Fully AutonomousRobo-CentricData Foundry
World's first autonomous data foundry running robots in real home environments. Generating contact-rich, household-specific training data.
Household
is the final frontier.
We believe the next revolution in robotics won't happen in warehouses or factories. It will happen at home.
What made GPT a cultural moment wasn't the technology — it was mass market accessibility. Millions of people using it, every day, in their own lives. Robots will follow the same path.
We don't know which technology will get us there. World Models? VLAs? Something we haven't invented yet? What we know is this: whatever wins, it will need data. Real-world, contact-rich, home environment data. At scale.
To build this foundry, we had to solve a hard problem first: how do you make a robot autonomous on a task without months of engineering? We built the One-Shot method—from hours of demonstration to minutes. From months of engineering to days.
Physical AI is stuck in the lab.
The “perfect run” flaw: robots are trained only on flawless trajectories, so they collapse on minor real-world variation. Tier 1 Robo-Centric data — with real failure and recovery — is the only path to models that survive outside the lab.
Robo-centric
High-fidelity, cross-embodiment data generated directly by autonomous robots. The layer no one was able to scale. Until now.
Interface-centric
High-fidelity teleoperation and haptic rigs. Fundamentally unscalable — capped at a 1:1 human-to-robot operational hour ratio.
Human-centric & Ego-centric
Cheap and abundant — YouTube, wearables, ego-video. Lacks precision, modality, and contact fidelity. Good for VLMs, not manipulation.
Cross-embodiment data generated directly by robots
Human operator uses expensive hardware to control a physical robot
Human-centric ITW data and human-operated interaction capture
Demonstrate once. Run forever.
One-Shot execution turns data generation from a manual engineering bottleneck into an autonomous factory — synchronized, contact-rich Robo-Centric data, around the clock.
- 01
Demonstrate once
One human demonstration. Multimodal sensing captures the full contact sequence.
- 02
Extract & adapt
Task logic extracted and adapted to the local environment — no hand-coded waypoints.
- 03
Run forever
Skills deploy across fleets 24/7. Recovery loops capture real failure and recovery data.
- Industry: 1+ hr / skillLili-o: <5 min
From months of engineering to days.
One-Shot Method: teaching robots new skills under a minute
Our proprietary One-Shot method reduces manipulation primitive creation from over an hour to under five minutes — the unlock that makes the foundry viable at scale.
Read article →We're looking for the teams building household robots.
We're opening the foundry to a small number of design partners — robotics companies and research labs who want early access to household manipulation data and are willing to help shape what we build.