<p>I've said what I can about "prompt engineering," why the idea is facile nonsense, the complete lack of theoretical and empirical basis, and the organizational impact of promoting on the basis of "prompt engineering" skills.</p><p>But there's something else to be said, I think, about the underlying fallacy that makes "prompt engineering" appear to be real: namely, the incorrect assumption that LLMs can be "programmed" at all.</p>