Veedrac comments on Supposing the 1bit LLM paper pans out