We present a lightweight text filtering algorithm intended for use with personal Web information agents. Fast response and low resource usage were the key design criteria, in order to allow the algorithm to run on the client side. The algorithm learns adaptive queries and dissemination thresholds for each topic of interest in its user profile. We describe a factorial experiment used to test the robustness of the algorithm under different learningpa rameters and more importantly, under limited trainingf eedback. The experiment borrows from standard practice in TREC by usingT REC-5 data to simulate a user reading and categorizing documents. Results indicate that the algorithm is capable of achievingg ood filteringp erformance, even with little user feedback.