RoboDojo sim "not even wrong": blog exposes physics flaws in popular robot benchmark
ZeYanjie · x · 2026-11-18
A blogger dissects RoboDojo, a robot simulation benchmark built on NVIDIA Isaac Sim (used in Fable 5.2's model card), showing videos of severe physics failures and concluding the simulation is "not even wrong" — finally understanding why robotics experts "hate" past simulators.
The author still argues from first principles that LLM-powered agents plus minimal but necessary real-world calibration is the right path to embodied intelligence: the real world is slow, expensive, hard to scale and label, while sim is not — so a trustworthy sim would partially solve the robot data problem and unlock compute scaling. The thread's commenters call for someone to actually fix RoboDojo's physics.
More from Embodied
- NVIDIA-backed Agile ONE SN humanoid robot for researchers ships early 2027 — xmercury_one · 2026-11-18
- Peking University Open-Sources PanDWM, a 360° 4D Driving World Model Beating DiST-4D — PekingUniversity · 2026-11-18
- Humanoids serve ice cream and spar in Shenzhen while APEC leaders visit — TansuYegen · 2026-11-18
- Otto by Rockwell explains using Gazebo digital twins for AMR deployments — ___Mufasaa · 2026-11-18
- Tokkey debuts thumb-sized AI hardware to run local models on your Mac — TejasKumar_ · 2026-11-18
- 'Nothing more physical AI than moving tons of rock': live demo at Web Summit shows real-world difficulty — MarvinTBaumann · 2026-11-18