Put him on a treadmill to power a data center

I like LLMs but I want to run them locally and for that I need RAM that I can't really afford. Especially Chinese models are good candidates to run locally, I wonder if that's part of the US strategy, I mean making local inference so expensive that people have to use their models if they don't want to send all their data to China

As if this trick wouldn't work on any of us...

Damnit, Sam! What are you doing under my desk?!?

Infinite energy unlocked!