Um, that's amazing... and it's free?
Maybe I'm thinking too far ahead but I'm hoping maybe I can cover the (future?) compute costs by charging for the heavy duty ML that would have to be run non-locally but IDK. This isn't critical yet but would love to hear your musings.