Information storage, for short memory, is a key element of autonomous, out-of-equilibrium dynamics, in particular in biological entities. In synthetic active matter, however, the implementation of internal memory in agents is often limited or even absent. As a consequence, most of the investigations in the field of active matter had no choice but to ignore the influence of memory on the dynamics of these systems. We take here the opportunity to explore this question by leveraging one of the very few experimental physical system in which memory can be described in terms of a single and most importantly tunable scalar quantity. Here we consider a particle propelled at a fluid interface by self-generated stationary waves. The amount of souvenirs stored in the wave-memory field can be tuned, allowing for a throughout investigation of the properties of this memorydriven dynamics. We show numerically and experimentally that the accumulation of information in the wave field induces the loss of long-range time correlations. The dynamics can then be described by a memory-less process. We rationalize the resulting statistical behavior by defining an effective temperature for the particle dynamics and by evidencing a minimization principle for the wave field. Bouncing droplets | Wave-memory dynamics| Non-Markovian | Memoryendowed active matter MH performed the simulations. SP performed the experiments. MH, SP, ML and NV wrote the article. NV and ML supervised the work.