But the question is where do the differences lie. My hypothesis for one difference is that we dont write code like painters paint.
That is bad. We should. Hence my advocacy for literate programming.
It isn't to be dismissive of painting, it is to say we put ourselves in a more precarious position by how little we support ourselves with our tools.
Painters (, mathematicians, etc.) are all working with pencil-and-paper in quite a radically different way.
I think this goes some way to saying, actually, there is a bit of a difference here.
Doctors and nurses, who also deal with complex systems, have gotten exceedingly good at this with charting.
We? Considerably less so. But I'd hazard improving every year.
Reduction of boilerplate and flexibility-for-the-sake-of-flexibility in languages has been a big win, IMHO.
From c. 1m30s to 5m30s
What is rarer is the moment of effective effort's nature, for many types of coding (to agree with your point).
The author's house of cards analogy trivializes it and makes it easier to dismiss as a rhetorical device. I'd use activation energy as a more apt comparison.
IOW, if you put in 30 minutes (or 3 days) understanding code, and do not make the correct 5 minute addition / change at the end of that, all that previous time is wasted. And moreover, that knowledge has some half life in my head, after which it must be reacquired.
And that correct addition / change cannot be made without first paying that cost.
This is not unique to coding, but is generally limited to professions involving the modification of dynamic complex systems. E.g. legal, biological, mathematical.
And I'd hazard to say of those, programming has a wider universe of possible approaches (aka "Why did the last person? Oh... Oh. Oh no."). Legal is probably the closest analogy.