I'm beginning to wonder if a lot of commenters have forgotten that waterfall is a failed methodology in part because you can't always just gather requirements by say introspection or interrogating the customer for them. In order to reveal the true, accurate requirements, you often need to write code. Not always, sure, just like some code is relatively easy to write, some requirements are easy to both accurately specify ahead of time and even satisfy. (An old XKCD hinting at this distinction: https://xkcd.com/1425/ Only nowadays we actually can easily satisfy a lot more of those old easy-to-specify-difficult-if-not-impossible-to-satisfy requirements. That's the way progress in the field often goes.) Decades of startups support the iterative approach, along with pivoting, the successful ones learning the lessons of writing less committal code and discovering requirements iteratively. And it's hard to actually pull that off -- so many companies have gone under or lost out significantly to competitors because they couldn't write code fast enough, or change code fast enough, either with the goal of discovering the real requirements or adequately satisfying those requirements once known. Code is hard in general.