It wasn’t just the LLM language, the sections “The corpus” and “What the data supports” were disjointed and unconvincing. Likely due to lack of human wrangling. Unfortunately they were the core of the analysis.