Model stacking is an ensemble technique that involves training a model to combine the outputs of many diverse statistical models, and has been shown to improve predictive performance in a variety of settings. 'stacks' implements a grammar for 'tidymodels'-aligned model stacking.