Tesla has unveiled what may be the most impressive demonstration yet of its humanoid robot, Optimus, showcasing a suite of new skills that mark a major leap forward in artificial intelligence and robotics. To the average eye it may seem mundane, but Optimus is autonomously performing a wide range of domestic tasks — all learned from watching humans on video.
From taking out the trash and vacuuming to stirring food in a pot and recognizing specific auto parts, Optimus now completes everyday actions without the need for direct programming or remote operation. Tesla says all the tasks were executed through a single multitasking neural network — a powerful AI model trained on first-person human demonstration videos.
The robot’s performance in the video goes beyond simple mechanical execution. It demonstrates smooth movement, spatial awareness, and coordination in tasks like opening cabinets, closing curtains, and using a dustpan to clean up a table. Beyond the home, Optimus also correctly identifies a Model X fore link, retrieves it from a labeled box, and places it on a dolly ramp — a sign of its growing utility in factory settings.
ALSO READ: Tesla Reveals More Info On Robotaxi Launch in Austin, Optimus Production Plans, and Dojo
Tesla AI lead Milan Kovac explained that the company recently made a significant breakthrough in training methodology. Instead of relying solely on manually collected data or teleoperation, Tesla can now teach Optimus directly from videos of people — even footage captured from a first-person perspective. This dramatically accelerates learning and enables the robot to generalize across a variety of real-world scenarios.
Kovac also noted that these new skills can be activated through natural language, either by voice or text. The simple commands were also included in the video.
Tesla has released a new video of Optimus performing a variety of household tasks, all of which were learned from watching videos of humans doing these tasks. https://t.co/WXt0PpC9cu
— Drive Tesla (@DriveTeslaca) May 21, 2025
