Oh, I don't disagree with that part at all, I just think that your aside about it encouraging just the right amount of design upfront deserves more thought. I

Oh, I don't disagree with that part at all, I just think that your aside about it encouraging just the right amount of design upfront deserves more thought. I tend to do design brainstorming on scratch paper, Emacs scratch buffers, and in Prolog, but I'm already convinced about the merits of having a good README, and would write one regardless. (Just like I write tests anyway, whether or not I do them upfront.)