to each their own. if you are having really high-sensitive conversations with your GAI that someone would bother snooping in your docker container, figuring out

to each their own. if you are having really high-sensitive conversations with your GAI that someone would bother snooping in your docker container, figuring out how you are doing inference, and then capturing it real-time - you have a different risk tolerance than me. i do think that cloud GPUs can cover most of this experimentation/learning need.