I didn't like some of the 7GUIs that weren't real-world enough

I like the idea of a set of benchmark applications for testing PL coverage. Rather than learning the features of other PLs and choosing which features to copy, I prefer a task-driven approach -- i.e., I want to make sure that any tasks that can be performed using other PLs can be performed using my PL, where I define a "task" as something minimal that a program does, such as adding two numbers together or writing to a file, as opposed to how a given PL does something. I had hoped to find an existing set of such tasks and code them in my PL, and if I succeeded I would know that I would have achieved my coverage goal. Unfortunately, no such luck. I considered slogging through the Rosetta Code tasks, but there appeared to me to be an enormous amount of task overlap there and no guarantee of the full task coverage I'm looking for. In the meantime, I have decided to code common applications like 'Hello, World!," a Guess-the-number game, Tic-Tac-Toe, Flappy Bird, TodoMVC, etc., and in different areas such as games, business applications, Internet applications, etc., until I find that I am a) using my PL elements in a consistent manner, and b) no longer adding new PL elements.

So I'm all for what you and @jonathoda are trying to do

For an advanced task you could consider games that would traditionally be considered more complex. Tower defense? How do you express a spatial query between two moving objects (bullet and enemy)? What would a path-finding algorithm look like? In most languages this requires complex data structures and algorithms. You need to be able to easily express "popular" algorithms such as these, i.e. that you want the "shortest path" between two things, or design your language & environment such that highly reusable pre-made solutions are readily available via a search bar (how does the user figure out what to search for?). Or if you think non-trivial games are out of the scope of end-user programming, then where did you draw the line on what end-users are "supposed" to build? Once they've built their timekeeping and social apps, what's next? As a kid I always wanted to make video games (and many kids do try). I've got several friends who design intricate pen-and-paper worlds on a regular basis (DnD) but don't have the ability to bring these worlds alive in a computer. Games are possibly the highest-demand end-user programming domain that exists.

And I like Nick's idea about choosing apps that an end-user might reasonably want to create.

And I'd add, apps whose specs are minimal and therefore quickly grasped, like Flappy Bird.