 Recent advances in brain-machine interface, BMI, technology have made it possible to predict seizures. However, transmitting large volumes of electrophysiological signals between sensors and processing units, as well as the associated computation, present two major bottlenecks for BMI systems. To address these issues, we developed C2SPnet, a framework that combines compression, prediction, and reconstruction without requiring any extra computation. This framework uses a plug-and-play compression matrix to reduce transmission bandwidth requirements, allowing the compressed signal to be used for seizure prediction without additional reconstruction steps. Additionally, reconstruction of the original signal can be performed in high fidelity. Our experiments show that C2SPnet is more energy efficient than existing approaches and achieves higher prediction accuracy. This article was authored by DiWu, Yixue, Ziyu Wang, and others. We are article.tv, links in the description below.