Software engineering. There has been very little experimental work done here, simply because it's incredibly expensive to perform interventional research. But

Software engineering. There has been very little experimental work done here, simply because it's incredibly expensive to perform interventional research. But if we had solid research which said "software development method X produces software faster and with fewer bugs than software development method Y" it would be a great boost to startups, and YC is ideally positioned to perform such research given that it is surrounded by (and funds) lots of software development projects.