First and foremost, I need to fix the garbage collector. It's currently stupidly slow, so much so that I turned it off entirely in order to run the self-hosting tests. Fixing this won't be terribly hard, but it may be difficult to hit the kind of speeds that I really want out of it without doing a lot of heavy lifting. In particular, the GC is currently backed by operator new() in the C++ implementation of the runtime, meaning that I don't get fast allocation, and I can't do pointer relocation. Pointer relocation is a major challenge and might be out of scope of what I want to do with the GC in the short term, so I'll probably settle for just optimizing things as much as I can.
Once the GC is in better shape, the compiler itself needs a lot of love. The code is disgustingly unorganized, and the basic algorithms and data structures are utter rubbish (although it is kind of fun to say I wrote an entire compiler using nothing but linked lists). It currently takes about 60 seconds to compile itself, which is just... no. The C++ implementation of the compiler, for comparison, can compile the Epoch compiler in about 600 milliseconds. So I need a hundredfold speedup in the compiler, which should be doable considering I've done 1000x in the past.
A tangential compiler feature that I desperately need is separate compilation, i.e. code in multiple files. The compiler will be a lot cleaner with that support in place, and it'll enable a lot of other stuff I want to do.
After the compiler is at a point where it hauls ass, it'll be time to start working on the IDE. I have some concepts for Era that I'd really like to prototype, but they require a ton of iteration so having a fast compiler will be important. I also want to start working on a generic UI framework so that I'm not writing a bunch of Win32 message processing garbage all over the place.
And all that should keep me busy for quite a while. Somewhere in there I'll package up Release 15, but I'm not sold on shipping a release just yet. I'd really rather improve some of the core language features before shipping another release, and that will probably wait until during the IDE work if not after.
So there you have it. As always, I will probably change my mind and wander around idea-land after about a week, so don't take this as a hard commitment :-P