A↗adaptive / table
ConnectingLOCAL WORKBENCH

PHYSICAL AI · TWO ARMS, ONE TASK

A little more adaptive.

Give the goal. Watch the scene. Follow every decision.

Development baselineLearned policy not loaded
Robot workspaceSeed 0
Starting
Live camera image from the dual-arm MuJoCo simulation
Preparing the simulation…
LIVE RGB OBSERVATION0.0 s simulated
PlateCupSpoonForkActual MuJoCo rendering
Your instruction01 / GOAL
Current goalNo goal yet

Set an instruction to build the task list.

Try the learned policy

Reaching only; does not pick up objects.

Pilot has not been run.

PLACEMENTS VERIFIED0 / 0
CAMERA OBSERVATIONS0
GOAL VERSION—
RECOVERY ATTEMPTS0
Five abilities. One shared system.02 / CAPABILITIES
01

Second Chance

Look again. Correct a missed move.

Awaiting action
02

Dinner Detective

Reveal an item before picking it up.

Awaiting action
03

Photo-to-Table

A picture becomes the placement goal.

Awaiting image
04

One More Guest

Update the task while keeping good work.

Awaiting update
05

Think Twice

Compare two moves before acting.

Awaiting preview

Activity labels describe this run. Full feature acceptance and learned manipulation require recorded evaluation.

Task sequence03 / PLAN

Your goal will appear here.

What the robot observedCAMERA ESTIMATES

Waiting for the first frame.

DECISION LOG 0 events