I enjoyed reading about this! I do not work in AI and only use the most consumer-friendly tools on the market right now (I do not host my own LLM instances), so my technical knowledge is limited. But I like the examination of the field's current direction and the proposal to essentially look toward learning from biological efficiency instead of maximizing a model of how brains work as the sum of many reduced mechanical parts. (As a music teacher, I have been recently very interested in how tactility and generative power were important parts of historical improvisation models whereas modern music theory is often abstract and too slow to be immediately practical.)
A part of me desired a smoother ramp between the pitch itself and the technical details of *how* you are pursuing new approaches to examining human neural architecture. I am not well-versed enough on this subject (either on the biology or on the computer science) to understand many of the links you provided, so if you are pitching to those who are not industry experts, it might help to have a primer article somewhere?