We have just-in-time compilers, garbage collectors, profilers, and all the rest. And AI and Expert Systems are getting better all the time. It's not unreasonable to expect that actual writing of code by machine is going to continue to get better.

But, on the other hand, just today I downloaded the drivers and utilities for a Windows-connected Dymo label writer.

165 MB installer.

It's ridiculous. :-/ (Of course, that was human written code that was dragging around some giant Windows C++ Runtime package, also too.)

I'm also reminded of my time working for a small banking software company, just after grad school. I was working on a hypertext "help" system for their banking software. It crawled on a 286 (as you might imagine), and when I told the group in a meeting about it on guy said, "Great! That means we can sell them more powerful computers with it too!!" :-/

drook is right that the modern way is to throw more hardware at the problem, until the next paradigm-breaker comes around to make people start down a different path ("What do you mean that it's faster to do the computations on my $400 graphics card than on my $1200 CPU??"). It's wasteful to not care about code efficiency and instead always think that progress depends on eliminating people with expertise, and makes it much, much easier for bugs (and back doors) to creep in. But until there are actual penalties for bad and inefficient code, it's hard to imagine things changing.

Of course, my copy of "Spontaneous Assembly" never got much of a workout from me, so... ;-)

Cheers,
Scott.