OTOH, Linus has put together a team that has put together an OS kernel that runs on everything from a mobile phone or a Pi Zero to the top N of the world's supercomputers. And Kieren?
OTOH again, not all his ideas have been great:
He's excoriated RDBMS developers for preferring to deal with raw disk rather than go through the file system. The reason they do this, of course is that plus or minus any buffering in smart interface boards or the drives themselves, they know that when a write call returns the data's on disk. As a sometime DBA that's the way I liked by database engines to work. Eventually, as available memory grew he realised that file system writes could be buffered a long time and the data was at risk if the machine failed in the interim. It provoked a comment about "what moron did that". One might reasonably have asked what moron allowed it in the kernel.
I also had an experience with Linux on a Cyrix board filling the log with messages to the effect that that particular processor didn't support speed throttling. Googling for the error message brought up a comment by Linus that the worst that could happen would that it would write that particular error message in the log. Yes it did. About once every second.