Arthur Hayes: FLOP miners should adjust hardware configurations according to actual inference needs.
Arthur Hayes stated on the X platform that FLOP miners' block rewards are calculated based on their proportion of the inference computing power they process; miners who process half of the total amount of inference power will receive half of the block rewards. If the purchased hardware is not compatible with the required AI model, miners will be unable to process inference requests and receive rewards. Miners can refer to popular models on OpenRouter to predict the models needed for the agent's work and adjust their configuration based on actual results.