Opinion: Astra could become the brain of OpenAI's robots
The author argues that OpenAI is building out the full robotics stack, and Astra's perception, planning and long-horizon execution make it the first suitable high-level brain for it.
One of the most underrated things happening at OpenAI right now is robotics. This is no longer some tiny side experiment. OpenAI is building across the entire physical stack: control systems, actuators, sensors, inference, real-world robot workcells, and models capable of operating at a much higher level of abstraction. And Astra may be the first OpenAI model that genuinely makes sense as a high-level “brain” for that stack. We’re already seeing it plugged into robotics experiments, and its strength in perception, planning, tool use and long-horizon execution makes the direction pretty obvious. At the same time, OpenAI just acquired Glass Imaging, while its device teams are explicitly working on cameras, microphones and other sensors to help AI understand people and the physical world around them. I’ve seen enough of this area to say that people are seriously underestimating how fast OpenAI’s physical-world work is moving. The robots are getting a brain now.
This post quotes another post: x.com/scav/status/2099648931152191977