Action anticipation and forecasting in videos do not require a hat-trick, as\nfar as there are signs in the context to foresee how actions are going to be\ndeployed. Capturing these signs is hard because the context includes the past.\nWe propose an end-to-end network for action anticipation and forecasting with\nmemory, to both anticipate the current action and foresee the next one.\nExperiments on action sequence datasets show excellent results indicating that\ntraining on histories with a dynamic memory can significantly improve\nforecasting performance.\n