>The last 50 years of research have no impact at all at current day computing, except for silicon and SOCs etc.
Doing new stuff in silicon is like shooting fish in a barrel. You just go find some research paper that has a good idea, then go and design it in hardware. I little hard math is enough to ensure that no one else is doing it.
Software seems stuck in a research time warp. Plan 9 was how long ago? Yet we don't have network transparent multi processing and per process name spaces in popular OSs.