It seems like the field has become stagnant in the decades following the enormously successful and influential Black Scholes model. (The original paper has been cited a staggering 25,000 times - more than ANY economics paper, by far.) There is also the CAPM and GARCH models, but those were decades ago. Now we have Black Scholes type models for every type of option under every conceivable condition. We have multiple derivations of these formulas (forier series, closed form, etc) . So what is next? The only new idea is http://en.wikipedia.org/wiki/User:Stockequation/sandbox which makes the jump from the stock market being a statistical system to a mechanical/physical one, like general relativity. It's essentially a 1-d ADS/CFT applied to the stock market and it generates fat tail option prices and vol. smile
These years, The new frontier has been around optimal trading, market impact and orderbook dynamics. They are plenty of sources around, one of them being this bi-yearly conference. You have all the slides on the web site: Market Microstructure: Confronting many Viewpoints.
The next challenge is to link previous quant and economic knowledge with micristructure.
It's a free peer reviewed journal, depending on your definition of discovery you might find interesting tidbits there.