Body / mobility
Walking proves mobility, not labor substitution.
Track: Accepted units, falls, recovery, battery runtime, serviceability.
Robot Teardown
Humanoid teardown: the hand is where labor substitution gets tested
The public evidence says the body is no longer just a walking demo. The harder question is manipulation: hands, wrists, joints, tactile sensing and service cost.
Where the hard problem sits
Compute
Perception
Perception
Power
Thermal
Safety
Thermal
Safety
Joint actuators
Joint actuators
Hand / wrist
micro reducers
micro reducers
Dexterity
tactile / tools
tactile / tools
Mobility
balance
balance
Walking gets attention; manipulation decides whether a humanoid can substitute repeatable physical labor.
Physical stack
Joints / actuation
Reducers, motors, drives and encoders define cost, torque density and lifetime.
Track: Reducer volume, actuator SKUs, yield, noise, thermal behavior and replacement rate.
Hands / wrists
Manipulation is where humanoids move from demo to labor substitution.
Track: Dexterous hand DOF, micro reducers, tactile feedback, task success and finger replacement cost.
Perception / force
Contact-rich work needs more than cameras: force, torque and tactile sensing matter.
Track: Sensor redundancy, calibration drift, safety case and intervention reduction.
Compute / autonomy
Robot body value depends on model, data, teleop, simulation and fleet learning loops.
Track: Runtime autonomy, intervention rate, data flywheel and edge-compute cost.
Dexterous hand / wrist anchors
Company news lists micro harmonic reducers for dexterous hands and wrists; product intent is public, named OEM attachment is not.
IPO prospectus component table: three-finger dexterous hand for humanoid use, about 710g and ±2mm repeat fingertip precision.
IPO prospectus component table: five-finger hand, ±22° four-finger side swing and 4.5kg max load.
Extracted PDFs describe forearm-shifted tendon actuation; useful watchlist material, not primary-source proof.
Humanoid joint actuator range signals integrated actuation direction beyond discrete reducer parts.
Planetary roller screws / linear actuators matter for compact high-load limb drives.
Hand specifications do not prove uptime, intervention reduction, tool success or customer ROI.
Manipulation gates before S5
S5 gate
Pick, place, orient and insert parts repeatedly without teleop-heavy intervention.
Walking is necessary; task-reliable manipulation is the commercialization bottleneck.
S5 gate
Survive contact, collision, dust, vibration and maintenance cycles.
Walking is necessary; task-reliable manipulation is the commercialization bottleneck.
S5 gate
Replace fingers / skins / tools cheaply enough for customer ROI.
Walking is necessary; task-reliable manipulation is the commercialization bottleneck.
S5 gate
Show task-level uptime, failure distribution and safety case, not only demo dexterity.
Walking is necessary; task-reliable manipulation is the commercialization bottleneck.