CAD and Classic Engineering are getting saturated now too.

I can't speak for math, but from a SWE perspective I'd say it's a very mixed bag.

One thing is clear that AI produces software faster, but the net results of that haven't really shown to be positive. There's undoubtedly more software now, but quality isn't improving. There's real downsides too like professionals unskilling themselves, and the cycle of low quality source being fed back into training.

Please show evidence that quality isn't improving. Also, maintaining same quality but delivering more is still a big win.