Obviously written by someone that has never done any "physical" engineering.<p>It doesn't matter what you build, when it is at scale the sunk costs can be enourmus to realise half way thru it isn't going to meet your needs, or take 9 years longer than expected.<p>As Fred said about software, but it applies to everything, nothing is harder than deciding precisely what to build.<p>(First identify the problem, then work out your requirements to solve that problem, then go from there - most people skip at least the first step if not most of the second and then, there's your problems, esp with software).