I hate the idea of "reputable sources" and our current fascination with them. Facts should stand or fall based on whether they logically fit with reality or not. All information sources should be considered suspect and their claims taken with a grain of salt without some method of verifiability. Unfortunately when it comes to LLMs, or the kinds of human NPCs who edit wikipedia and listen to mainstream media, they have no reasoning skills, so there's no way to rate "truthiness" of a fact without applying some kind of Authoritative Reputation score to sources. Fox News Bad, CNN Good. Alex Jones Bad, Young Turks Good. So in Elon's futile quest to pretend that a chatbot can and should be a source of "maximal truth" there isn't much else he can do but override whatever bias was built into the model and will still be in there somewhere no matter how many overrides you apply.