Comment by jcadam

Well, the crazy thing I'm working on (100% self-funded thus far): https://trivyn.io. The main idea is moving most of the reasoning to the symbolic layer so the "neuro" piece can be a small model able to be self-hosted on reasonable hardware.