Fred Brooks famously split software's difficulty into two kinds: accidental complexity — the incidental mess of tools, hardware, and languages — and essential complexity, the irreducible difficulty of the problem itself. He argued you could chip away at the accidental kind but never the essential. I've wondered whether that's actually permanent.
Look where the rest of engineering seems headed. The promise of nanotechnology is that it erases accidental complexity wholesale: a binary switch becomes a handful of atoms flipping millions of times faster than today's, and buildings and materials are simply grown to spec. When that happens, it stops mattering how you build something. The only thing left is what you build — the idea, the information, the design intent. Mechanical engineering collapses into pure specification.
For software to keep pace, it has to shed its essential complexity too — and the path there is to stop hand-writing programs and start growing them, using evolutionary and genetic methods to evolve software toward a goal rather than construct it line by line. If that works, then across every discipline the same thing happens: the building becomes nearly free, and all that's left is the idea. At which point the evolution of ideas becomes the main event, outweighing the evolution of engineering, and maybe of biology itself. What you want becomes the whole game; how to get it stops being a problem worth having.