Opinions on where this is going vary a lot, understandably, because we're going through a time of unusually rapid change. If you even consider how much opinions have changed within the last three years alone, you'll see that any guess of mine is equally likely to have huge error bars. But I'll make a few observations.

The cost of producing software is definitely going down. The speed of producing software is definitely going up. What's unknown is how far these two things have yet to go, and in what time frame. There is a lot of talk recently about Jevons Paradox because it's unclear whether the decreased cost and time of software will lead to much greater demand or whether it'll shrink the workforce. I don't have an informed opinion here, but I'll observe that in my own life, I have been building far more software now, a lot of it custom for my own use, simply because the cost of building has gone down. I've also been building more prototypes that could become useful tools for other people. At the same time, I canceled my contract with a Brazilian programmer with whom I had been working for the past two years building software together because the state of AI has gotten to the point where his salary became way too high in comparison. So, in my own life, both things are going on: I am building more software because its cost has gone down, and I have eliminated the job.

I'm not sure if it will all end up in just taste and judgment, though. If AI building software really is like everyone having a word processor, then it's true that all that remains are the thoughts of the people who know what to write on a page. But I wonder if it's more like a carpenter with a nail gun (something I've personally used before). While it's true that even amateurs can use a nail gun (and occasionally might even hurt themselves seriously), a professional carpenter is going to be much better at using the same tool. Professional carpenters continue to be employed; they're just each personally more efficient than they were before nail guns were invented.

I don't know which direction all of this goes. I will say, however, that the last two decades in software will likely come to be viewed as the halcyon days of software development as an amazing job from a compensation perspective. My intuition, given this uncertainty, is that people should only get into software, as they did in the 80s and 90s, because they truly love the creativity of it, not because they feel like there's some sort of guaranteed awesome job at the other end.