BUILT-IN SKILL
PRIMITIVE
Composable skill units
designed for modular
scaling across tasks.
WORLDACTIONMODEL
Learning physical intelligence throughobservation and interaction — enablingrobots to understand and act in the realworld, and generalize across tasks,environments, and embodiments.
INTERACTION-DRIVEN LEARNING
FROM PHYSICAL INTERACTION TO WORLDUNDERSTANDING
Learning physical intelligence through real-worldinteraction — enabling robots to manipulate,adapt, and generalize across tasks, environments,and embodiments.
OPEN
MODELWORLDSKILL
BUILT-IN SKILL
ADAPTIVE
TACTILE
INTERACTION
Millisecond tactile feedback
stabilizes manipulation
behavior in real time.
INTENT-DRIVEN
MANIPULATION
Robots to perceive, plan, and act
from intent — without scripts or
task-specific programming.
OPEN SKILL
ECOSYSTEM
A shared skill library where
manipulation intelligence
accumulates, is reused, and
continuously evolves across
deployments.