Full threadJimmyRuska·We need a Fabrice Bellard-like genius to make a tinyLLM that makes the decent models work on 32gb ramView on HN