This reminds me of the Microsoft research talk (link pending) about the advice he gave to his researcher students about not waiting until having ideas clear to write them, but write ideas to make them clear.

In my case, documentation has been a pretty active part of Grafoscopio and Brea, but the closer prose, code and workshops are, the clearer I'm about how and what to teach (and the code and prose to write and to replace/erase).