As the anthropogenically forced climate signal continues to emerge and demand for skillful Earth System predictions on all timescales grow, ensuring accurate representation of key processes in Earth System Models (ESMs) is increasingly critical. This presentation will discuss three lines of work focused on process fidelity across timescales, spanning long-term climate change, seasonal prediction and decadal variability. For the long-term change, recent progress in understanding the inability of ESMs to represent historical atmospheric humidity trends, a quantity that is highly relevant for wildfire, will be discussed. On the seasonal timescale, improvements in the representation of the stratospheric Quasi-biennial oscillation and its associated teleconnections will be considered, and on the decadal timescale, the inability of ESM’s to represent multi-decadal variability in the North Atlantic jet stream, likely through ocean-atmosphere coupling will be discussed. Together, these examples will illustrate the importance of improving process representation across timescales for building confidence in both projections and predictions of the Earth System.