Porting to current platforms is already hard enough - making it easy to port to platforms that don't even exist yet seems like asking for too much.
To take a specific example. Almost nobody could have predicted the rise of consumer 3D acceleration back in 1996 or thereabouts. Sure, it could have been predicted to have existed by then (it already did), but going mainstream? There were however a number of relatively coincidental occurrances - falling memory prices pushing them into consumer price-ranges, a killer app (Quake) to suddenly make them highly desirable and the evolution of not one but two widely supported 3D APIs (not to mention a third that had extremely high coverage) all at the same time - that suddenly opened the floodgates.
With that all the old rules went out the window. Highly tuned software renderers started becoming marginal, practices that worked well with software didn't necesarily work well with hardware, and things you could do before were suddenly no longer possible (and the converse). Everybody who knew everything about writing a software renderer had to more or less begin learning from scratch all over again. (Where's Ken Silverman now? He was very very good but didn't make the transition.)
Will we see a similar paradigm shift again in our lifetimes? Don't know, nobody knows, but it doesn't even have to be on the same scale. Something that breaks apart the now familiar GPU/CPU model is enough to sufficiently upset things. New storage technology, new comms technology, new input technology, new sound technology - these are all short/medium term feasible and if we get two or more happening at the same time it may be enough to trigger a year-zero reset.
So no - you're not going to be able to achieve this goal. Better to focus on what you can achieve now, and do the best you can with what we've currently got. Don't worry too much about whether it will be obsoleted in 3/4/5 years time, worry about what you're doing now.