Next up, our own Ignace Bleukx on an efficient hyperparameter method that combines SMBO with addaptive capping (=what exact solvers really want), and for which you can learn priors to speed up initial iterations further (they matter most when doing adaptive capping)