@Plinz @jon3k @grok take this thread literally and figure out the scale and logistics required to run your inference calculations manually - humans with paper, pencil, and basic calculators. how many people and what's the speed of tokens output?
More notes on nearby ideas
Quote tweet · Sep 21, 2026if we ever run out of real jobs, llms could always pay us to run a few tokens of inference manually for their amusement - grok says the 8B of us alive…Reply · May 30, 2023probably lol... maybe 3.5, and def eval in batches, and maybe w some heuristics to pre-filter?Tweet · Aug 23, 2023does your average GPT call use more input tokens or output tokens? | are you using 3.5 or 4 more often?Reply · Dec 3, 2023@lumpenspace do you have an AI model of me and does it emit high fidelity tokens?