Pearl and AI inference
Pearl's proof of work is a matrix multiplication, so the cards that serve AI models can mine it. What that earns, what it costs the inference beside it, and how much it could take off what open models cost to call.
Which of Pass’s three worlds is Pearl in?
Rafael Pass’s economics of proof-of-useful-work sorts a network like Pearl into three equilibria by two numbers: whether running the miner beside inference beats splitting the card (Δ), and how large block rewards are against the inference market (θ). Δ and the threshold θ must clear come from the overhead below.