This talk presents a statistical testing framework that allows researchers to analyze an incoming i.i.d. data stream with any arbitrary dependency structure, and test whether a feature is conditionally associated with the response under study. By processing data points online, we can stop data acquisition when significant results are detected while using powerful machine learning algorithms to enhance data efficiency. To develop our method we draw inspiration from the model-X conditional randomization test and testing by betting, resulting in a flexible approach for sequential conditional independence testing.