Thinking over things since yesterday, I decided I wanted to try adding some more structure and directionality to the network. Instead of just round-robining edge creation between the different io clusters (a simplification I hoped I could get away with, but think I can’t now), I’m going to create at least 2 or 3 intermediate layers of neurons, then make them link up progressively rather than just cross linked to each other. I think this should cut down on the cycle creation that I think is killing the system.

Also, I realized I hadn’t adjusted the short circuiting logic in the interrogate methods to work with the paradigmatic deconfliction I did earlier, so I fixed that (https://github.com/silasray/aiexplore/commit/9fac38b263bf8c9d990a76a0af37383cbb848d3f) in the time I had to work today. Tomorrow, hopefully I can make it through the structuring enhancements.