Learning at industrial scale is sometimes temporally cheaper because it allows you to conclusively exhaust a design path through efficient and spectacularly memorable failure. ;)

HN should have a 'comment of the day' section. I'd nominate this one.