In some month or two we can expect partially-functional thing that involves a person interacting with it. Fully automated is a multi-month thing of labelling and training. Looking for more engineers to join our effort.
I used a similar program from the first year or so of my scripting life. It worked fairly well, but wasn’t perfect because camera angles change and the movements are sometimes lost, but I am hopeful that we can move towards something like this where the AI can decipher what it needs to code better. I am excited to see how this project evolves.