I think we've not even yet tapped this. Not only we're not over "peak software" hump - it's not anywhere in sight, we're not even on the main exponential growth curve.
Software based pill: There's already a pill the uses a sensor to measure it's place inside you're gut and release it medication in a specific location. There's also an in-body device that release medicine on regular intervals.
Software based glasses: augmented reality glasses. They might be a growth area for apps.
Programmable shoes: There are shoes with a programmable pattern and shoes for people with Alzheimer that transmit you're location.
Programmable cars: cars are already programmable. they contain huge amounts of software.
Programmable chairs: some dental chairs are already programmable. You have programs for different seat positions.
My guess is, to create a new software employment boom, you need to create a new mass market programmable platform, that needs a lot of small volume applications and that developing for it is neither too complex(windows gui before VB) and neither too simple(excel).
They are just blossoms here and there, not a mesh of grass.
Everything will not be simply programmable: it would be flexible, changing. Your wall would not be concrete with some software. It would be software with some concrete.
It has the potential to be very relevant. 3D plastic printers are interesting, but too limited at the moment. 3D concrete printing has existed for a while though, that can definitely allow software to be/create the walls of your house.
For medicine it's just a case of getting the simulation down. It'll also be interesting if robots get better. There's a lot of boring lab work in the natural sciences, if this can be automated and controlled by software... Your testing can be done by software... Genetic algorithms could be used to find cures to things. (Although using robots to physically test would presumably be much much slower and more expensive than a computer simulation. :( )
Or perhaps dozens of little robots built the wall in the first place (by lashing together sticks and filling the space between with gravel, say) under software control, and can disassemble it and reassemble it in a different form under software control in an hour or two. In this case the sticks and stones are the pixels.
Here you will have the real thing. Real but no longer boring.