It's something of a truism that most of the great software that exists does so because the programmer had an itch he wanted to scratch. (This probably explains why a good programmers' editor is so much more satisfying to use than a word processor, the people who use the programmers' editor are the people who program it. If there's something that bugs them about the app, they fix it and move on.)

It seems to me that, if you really want to get better at your craft you should find something that you need and implement it. Assuming you have the time, implement it yourself. Then go out and see if someone else has done it. If they have, compare your wheel with theirs and work out why theirs is better (or worse) than the one you made.

Then find another itch and do it again. Rinse, lather and repeat. If you don't get better at your craft in this process I'll be amazed.

Of course, reading around the subject couldn't hurt either. Grab one of the classic computer science/engineering texts and read that. Read up on (and play with) a few languages that do things differently from what you're used to. See how those ideas map to perl (and vice versa).

Because you'll always be working on stuff that is useful to you, or that you find interesting, you'll be able to track your improvements. If you're working on something because it's popular, or because it seemed like a good idea at the time, chances are you won't really engage with what you're doing and you won't really gain much from the experience.