“Effective prompt engineering for language models involves a combination of philosophical clarity in defining concepts and empirical, iterative testing of edge cases.”