I think code quality is always important, but it's less important for terminal code which implements a specific set of features using robust APIs underneath. Low code quality becomes a much bigger issue when the code is supposed to deliver well-defined abstractions on which other code can rely.

If you are a serious software developer, then you will probably be able to explain both what your code does (i.e. what spec it implements) and how it does it (what does it call, what algorithms does it use, what properties does it rely on?). With the advent of LLMs, people have started to accept not having a clue about the "how", and I fear that we are also starting to sacrifice the "what". Unless our LLMs get context windows that are large enough to hold the source of the full software stack, including LLM-generated dependencies, then I think that sacrificing the "what" is going to lead to disaster. APIs will be designed and modified with only the use cases that fit in the modifying agents context window in mind with little regard for downstream consequences and stability of behavior, because there is not even a definition of what the behavior is supposed to be.