial record.
Synchronization
<1ms
Cross-device sync ensures every gesture lines up with its ground truth.
Output formats
4
HDF5, ROS Bag, RLDS, and Open X-Embodiment. Drop into any pipeline.
Models that benefit
∞
Open refinement layer. Every team trains, every team contributes.
Nothing intelligent has ever evolved alone.
Sixth Sense Labs · Founding Principle
Act IV · For engineers
Commodity hardware.
Pipeline-ready output.
Devices you already own. Formats your team already uses. No exotic dependencies.
Supported devices
- Meta Quest Pro · Quest 3 FIRST-PERSON
- Apple Vision Pro FIRST-PERSON
- iPhone 12 Pro+ LIDAR · EXOCENTRIC
- iPad Pro LIDAR · EXOCENTRIC
Output formats
- HDF5 HIERARCHICAL
- ROS Bag ROBOT-NATIVE
- RLDS TENSORFLOW DS
- Open X-Embodiment INTEROPERABLE