I think there is a stark contrast between the plus or minus micron level tolerances, and whatever is being churned out of LLM ('not getting it perfect all the time')