At this stage all you need to care about is to keep stretching your brain in many different directions. There still will be a lot of time for you to learn all the amazing theories, useful tools or amusing curiosities. Let your imagination run free and enjoy the math you do (but be serious, don't slack off).
If I were to give you one advice on what to do: experiment!
Each mathematician has his/her own best set of things to do and these differ considerably. I have no idea what will work best for you, I can only suspect and even that is warped by my own subjective perspective. Of course, I have some observations about what worked and what did not for my friends or students I taught, on this observations and various musings of others I based the advice that follows. However, you won't really know what works until you try it - each to his own.
If you had to choose between knowledge and thinking, at this stage I urge you to pick thinking.
Both are important, but there is only a finite amount of time you can spend on them and
- you will pick up some knowledge along the way and that will be enough to prepare yourself for more knowledge in the future,
- you will be able to learn new things for many, many, many years,
- there are some quite good substitutes for knowledge (e.g. books, internet),
- I know no conclusive research on this topic, but to me it seems that thinking capabilities work similarly to other human traits, that is, the early years determine the limits you can later achieve (the cap on your abilities).
It doesn't really matter that much what kind of math do you do, as long as you keep stretching your brain in different directions. Also, whatever you do, be serious, don't slack off.
More personal advice
Personally, I would advise you against learning advanced material and study university topics ahead of time (more on this below). Instead, practice a lot of Math Olympiad problems, but optimize for brain-stretching, not for contests (more on this also below).
Another thing that would be great to do is to experiment with math. Take some math and use it, or find some practical (getting purely abstract too soon won't do you any good) problem (it might small and insignificant, doesn't matter) you would like to solve and try to do it with math (perhaps learn some new things in the process).
I will not only give you an example, but three tasks that will be great for your thinking (they will take some time, but not that much):
- Make a simulation of a lunar lander (the graphics can be trash) according to physics equations, including (at least) mass, gravity, atmosphere density (to have some damping effects) and thrust forces from engines. Represent them as differential equations and solve numerically (use something simple like Verlet).
- Transform an audio voice recording (e.g. some MP3) using your own FFT algorithm (it can be DFT, doesn't matter), e.g. to be a bit higher. (To convert the numeric data to and from audio files use some tool or library.)
- Code a minimal raytracer. Try to use homogeneous coordinates inside and Monte Carlo methods during rendering. Also, play with transformation matrices.
If these sound intimidating, don't worry, they are much easier than they seem and you can find a lot of help online (tutorials, forums, stack-exchange sites, etc.). Besides, you don't need to understand all the details of the underlying theories, to complete these tasks you only need to learn a few bits for each (e.g. derivatives and integrals, complex numbers and polynomial evaluation, matrix multiplication and some randomization). In fact, a few paragraphs below I will argue that you should not learn too much advanced material too early.
The above tasks require some programming skills, but I guarantee that coding ability will be useful to you (see also here). Each has an underlying topic (differential equations, FFT, linear algebra) which is highly relevant, frequently used in applied math and amazing by itself. You will have a better understanding of what math can do, and believe me, doing it personally is a whole other experience than just hearing about it. Finally, it is great fun seeing it run!
(If you are more focused on abstract math, these problems will still be helpful.)
On rediscovering the wheel
Whatever you will ultimately do, at your age, don't be afraid of rediscovering the wheel. Although in research it's better to rely on known theorems rather than prove everything from scratch (human lifespan is limited), there is no better way of learning how to invent new things (I would recommend against true open-ended research, because it's easy to get frustrated).
Basically all standard textbook problems rediscover the wheel (they were solved before, weren't they?), but somehow we understand that this is precisely their intention. However, ordinary theorems or folklore facts can also be though of as challenges. This is standard in geometry, so why not in other areas? For example, could you solve the quadratic equation without knowing/using the exact formula for roots?
On university curriculum
I have nothing but anecdotal data to support this claim, but I think that studying advanced books too early may not be such a good idea. You can find a more detailed description of my worries here: Effects of early study of advanced books, but to briefly summarize main points:
- such approach favors knowledge over thinking,
- when you study on your own, you might miss some important pieces,
- you train yourself in following the beaten track (rather than creating your own).
On Math Olympiads
Math Olympiad problems are one of the best mind stretchers I know. If you approach them right, a lot of them (depends on country) become great tryouts for your out-of-the-box thinking.
The most important thing here is not to optimize to get far in the contests, but to expand your mind. For example, there is a certain set of mathematical theorems which with enough practice allow you to solve many olympiad problems with little or no insight (e.g. the Muirhead's inequality is notorious for killing fun). So when you spot a problem that is easily solvable by some "overpowered" theorem, either try to find another solution or go on to next problem. Similarly, do not measure your time when solving Math Olympiad problems. If you are stuck, try harder (rather than moving on or reading solution). If you can't stand the problem anymore, leave it and perhaps try again next week.
Don't just learn math, or as Paul R. Halmos have said: Don't just read it; fight it! (the picture comes from the Abstruse Goose webcomic).
I hope this helps $\ddot\smile$