Comment by sublinear

I have never been able to understand this perspective. It's come up so often that it almost feels like a pro-AI propaganda strawman in the style of the "virgin vs chad" meme.

You can and should still write most of the code yourself and use AI to assist when you get stuck. Apart from saving yourself some hours because you no longer hit dead ends on google, I don't understand how that's meaningfully different from using a search engine and finding solutions on StackOverflow and GitHub. I don't even understand how people think this traditional workflow I'm suggesting is any slower than reviewing LLM output either. It takes the same cognitive load and the same time in the end. The only people who believe they can go faster and reduce cognitive load aren't just inexperienced, but also seem to have ADHD or something. Like they just wanna bask in the vibes of the machine-generated text scrolling by.

If your goal is to make something you're going to live with and maintain long term, or is more generally the foundation of a bigger idea, you can't ignore any part of the code or delegate it away. It's still writing that you're bound by like any other document.