lowspecai

I publish measured notes on running local AI on hardware people already own: memory bandwidth, tokens per second, and used mini PCs.