Give Physical AI
a replay button.
Inspect a policy failure. Orbit a deforming sheet. Turn physics into a motion sculpture. Follow the recorded behavior all the way back to its evidence.
Runs were recorded in advance. No install, live inference or simulation service is needed to explore.
Choose your experiment.
Pause, compare, share and download.
Same task.
Change the view.
All 30 trials, failures included. Group paired gains and losses, open the source recordings, and export the full outcome report.
Open the Stress Lab ↗
Same sheet.
Three ways to fall.
Orbit and overlay recorded meshes. Export a 1080p figure, verify a shared sample, and bring the original geometry into Blender.
Open the Cloth Lab ↗
0.05° apart.
Worlds apart.
Turn a small release difference into an inspectable motion sculpture. Follow every original pose and find the largest measured separation.
Open the Butterfly Lab ↗Keep the run behind the image.
Each lab includes its source clock and data. Share a sample, download the evidence, or use the repository's standard-library checks to reproduce the measurements.
Take the scene into Blender.
Cloth and pendulum OpenUSD exports retain recorded geometry and poses. The full project adds Blender films, agent-directed storyboards and native Rerun inspection.
Explore Physics → Cinema ↗The policy experiment covers one LIBERO task and is not an official benchmark or a real-robot result. Cloth coefficients are not calibrated fabric properties; collisions and self-contact are disabled. See the policy media notice, each lab's method, and the source commit and file checksums.