It started with scratching my own itch, and I've done a lot of customer validation since then (not as much as I'd like, but enough). So in this case, the MVP is about limiting scope, not doing things that don't scale. Doing it manually is not "perfectly fine", because the problem is that it's already being done manually, and needs to be automated. "Manual" and "viable" are incompatible in my case.

Which gets to the author's point.