Engineer Flags Missing Evals in Local LLM Shares

Nate Raw posted that the most annoying thing in the local LLM space is everyone sharing their deployment speeds but mentioning nothing about evals. The post comes from Raw, a Senior ML Engineer at Splice with prior experience at Hugging Face and Lightning AI. It was retweeted by Charles Frye, an engineer at Modal focused on serverless GPU infrastructure and LLM inference optimizations. The statement stands as Raw's direct observation on visible posts in that community.