I mean, I get it. If I was ONLY using that 32GB for local AI models then we could take about efficiency and centralization but that’s not the case for most people. When I run local models, I am not running my music DAW with tons of plugins, I’m also not gaming or doing video editing. When I shutdown the local model, I’m usually starting up a game or a recording session. I could just be in the middle of a workday with a lot of teams, Firefox, and vs code running. It could be a video editing day. His concept could be accurate if that memory was used for one thing only on my machine but that’s not the case.

Why you snitching?