This prototype uses carefully selected gear ratios to mimic the weight adjustments in a neural network. The mechanical system processes inputs to produce outputs analogous to a digital neural network, all while operating with minimal energy consumption.