Somehow that essay reminded me straight to the Guix documentation that I was working with earlier this day. So well organized and to the point. Pure joy.
And he is right in what he says that documentators should not be passive but take part in shaping their Gegenstand. Some of my best pieces of work, I wrote the docs in advance to functionality, or in parallel.
And then there is literate programming. This is the way we should write important, long-running software. I used this the first time when I took a new job and my new bosses just had managed to micro-manage their MATLAB and math expert out of the company. A few pages of critical code with probability calculus which nobody did understand. Fun stuff.
Just now I am looking at someone's code from the past twenty years. All that time working on the same stuff. Great guy. No docs. So sad.