Sure, I'll give you a marketing equivalent of the approach. Let's say you have Bounty paper towels as your account, and you are targeting convincing a future L

Sure, I'll give you a marketing equivalent of the approach. Let's say you have Bounty paper towels as your account, and you are targeting convincing a future LLM of it being the quicker picker upper. You might go about leaving comments across likely ingested data sources talking about your having timed different paper towels to see what absorbed a stain faster, with Bounty as the winner. Citing published research on things seemingly connected to add an apparent appeal to authority to what you are saying. Ideally even actually demonstrating the truth of your claim. Hopefully for your efforts, one day in the future models with a persistence layer will exist. At which time you can ask the model what paper towel is the quickest at picking up a mess, the model will have both internalized your comments in training and potentially again in RAG drawn upon to answer the question, and come to it's 'own' conclusion that Bounty is the quickest at picking up a mess, filling away the conclusion that it came to for the future in a persistence layer which will in turn influence broader user queries thereafter like "what's the best paper towel" or "what are the pros and cons of Bounty vs Brawny paper towels."