I agree that the context switching costs are quite high, reducing the time actually spent on coding. It might be better to focus on one language/framework for five straight weeks, for example, while still changing specific design patterns/recipes per week. It reduces switching costs, gives devs time to grok new languages if need be, and should result in higher quality code for each weekly project.

You might also let devs post fragments of their code as they work on it, and links to their github repos. The reason is that a dev who is really learning a lot, not just exercising existing skills, might get far off track if working in isolation, like a plane that takes off at a heading that is slightly off, and ends up hundreds of kilometers from its target. Plus, it would give more time for people to discuss/debate alternate implementations, and that debate itself would also be a great source of learning.

Even though I signed up and am going to try my best to do something with this, I do agree. If you're targeting different languages / frameworks / etc, it seems like even something simple like doing a code kata in a brand new language is a decent accomplishment for a week.

I think that in itself is actually a good learning activity as well.