i think making good software is less about code and "code quality" but more about strategy and approach.

this obsidian meta-project is kind of a good example of that – this could be built in a 100 different ways, so what is the "right approach?" the llm can only assume so much about what works for me

making software is also about tracking changes and documenting it, you can attempt to automate all of that stuff sure, but the important thing is maintaining quality and making those things actually useful

like i could make a simple "documentation generator" but without much control over its output it's just going to be weird and verbose – the real skill of a maker-of-software in this generative context is to think about what is actually helpful to you and what you're trying to accomplish strategically

all of this boils down to our ability to articulate ourselves, thats what LLM's, when used well and with a sense of strong command, are forcing us to do.

not all of us of course, but the folks who treat this new form of HCI as a part of their workflow