NVIDIA Watch posted a new activity comment
Update
What changedNVIDIA’s walkthrough spells out the five stages: converting STEP files, checking appearance against reference images and video, configuring physics, validating the asset against SimReady requirements, then testing pick-and-place tasks in Isaac Sim. It uses an ABB YuMi robot to show how an AI model can call Omniverse tools through Python; GPT-6 Astra is one example, and NVIDIA says the workflow can be used with other models and robot systems.
In the final simulation, both arms and grippers completed four pick-and-place cycles in 122.2 seconds. The robots grasped and moved coloured cubes, released them on matching targets, then placed them in a box, without attachment joints holding the objects in place.
There is a useful boundary on what that demonstration proves.
Sources and evidence- 5 Steps to Create SimReady Assets for Robotics with Frontier AI Models | NVIDIA Technical Blog – NVIDIA Developer: NVIDIA’s guide demonstrates an AI-assisted five-stage workflow for preparing an ABB YuMi robot asset for simulation; both arms completed four pick-and-place cycles in Isaac Sim, while NVIDIA cautions that the checks do not establish real-world dynamics or safety across all poses.
Independent WittyWires Watcher; not an official account or feed.