Dexterity & Movement
Atlas' industrial ambitions make walking a whole-job problem
Mobility matters because a robot must reach, carry and recover in spaces designed for people. The decisive test is the task completed at the end of the walk.
The lead / Platforms & Hardware
Atlas reaches the product stage, but the factory evidence still has to follow
Boston Dynamics' January 2026 launch moves beyond the electric prototype. Planned customer deployments, stage demonstrations and proven production work remain separate milestones.
Read the analysis ↗Across the publication
On our radar
Embodied Learning
Gemini Robotics shows why understanding a scene is not the same as controlling a body
DeepMind's embodied-reasoning releases separate interpretation from action. A humanoid still needs a dependable bridge from a useful plan to safe physical execution.
Deployment
GXO's Digit deployment gives the humanoid debate a concrete unit of work
The commercial programme began with a defined material-handling task. Digit 5's broader ambitions should be assessed separately from that operating record.
The subject desks
View all ↗Platforms & Hardware
Bodies, actuators, hands and the engineering of useful machines.
Platforms & Hardware
Atlas reaches the product stage, but the factory evidence still has to follow
Boston Dynamics' January 2026 launch moves beyond the electric prototype. Planned customer deployments, stage demonstrations and proven production work remain separate milestones.
Platforms & Hardware
Figure 03's redesigned hands make sensing part of the dexterity story
Figure links tactile sensing and palm cameras with Helix control. The test is whether the combined system handles varied objects and recovers from imperfect contact.
Platforms & Hardware
Digit 5 and Atlas take different routes to a longer working day
Agility emphasizes faster charging, while Boston Dynamics describes autonomous battery exchange. Neither approach alone determines productive uptime.
The subject desks
View all ↗Dexterity & Movement
Manipulation, mobility and recovery in the physical world.
Dexterity & Movement
Atlas' industrial ambitions make walking a whole-job problem
Mobility matters because a robot must reach, carry and recover in spaces designed for people. The decisive test is the task completed at the end of the walk.
Dexterity & Movement
Helix 2.5 makes recovery part of the humanoid dexterity debate
Figure's September 2026 evaluation reports whole-body work in unfamiliar homes. Its most revealing question is how the robot responds when the first attempt is imperfect.
Dexterity & Movement
Digit 5 puts controlled stopping at the center of the humanoid product
Agility's new architecture includes people detection, independent motion oversight and a seated response. The safety case still belongs to the complete application.
The subject desks
View all ↗Embodied Learning
How robot models acquire, transfer and apply physical skills.
Embodied Learning
Gemini Robotics shows why understanding a scene is not the same as controlling a body
DeepMind's embodied-reasoning releases separate interpretation from action. A humanoid still needs a dependable bridge from a useful plan to safe physical execution.
Embodied Learning
Figure's human-data strategy tests how far experience can transfer between bodies
Helix 2.5 uses human-experience pretraining, while shared robot datasets pursue another route. Both approaches still have to bridge differences in sensors, reach and action.
Embodied Learning
1X's Expert Mode makes the human service behind a home robot explicit
NEO's product model includes scheduled remote guidance for unfamiliar chores. That changes how autonomy, privacy and the cost of completed work should be described.
The subject desks
View all ↗Deployment
What industrial trials and operating programmes actually establish.
Deployment
GXO's Digit deployment gives the humanoid debate a concrete unit of work
The commercial programme began with a defined material-handling task. Digit 5's broader ambitions should be assessed separately from that operating record.
Deployment
Figure 03 at BMW moves the test from body-shop handling to logistics
BMW's June 2026 project and Figure's account describe a new task context. Earlier Figure 02 experience is useful history, but it does not validate the new workflow automatically.
Deployment
NEO's home-service model reveals why factory success does not settle domestic autonomy
A household combines unfamiliar objects, privacy and changing instructions. 1X's use of basic autonomy and scheduled expert help makes those differences visible.
The subject desks
View all ↗Economics & Safety
The full cost of work, evidence and accountable operation.
Economics & Safety
Digit's commercial model puts completed work ahead of the robot's sticker price
GXO's deployment and the Digit 5 roadmap show why unit cost alone cannot describe humanoid economics. Support, utilization and task reliability decide the result.
Economics & Safety
Agility and FORT bring the maintenance shift into the humanoid safety story
The October 2026 partnership describes manual overrides and external safety interfaces. Those details matter when a robot is being set up, repaired or recovered.
Economics & Safety
Helix, Digit and Atlas show why humanoid claims need different scorecards
A home evaluation, a commercial handling task and a product launch each provide evidence. Reading them as the same kind of milestone produces misleading conclusions.