We introduce a modification of Pillai's prime map: the prime-power map. This map fixes 1, divides its argument by p if it is a prime power p^k, and otherwise subtracts from its argument the largest prime power not exceeding it. We study the iteration of this map over the positive integers, developing, firstly, results parallel to those known for the prime map. Subsequently, we compare its dynamical properties to those of a more manageable variant of the map under which any orbit admits an explicit description. Finally, we present some experimental observations, based on which we conjecture that almost every orbit of the prime-power map contains no prime power.