I think there's another implicit point as well, echoed by a lot of more recent startups, which is to fail early and often. If you're going to make the engineering decision to go down a path which will take a year+ to validate (i.e. developing a "PhD thesis" of a filesystem), then you'd better have a very compelling reason to do so over the current solution.
Comments
Most valuable point in here, vital for startups: You can’t sell someone the solution before they’ve bought the problem
I think there's another implicit point as well, echoed by a lot of more recent startups, which is to fail early and often. If you're going to make the engineering decision to go down a path which will take a year+ to validate (i.e. developing a "PhD thesis" of a filesystem), then you'd better have a very compelling reason to do so over the current solution.