Copper now goes end to end. We added an open-source ViTFly-based autonomy example that flies through complex environments at high speed, while also deploying an NVIDIA TensorRT-based self-driving stack with a customer. From Candle 🤗 to CUDA, Copper brings learned models into the same deterministic, observable, and replayable robotics runtime as the rest of the system.

If the “small brain” in your robot can fully understand and verify what the “big brain” is doing, why did you build the expensive big brain in the first place?

This is the trap autonomous robotics has been stuck in for a decade…