01
Real-world variation
Labs control lighting, floors, and clutter. Policies that ace the lab fail when any of them change.
HereLight, layout, surfaces, and clutter change by the hour and by the season. Nothing is reset between runs.
Humanoid field trials
Test where nothing is staged.
Labs and demo floors control the variables that matter. A working farm doesn't. Bring your humanoid to uneven ground, changing light, dust, mud, real weather, and tasks that are never set up the same way twice.
01 What humanoid teams need
The gap between the demo and the deployment.
The bottleneck for humanoids isn't another clean demo. It's evidence from messy, variable, real places. That's what we provide.
02
Failure & recovery data
Demos show successes. Production needs to know what happens after a slip, a drop, or a blocked path.
HerePlenty of honest failures, logged and on video, plus room to practice recovery under supervision.
03
Reliability over time
A one-day demo proves very little. Buyers want weeks of consistent performance on a defined task.
HereMulti-week pilots and full-season residencies on a site that operates every day of the year.
04
Degraded sensing
Dust, glare, low light, moisture, and grime break perception in ways simulation rarely captures.
HereDusty interiors, bright doorways, dark corners, condensation, and sub-zero winter mornings.
05
Hard terrain
Flat concrete hides locomotion problems that gravel, mud, slopes, and thresholds expose fast.
HereGravel yards, ruts, wet and uneven floors, ramps, grates, snow, and ice, all within walking distance.
06
Data you control
Teams need diverse teleop and autonomy data from new scenes, without handing it to someone else.
HereCollection plans agreed up front. Your data, your footage, your approval on anything shared.
02 Task library
Real work, ready to try
A starting menu of tasks we can stage, from teleoperated first attempts to supervised autonomy.
T-01
Gates, latches & doors
Dozens of different mechanisms, many bent, sticky, or improvised.
T-02
Pushing & moving material
Repetitive, heavy, time-sensitive work along long alleys.
T-03
Bedding & cleanup
Carrying and spreading loose material in tight, dirty spaces.
T-04
Hoses & wash-down
Wet floors, water pressure, and slippery surfaces.
T-05
Carrying supplies
Bags, buckets, and tools across uneven ground and thresholds.
T-06
Inspection rounds
Walk a route, read gauges, check equipment, and report back.
T-07
Tool use
Shovels, scrapers, forks, and hand tools designed for people.
T-08
Outdoor terrain
Gravel, mud, slopes, snow, and ice between buildings.
03 How a trial runs
Start narrow. Prove it. Expand.
The humanoid deployments that work start with one task, reach high reliability, then grow. Our trials follow the same path, with engineers on site who can work on your stack, not just watch the robot.
-
Step 1
Scope one task
Pick a single, well-defined task and agree on the success metrics before anything ships.
-
Step 2
Supervised start
Begin teleoperated or tightly supervised, in controlled zones, with written stop conditions.
-
Step 3
Extend the run
Repeat the task across days, weather, and conditions, and log every failure and recovery.
-
Step 4
Hand off the evidence
Data, video, and a field report you can put in front of investors and customers.
04 FAQ
Humanoid questions
Can we start with teleoperation?
Yes. Most humanoid trials should start teleoperated or tightly supervised and progress toward autonomy as the plan allows.
How do you keep people and animals safe?
Every trial has a written plan covering operating zones, supervision, and stop conditions. Our team decides when and where operation near people or livestock is appropriate, and many trials never need to go near animals at all.
Can we collect training data?
Yes. Data collection is often the main goal. We'll agree on scope, privacy, and ownership up front. Your data stays yours.
How long should a trial be?
Long enough to see failures repeat. A few days will surface the obvious problems; reliability claims need multi-week pilots, and seasonal effects need a winter.
Next step
Bring your humanoid somewhere real.
We're scheduling a small number of founding-partner humanoid trials for 2027.