Solution? Move processing to the edge. I want a local model on my desktop. I want training to be some peer to peer open source processing network. We did it for SPA's time to do it for LLMs. I'd pay up to 3000 for a compute module that can do this stuff locally.