Again, I'd argue that focusing on being proficient at the level someone is interested in is much, much more useful than focusing on making a lot of different toys at a bunch of different levels and in a bunch of different languages. You are going to learn something either way, but jumping around with a bunch of small projects all over the place probably means that you aren't going to retain much and you're not going to have any particular strength at the end.
Conversely, if you focus you're going to retain a lot more, since you can incorporate a lot more of your previous knowledge as you get more knowledge - in fact, if you work on a real project instead of a toy, you're probably going to be seeking out a lot more knowledge on your own as issues come up. In the end you're going to be able to say your proficient in your chosen area, and you'll probably be in a good place to expand to other areas (and you'll have a good sense of where you should expand to).
Have you ever written an assembler?
Have you ever written a compiler?
An Assembler is far simpler then a compiler and after building the Assembler a compiler is an extremely strait forward next step as you've already finished the first quarter of what you need to do in a compiler. You've tokenized (broken the problem into small bits) and you've generated some opcodes. That's the perfect starting place for any systems programming background and a systems programming background is the perfect place to start out at higher level languages.
In a sense most games are a compiler. You feed it in source (images, dialog, models, textures, etc) and it spits out an interactivly compiled narrative.
Websites are also compilers. They just operate on a higher level of abstraction then the simple toy compiler we would right.
The information to someone new is invaluable so long as they learn it themselves and aren't forced through it.