1. Push for coreML - live update apps with models, approve your app w/ Apple just once
2. Heavy metal for training your coreML models - createML on a laptop stops making sense pretty fast
How does this compare to CoreML? What are the advantages of running models on-device vs in the cloud via an API?