LLMs have a hard time producing simple code. One of the reasons is that determining what is "simple code" requires a lot of context that lives *outside* the code & docs. It depends on your team dynamics, your company business strats, your suppliers or the latest trends. Peter Naur's paper "Programming as Theory-building" (1985) states this knowledge in devs minds **is** actually the program, the Theory w/ capital T as calls it. LLMs are increasingly solving & automatic code & docs, but that's just a small piece of the full cake. Being aware of it helps understand many LLMs failure modes and also how to more effectively leverage them. To learn more watch my talk @PyBCN last week! youtu.be/LwchtmM-d-s