I might be behind the times, but the thing that was by far the most interesting to me was Servo [0]. He's absolutely right - all of the current rendering/JavaScript engines are going to need significant work to take advantage of massively-parallel hardware in the next few years. (I say this after seeing the work required to port ordinary CPU code to CUDA. OpenCL should be similar.) The idea of Mozilla not just switching to WebKit but actually trying to leapfrog it to a new, even better rendering engine is awesome.
[0] https://github.com/mozilla/servo