The AI learning issue

AI is super useful for explaining how a function works. It's also helpful for finding libraries to import into your module, and even writing code quickly. I'm not sure what general consensus is, but I for one am not a fan of using AI for anything besides clarifying how something works when I'm learning.

I don't particularly enjoy being a "prompt engineer" - something I've experimented with. I will say, yes, code can be written quickly, and the model generally explains as it codes. Everyone says, "just review the code like you normally would when inspecting a pr," and I get it. Perhaps this is where the industry is headed. Maybe 10 years from now, all of our software will be "vibe-coded." But I do not enjoy it one bit.

I learned computer science and programming to write code, not watch a computer write it. I don't really care if it's faster, as odd as that may sound. However, I do find it useful for automating tasks that I myself don't enjoy doing. Mainly writing documentation, and GitHub READMEs (and always reading through what it's written, editing as needed).

What about using AI to help me learn? AI can be great for finding resources. I would say just don't use it to write the code for you (at that point, what are you learning?) Typing the code yourself is like building muscle memory, even if it's boilerplate.

Seeing where AI will go next makes me both excited and uneasy. On one hand, agents such as Claude come in handy for explaining snippets of code to me. On the other, I don't want my editor to be infiltrated and poisoned by Copilot.

If you're an experienced programmer, taking the time to learn how to be most effective (both in terms of functionality and token consumption) when using LLMs will ensure that you're adapting to the current tech field. If you're a newbie, really focus on the basics of programming and computer science. Then, when you use AI for more complex tasks, you'll be more effective in your pair-programming.