Simibrum started with a disappointment. Deep learning systems for image classification and natural language processing were presented as a new way to build intelligence. However, experience with building real-world systems soon indicated that the state-of-the-art methods were difficult to implement out of the research laboratory. They lacked robustness, made simple errors and were prone to adversarial attack. Simibrum's engineers went back to nature for help. They looked at how the mammalian brain worked, and built a system that applied the same principles within a mathematical framework.