I disagree, but I agree that this is the wider perception.
> In this context, unless you can write code faster than a machine, you have no choice
This is the main misconception. Speed isn't as valuable as reliability and correctness. Is it better to code something that takes 2 hours but constantly breaks or something that takes 5 days but never breaks? The problem is that it's hard to quantify whether the 2 hour unreliable code is a good tradeoff. How many hours of fixing will there be, and what other things is it going to cause as a result of breaking the code? It could be simple bug fixes but it could also be production downtime. In order to avoid all of that you can simply code something with strong guarantees it will work, and it will work well.
Unfortunately I think the industry has spoken, and they are willing to go down that path of doing it as quickly as possible at whatever the cost. After all most companies will hire many mid-level engineers, and that's easier than finding a few really good ones. So I guess my point is that while this is changing across the industry, either things will start breaking and AI will become less used as everyone realizes you really want reliability over speed, OR AI does get better and better and then engineers are truly replaced. Nobody has to think of coding ever again. Right now we're in this limbo phase where we pretend AI has changed things but engineers are still needed. I don't think there is an in-between here.
I disagree, but I agree that this is the wider perception.
> In this context, unless you can write code faster than a machine, you have no choice
This is the main misconception. Speed isn't as valuable as reliability and correctness. Is it better to code something that takes 2 hours but constantly breaks or something that takes 5 days but never breaks? The problem is that it's hard to quantify whether the 2 hour unreliable code is a good tradeoff. How many hours of fixing will there be, and what other things is it going to cause as a result of breaking the code? It could be simple bug fixes but it could also be production downtime. In order to avoid all of that you can simply code something with strong guarantees it will work, and it will work well.
Unfortunately I think the industry has spoken, and they are willing to go down that path of doing it as quickly as possible at whatever the cost. After all most companies will hire many mid-level engineers, and that's easier than finding a few really good ones. So I guess my point is that while this is changing across the industry, either things will start breaking and AI will become less used as everyone realizes you really want reliability over speed, OR AI does get better and better and then engineers are truly replaced. Nobody has to think of coding ever again. Right now we're in this limbo phase where we pretend AI has changed things but engineers are still needed. I don't think there is an in-between here.
Writing code by hand is something from the past? Sometimes I think about this. Here is one of these thoughts.