Regarding LLMs and the State of the Art in Software Development, I'm wary of the notion that AI has a negative impact for developer productivity. And to clarify, I don't doubt "... that (currently) when developers use AI tools, they take 19% longer than without—AI makes them slower" (ref).
What I do assert, is that the Software Development discipline needs to update their practices and potentially create new workflows to take advantage of the leverage that LLMs provide. I think Sean Grove's gauntlet (ref) that "specifications, not prompts or code, are becoming the fundamental unit of programming", is a step in that direction. Specifically that
"In an era where AI transforms software development, the most valuable skill isn't writing code - it's communicating intent with precision. This talk reveals how specifications, not prompts or code, are becoming the fundamental unit of programming, and why spec-writing is the new superpower."
LLMs and Developer productivity have been a recurring point I've encountered with my friends and colleagues. Specifically that the development of the best-in-class / frontier models, requires people and organizations to rethink their entire workflow and set of patterns for developing software. This is a point I touched on in my post "Artificial Intelligence and Human Progression" (ref).
But the point bears repeating as it surfaced again in a text exchange I had with a friend. An exchange where a famous Paredit quote served as an analogy to the change that I percieve is required. The quote I think requires a bit of context for non LISP / Emacs users: "If you are not the kind of person who can deal with Paredit or smartparens, you need to become that kind of person". It's like that, but with LLMs.
That quote has a special lore in the Lisp / Emacs community. Most editors, including Emacs, have users operate on text as the basic building block. Paredit though, treats source code as a data tree, allowing users to navigate and operate on nodes in that tree. Before the tool and accompanying philosophy came into fashion, the uninitiated pushed back with the objection that they were unaccustomed to this kind of editing. Hence where the quote derives. And I've experienced the epiphany myself... Once you've used Paredit to navigate and manipulate a tree, there really is no going back to editing source code just as blocks of text.
The tool offers such an advantage and paradigm shift that the user needs to re-think and update the way in which they approach their editing. And my assertion is that that maxim equally applies to the advantages that LLMs provide to the software discipline writ large. Having such a powerful tool at hand, I think, requires people and organizations to rethink their entire workflow and set of patterns for developing software.