I’m proposing a two-stage experiment investigating an alternative architecture for language-model inference. Stage 1 — Build a collective LLM Instead of constructing a language model from conventional neural computational units, investigate a network (or…