SPATIAL INTELLIGENCE
Ground intelligence in three-dimensional physical reality. Geometry. Physics. Closed-loop action.
Full 3D understanding with temporal dynamics. Geometry, physics, and action unified in a coherent world model.
Geometry
Spatial intelligence requires geometric grounding. Objects have position, orientation, extent. Surfaces connect. Volumes contain.
Physics
Recognizing objects is insufficient. Spatial intelligence must anticipate how objects move, collide, and interact. Physics is not simulation — it is prediction under uncertainty.
Closed-Loop
Static perception is not intelligence. The loop must close: sense, model, predict, act, observe outcome, update. This is where abstract reasoning becomes grounded agency.
Persistence
Objects do not cease to exist when occluded. A spatially intelligent system maintains identity across time, viewpoint changes, and temporary invisibility. The world does not reset.
Track identity across frames
Maintain hidden objects
Invariant to camera pose
Match after long absence
Embodiment
Spatial Intelligence is not purely computational. It becomes real through physical instantiation — robots, vehicles, agents that sense and act in the world.
Physical grounding converts abstract knowledge into actionable policy. The body is not peripheral — it is constitutive of intelligence.
Grounding
Abstract representations must anchor to physical reality. Grounding is the process by which symbols acquire meaning through connection to geometry, physics, and action.
Depth is not decoration — it is constraint.
Object Permanence
The cognitive ability to maintain awareness of objects even when they exit the visual field. Critical for coherent world modeling.