I wanted to know whether Apple Silicon performance gains + local LLM models - frontier labs' starvation for energy would ever yield a world in which your MacBook Pro can deliver "good enough" LLM inference compared with the frontier labs. Turns out, that may be coming in a year or two.