I believe Ravyne has it spot on.
I've seen so many people getting into the 'limbo' of creating the perfect (3d) engine, without actually finishing anything (no offense though).
Somewhere there's a dot on the horizon in the future, that's were you want to go.
Take that dot, go back step by step till where you are today. And simply take the next step :)
Whether that be creating your own engine for learning (if so, make a scope/ feature set, not limitless), or ship a completed game, or learn engine features a to j etc.