I agree growing things is the way to go, and it's worked for me. Mistakes in some areas are harder to fix, such as in public APIs that others rely on - but they

I agree growing things is the way to go, and it's worked for me. Mistakes in some areas are harder to fix, such as in public APIs that others rely on - but they still can be fixed by deprecating older approaches. It's painful; but perhaps even worse to keep delaying til it's perfect... A hard one is patents: you can't patent after launching, selling or even demoing it publicly, so you have to finalize the design well-enough to describe it in detail, without the benefit of experience with actual users. You can file a provisional application beforehand, but it has to be as complete as a full application, because you can't add materially to it. I Believe that in the USA, you can apply up to a year after inventing it, but you'll still need sufficiently good documentation to be able to prove that the invention date, if challenged. A meticulously kept notebook is probably sufficient.