While I think this is true at the moment, it seems that there is an interesting question behind that, as well. Why don't more people want to create their own software? I do think the author has an implicit point: because, frankly, creating software stinks. Jon Skeet's talk  demonstrates this admirably. I spend a shockingly large part of my time working around bugs and leaky abstractions in other software rather than implementing my own ideas.
I don't know if it is possible to design a system with a solid enough abstraction that these problems don't exist. I do know that HyperCard came unusually close, as I know several folks who made HyperCard stacks who wouldn't imagine creating software in any typical fashion.
I hold out hope that such a useful system could be created again, one that gives people enough flexibility to create software solutions of their own (perhaps within a genre of software) and that is able to combat the currently-accurate stigma of programming being "hard".
Granted, that _may_ be a question of the tools. But I do find it telling that in the long time since the "death" of HyperCard nobody could come up with compelling tools.
And creating software is hard because ultimately, it requires analytical thinking. Which, by itself, is hard. Yes, the choice of tool modulates the hardness of the problem - but the underlying issues are still hard. (Note: I do not claim non-programmers _can't_ reason analytically. I claim the effort/result ratio is not right for them)
It's the same reason most people buy furniture instead of building it. Acquiring the necessary skill set is simply too much effort for the result.
I do think, though, that not being able to make projects of large-ish size isn't necessarily important for a wide variety of use cases. Most of the best apps I know of do only a very small thing—but do it very well. I think HyperCard allow people to do small things exactly the way they wanted. Which maybe wasn't very well, but was good enough for their needs and better than the alternatives.
As I mentioned in my other comment , I wonder if the parallel is somewhat like the UNIX command line. You probably aren't going to write a MySQL competitor in bash. But there are a whole class of small custom tools that you will write.
To perhaps restate my original point (and this is where perhaps we agree), the abstractions of HyperCard eventually didn't (couldn't?) adapt. Perhaps, as you said, because of cards. As someone else commented, also partially because of the codebase, but also, I think, because things like color and networking and a greater number of standard UI widgets all had to be added.