Learning is fundamental for interacting with a changing environment. Critical to this process is the ability to rapidly adapt previously acquired skills to external perturbations while avoiding catastrophic forgetting of previous tasks. For the learning system, this corresponds to a trade-off between being able to maintain stable memories of learned tasks, and the ability to swiftly adapt such memories to changes. We hypothesize that such a problem can be addressed by using a two-timescale [1,2] strategy, where memorization and adaptation are performed by two separate modules. This structure allows the use of sub-networks that are best suited for each of the two objectives, i.e. adaptation and memory. By imposing a specific interconnectivity between the two modules, we extend the classical memory-adapter model (Fig. 1.a, [3, 4]), showing how credit assignment for the memory network can be simplified (Fig. 1.b), making it possible to consolidate the adaptation signal through a simple, local plasticity rule. The model was tested on a centre-out reaching task [5,6], where the two modules, implemented as rate neural networks, jointly generate the behavioral output by controlling a two-joint motor plant (Fig. 1.c). After an initial training phase, the system has to perform motor adaptation to a rotational perturbation of the target trajectories. We find out that our system outperforms a standard recurrent neural network, both on adaptation and washout cases, showing features resembling real-life evidence, being able to implement the two-timescale learning strategy (Fig. 1.e-f). We finally argue that our system might represent a model of the mammalian cerebello-cortical loop (Fig. 1.d), given the resemblances in the features shown by the two [7]. We conduct ablation and inhibition tests on the system, qualitatively reproducing some of the recent experimental evidence on mice [8,9]. Altogether, our results provide a neural substrate hypothesis for distributed motor learning in the brain, and that can be easily bridged to generalized, non-motor learning.