For me, it's always writing an actual application...no toy or example programs. I suggest rewriting something you've implemented before, rather than new. The

For me, it's always writing an actual application...no toy or example programs. I suggest rewriting something you've implemented before, rather than new. The reason for this is that you've already solved the problem, you're just left to language semantics. Trying to tackle a new application is fine, too, but may cloud your judgment on the language if it becomes a hard problem to solve, or just plain boring.