Climate predictions are our most valuable tools to support socio-economic decision-making from regional to local scales and develop successful adaptation strategies. Their level of accuracy is determined by our understanding of the climate system and our capacity to simulate it correctly. While current prediction models exhibit high skill in predicting sea surface temperature in key regions – like the North Atlantic or the Tropical Pacific – from months to years in advance, their ability to predict the atmospheric circulation and through it the continental climate is undermined by structural model problems. These problems point to a misrepresentation of key processes and interactions. PREDDYCT focuses on the North Atlantic, the region where the structural problems manifest more clearly. Its aim is to bring a new fundamental understanding of the physical mechanisms that need to be realistically simulated to provide climate predictability to its neighbouring continents. The main working hypothesis is that mesoscale eddies – whose contribution is unresolved in current prediction models – are the key element. This is supported by the fact that in the North Atlantic region, resolving the effect of mesoscale ocean eddies and the atmospheric eddy feedback onto the midlatitude jet has been shown to critically improve the realism of air-sea interactions and their influence on the large-scale atmospheric circulation. In light of this, PREDDYCT will combine a new generation of predictions at ground-breaking resolutions, an innovative tuning framework to enhance their accuracy, and extensive process-oriented analyses to, for the first time, resolve and understand the contribution of mesoscale eddies to North Atlantic climate predictability. PREDDYCT’s new conceptual and methodological insights will pave the way for further forecasting advances at the global scale and contribute to achieve a long-awaited breakthrough in the realism and trustworthiness of climate predictions.