Good post, but in my experience full mockups and pseudo-code is still not close to 90% of the final product unless I'm doing something that is trivial for me. There is just too much complexity that emerges in places which are difficult-to-impossible to anticipate until I've dug into code and physically tested the ideas I've laid out in pseudo-code. I'd probably put the number more in the range of 25-50% depending on the scope of the project.

Of course for simpler projects or in domains where the developer is extremely experienced and can anticipate every hiccup that might occur, that number could go much higher, but I don't think this is usually the case in the wild.

I definitely agree that in terms of effort expended implementation is almost always >=50%, but effort can be of various types. Sometimes when I've implemented ideas, the implementation took 90% of the time, but it was basically straightforward code-slinging, so I wouldn't credit it with more than 10% or so of the total intellectual effort expended on the project. Other times, important stuff does come up when implementing that sheds new light on the idea (or shows that it wasn't as fully developed as I initially thought).

When I'm implementing someone else's paper, a heuristic I use is something like: did I "just" implement this paper, or did I learn important things while implementing it that aren't actually mentioned in the paper, and which I should probably write up somewhere for other people's benefit? Sometimes the answer is "just implemented" even if the implementation took a long time and was hairy.

As you say, it depends on the problem. Working in mathematics, if I have the algorithm, turning it into code is almost always very easy, but getting that algorithm can be very hard indeed.

Even if we put the number somewhere between 25-50% depending on the project, that is still a long way away from just having an idea. That still constitutes substantial progress and a solid start on execution.