While things may be worse now with AI making source code itself disposable, and perhaps attention spans really are shorter (I wouldn't know either way, though it does sound a bit cliché for old people to complain about Kids These Days):
Time-sucking meetings have always been a problem, the difference now is just that they're online and you can do something else on another screen, rather than been stuck in a chair with a notepad to doodle in.
One of the two worst software developers I've ever had to work with was heavy on the copy-paste in the early 2010s, I think they'd been at it for a decade by that point already. They were using C++ and ObjC with manual memory management (and proud of it!) due to a complete lack of interest in learning the better ways. (The other one was bad in a different way, treated me the way I'd treat ChatGPT).
> The good news is that with this growing competence/compatibility gap it gets easier and easier to identify candidates that can perform versus those that absolutely have no current hope.
Is it, though? AI probably interviews better than I myself do — and yet, my main competitive advantage over AI (and, from your description, over my human competitors) is that I can actually focus on long-horizon tasks. Leetcode, how does {library de jour} perform {task}, what's the difference between {approach 1} and {approach 2}? That's all stuff that most of the LLMs can one-shot.
Back when I was a product manager in the late 80s and 90s, I had a ton of meetings where I would often have to walk to a different building and sit through an hour of meeting with no laptop/connectivity even if only 20% was relevant. You really couldn’t multitask.
Time-sucking meetings have always been a problem, the difference now is just that they're online and you can do something else on another screen, rather than been stuck in a chair with a notepad to doodle in.
One of the two worst software developers I've ever had to work with was heavy on the copy-paste in the early 2010s, I think they'd been at it for a decade by that point already. They were using C++ and ObjC with manual memory management (and proud of it!) due to a complete lack of interest in learning the better ways. (The other one was bad in a different way, treated me the way I'd treat ChatGPT).
> The good news is that with this growing competence/compatibility gap it gets easier and easier to identify candidates that can perform versus those that absolutely have no current hope.
Is it, though? AI probably interviews better than I myself do — and yet, my main competitive advantage over AI (and, from your description, over my human competitors) is that I can actually focus on long-horizon tasks. Leetcode, how does {library de jour} perform {task}, what's the difference between {approach 1} and {approach 2}? That's all stuff that most of the LLMs can one-shot.