Correctness+Latency for Gesture Control and Observation
Speed is nothing without control. To optimise for low latency alone misses the most important part of Physical Ai: correctness. I reduce VLM from ~4Sec. to 33ms gesture input + 500ms per. observation.
