Because we're all working at different companies on different codebases in different languages doing different things, we're all talking abstractly about something we think is the same, when really it isn't. Obvious to every programmer is to make a library if you're copy and pasting code multiple times,
but the overhead of that means you don't do that if you only do that once or twice. the problem, in humans, as with LLMs, is the context window. after running create-react-app or whatever, there are a number of steps to do that are repetitive, but because that doesn't happen often enough to fully automate it, you just do it manually. LLMs let you do that level of boilerplate without all the overhead of manually configuring snippets in the IDE.
