Combining large amounts of publicly available structure-activity data with assays that have carefully curated annotations opens the door to a number of ways to analyze the data behind the scenes. Combining fully machine readable input for a diverse variety of projects with modelling techniques that can be used without fussy parametrization allows models to be created and updated whenever new data arrives. Predictions from these models can be integrated into normal searching and visualization workflows, without any need for the user to opt-in or make extra decisions. This approach is novel and different from the way structure-activity models are normally deployed: useful predictions can be presented ubiquitously with literally zero additional work on behalf of the user. We will present our efforts to date regarding ways to both passively and actively draw attention to important drug discovery trends while exploring compounds and assays.