Today's protip: If you *must* interact with a large language model, try telling it to only answer you in ASD-STE100 style. Even Gemini Flash (the lightweight model built into free Google searches) understands this directive, which constrains its output to conform to concise, clear scientific manual style. This also sharply reduces the amount of hallucinated output you get--but you do still need to CHECK what it says, because it WILL still hallucinate, despite the focus on brevity and clarity.