> building Rube Goldberg machines that cannot be reasoned about

What a great phrase!, can I use that in conversation?

Isn't a complex computer system, at the lowest levels, an example of a Rube Goldberg machine ?

If we could see the memory bits and the CPU internals, I think it would be good candidate, particularly branching and branch prediction.

There are methods for building complex systems our of smaller pieces. Humanity has done this forever. An internal combustion engine is very complex. An airplane is very complex. You put a boundary around your system, you build it to a spec, you test it to a spec, you define the behaviour across anything you can see. When two engines fail on the airplane it can maybe still at least glide. Maybe there's an ejection seat. Because someone is doing the system thinking with some level of discipline.

A brick is very complex. A bearing is complex. A bolt and a nut and a washer are complex. An integrated circuit is complex. We build stuff out of that. Like buildings that don't fall when you sneeze on them.

In software we use duct tape - at best. But that's not just Microsoft, that's pretty much everyone. With maybe few exceptions.

Software systems are by nature more complex than physical systems. They also lack any tangible interfaces beyond code and diagrams and collective knowledge in the minds of the humans who build and maintain them.. The rapid evolution of software is no help either..if anything AI is accelerating that rapidity as well.

In my experience it is always the "business" people that do not want to take the time to do it correctly.