Extreme heat waves and heavy rainfall are increasing in intensity on a global scale, trends which will continue with future global warming. Summer, with most biological and agricultural production, is probably the season when changes in extremes will have the most-severe impacts on humanity. Summer extremes are particularly devastating when they persist for several days: Many consecutive hot-and-dry days causing harvest failure, or stagnating wet extremes causing flooding. Despite this importance, persistence of extreme summer weather has largely been neglected by the climate science community. What maintains stagnating summer weather? Do climate models capture persistence and the underlying processes accurately? What is the role of global warming? Persistence is linked to sea-surface temperature, soil moisture and atmospheric circulation which are expected to change with future warming but the uncertainties are large. The proposed research fills this knowledge gap. I will study mid-latitude summer circulation and its influence on weather persistence focusing on the most high-impact, persistent summer extremes. The project innovatively combines novel methods from disciplines which historically evolved largely independently: (1) Machine learning guided by physical theory and (2) climate modeling experiments using state-of-the-art global Numerical Weather Prediction models. I will quantify persistence of summer extremes and their local and remote drivers in past, present and future climates, focusing on Western Europe and eastern U.S., i.e. two major population centers critical for global food production. In recent publications, I have reported a pronounced weakening of boreal summer circulation since 1979 and hypothesized that this leads to more-persistent, and therefore more-extreme, summer weather conditions. This overarching hypothesis will be tested, using the described methods, in observations and modeled data at different warming levels. This work will reduce societal risks from future summer extremes by improving existing forecasts and developing novel early warning systems based upon optimal empirical prediction methods.