I am dreading our LLM-written incident report future (opens in new tab)

LLM-generated incident reports pose unique dangers because their plausible-sounding errors lack immediate verification mechanisms that catch problems in code…