General-purpose robotics requires three things: compute, models, and data. Compute is already concentrated in the hands of a few. Models only scale if the data exists to train them. Unlike LLMs, which were able to feast on the open internet, robots have no equivalent corpus of real-world interaction. Simulation and teleoperation help, but they don't scale, fail to capture the long tail of reality, and break down at deployment.
We are building the infrastructure for a post-scarcity world by scaling physical-world data. A future of abundance begins with robots truly understanding the world we live in.
Reach us at shashin.bhaskar@gmail.com