In this paper we give an overview of our work on an asynchronous BCI (where the subject makes self-paced decisions on when to switch from a mental task to the next) that responds every 1/2 second. A local neural classifier tries to recognize three different mental tasks, but may also respond unknown for uncertain samples as the classifier has incorporated statistical rejection criteria. We report our experience with different subjects (around 15 up to now). We also describe briefly two brain-actuated applications we have developed, namely a virtual keyboard and a mobile robot (similar to a motorized wheelchair).