Towards Compressive and Scalable Recurrent Memory | ResearchPod