I’d like to thank @axisroboticsTR and @Ferhat_NZ for the merch package and the jest. While opening the package, I noticed this latest demo shared by @axisrobotics. They’ve trained expert models using reinforcement learning in simulation and directly transferred them to a real robot via visual randomization. In the test shown, the objects’ positions aren’t fixed—they’re completely randomized—but the robot completes the task without any issues. Bridging the gap between simulation and the real world in physical AI is one of the hardest challenges. It’s great to see concrete, working models—not just talk—and kudos to the team for making it happen.