I think he's mostly right. I think it's driven by some of the same panic that pushes so much of the AI boom: companies trying to find the "next big thing" in software. A real possibility is that we are near the end of the era of software innovation. Not that there will be none, but that it will slow down, because so many problems are, well, solved -- at least to the point where most potential customers are satisfied with current solutions.
People will continue to innovate, and some of that will involve software, but it doesn't mean there is a desperate need for software innovation.
There was a point where radio innovations were coming fast and furious. There was a point where mechanical engineering was progressing in leaps and bounds. Fields eventually slow down. It didn't put mechanical engineers out of work. But it's not the kind of field it was in (guessing) 1850.
I also think he's mostly right but even solved problems can still see much improvement. Take for example algorithmic improvements to improve speed or reduce memory requirements. Then there's also massive power usage improvements to be had by re-implementing existing solutions using more efficient languages [1].
People will continue to innovate, and some of that will involve software, but it doesn't mean there is a desperate need for software innovation.
There was a point where radio innovations were coming fast and furious. There was a point where mechanical engineering was progressing in leaps and bounds. Fields eventually slow down. It didn't put mechanical engineers out of work. But it's not the kind of field it was in (guessing) 1850.