1. No production implemented the feature we open sourced. This is unique. 2. Whole thing(AI Inference/Classifier training) are offline. 3. ARM did it.

1. No production implemented the feature we open sourced. This is unique. 2. Whole thing(AI Inference/Classifier training) are offline. 3. ARM did it. 4. We have classifier training on the embedded system. 5. I calc the theory/real, arm is better than Nvidia GPU.