I’m personally excited by this paradigm. A few years back I had success using a similar architecture for polynomial root finding. I think it’s entirely possible to be really ambitious and reverse engineer new and useful generalized functions.