本当に? [1]
My take on it is that you should have a budget for thinking about AI tools and that budget should not be very much, not more than 10-20% of your time.
I mean, tools are going to evolve, I learned that Gemini can correctly answer programming questions that Copilot can't, so I changed my habits. But I do not go looking for new tools every day, I do not ask who I should be following on X to charge up my FOMO, etc.
There is a lot of talk about why some people get good results with AI coding and others don't. I think it comes down to
(1) Software development knowledge
(2) Subject matter knowledge
(3) A.I. programming knowledge
in that order That is "knowing how to prompt" is mainly about (1) and (2) and less about (3). I might even be wrong about the order of (1) and (2) but the gap between those and (3) is huge.
Let's say you want to write an application to help people do easy tax returns.
It helps to be an accountant or other tax expert because you have (2). I could get away with it because part of my (1) skills is checking out a backpack full of books at the library and learning enough about a subject to impersonate the subject matter expert. Somebody good at (3) could certainly get Claude Code to make something that looks like it could fill our your tax return for you but it won't really work.
[1] Really?
