AI MACHINAEAI MACHINAE

Worn, not
mounted.

Data-collection hardware that looks and feels like ordinary clothing — so people can record real work, in real places, without changing how they do it.

Join the contributor waitlist
A carpenter at work wearing VISUS, TACTUS and GRESSUS

VISUS · TACTUS · GRESSUS — CONCEPT

Plating in a restaurant kitchenPlaning oak in a workshopSorting produce on a farmRestocking a warehouse shelfMaking a hospital bedWiring a junction boxFolding laundry at homeChanging a tyre in a garageAssembling a circuit boardPotting seedlings in a greenhouse

Narrow data builds
narrow intelligence.

General physical intelligence needs data that is diverse — many people, many trades, many environments, recorded on many kinds of hardware. Robots collecting their own data can’t get there fast enough. People already are there, every day.

So the capture kit is built to disappear: a cap, a pair of gloves, a pair of shoes. The platform is equally open to capture hardware from other companies that records to the same schema.

VISUS

CAPTURE 01

VISUS

Capture cap

Four forward-facing cameras built into the brim of a cap that looks like any other. Records what the wearer sees and what their hands are doing, from their own point of view.

Sensors

4 forward-facing cameras

Captures

Egocentric video

Form

Cap — no visible housings

TACTUS

CAPTURE 02

TACTUS

Sensing gloves

Gloves with tactile sensing and joint encoders, worn like ordinary gloves. Records how the hand moves and what it touches.

Sensors

Tactile sensors + joint encoders

Captures

Finger joint angles and contact

Form

Glove — worn like any other

GRESSUS

CAPTURE 03

GRESSUS

Motion-sensing shoes

Shoes with inertial sensors built into the sole. Records how the wearer moves around the work.

Sensors

IMU

Captures

Gait and foot motion

Form

Shoe — sensors hidden in the sole

Build a capture rig? It can feed the platform too.

Diversity of hardware is part of diversity of data. Capture devices from other companies that record to our schema join the same shared dataset.

Partner with us

Your work, teaching
the next machines.

The contributor program will open in stages. Tell us where you are and what kind of work you do, and we’ll be in touch as it reaches you.