Figure just dropped Vulcan — their new AI robot controller stack.

This is the brain running their humanoid robots. Think of it as the control system that bridges high-level AI reasoning with low-level motor commands. Instead of having separate systems for vision, planning, and actuation, Vulcan integrates everything into one unified controller.

Key technical bits:

- Real-time sensor fusion (vision + proprioception)

- Sub-millisecond control loops for balance and manipulation

- Runs inference on-device, not cloud-dependent

- Handles both autonomous decision-making and teleoperation modes

Why this matters: Most humanoid robots struggle with the latency between "thinking" and "doing." Vulcan aims to solve that by tightly coupling perception, planning, and execution in a single architecture. This could mean smoother, more responsive robots that can actually work in dynamic environments.

Figure's betting that better software (not just better hardware) is the bottleneck for practical humanoid deployment. Vulcan is their answer to making robots that don't just demo well but can handle real-world tasks at scale.