Breed yeasts to produce hexanol, rather than ethanol
In conventional fermentation, yeasts turn sugars into ethanol. Ethanol, as everyone knows, is promising as an alternative fuel. The problem is that yeasts die at concentrations higher than about 10% ethanol by weight, and so the fermentation process can at best produce alcohol that is 90% water. Obviously, this "beer" cannot be burned as fuel, and the excess water must be removed somehow, by distillation or adsorbtion, which adds a significant energy cost to each unit alcohol produced. At the earth's equator, solar energy can be relied upon to make up this energy cost. At the more extreme latitudes, that's not necessarily the case.
Ethanol is not the only alcohol produced in fermentation. Higher alcohols such as butyl, amyl, isoamyl, and 1-hexyl are also produced, albeit in trace concentrations. As a fuel alcohol, 1-hexanol has a lot going for it compared to ethanol. Firstly, it's much "greasier" than ethanol, having a thrice-longer hydrocarbon tail, and thus will handle and burn much more like the hydrocarbon fuels we're already using. Second and most importantly, however, unlike ethanol, 1-hexanol is *not* infinitely soluble in water, meaning that at some concentration the fuel and the water will simply phase-separate. Now, instead of having to spend energy to dry the alcohol, you just tap it straight out of the bioreactor at burnable concentrations.
The only reason we're not doing this already is that (known) yeasts don't produce useful concentrations of 1-hexanol. But because they're microorganisms and they reproduce rapidly and in huge numbers it's not inconceivable that they could be bred to do so. What's needed is a rapid, colorimetric, quantitative assay for hexanol concentration so that thousands of individual yeast cultures can be rapidly screened in high-throughput equipment like plate readers. Without such an assay, chromatography of some sort is required, slowing the process of screening down by many orders of magnitude. With the right indicator, though, it would be possible to screen yeast cultures almost as fast as they could be selected and grown. A rate of 10000 generations per year is entirely reasonable. Note that 10000 generations is approximately the same "distance" that separates homo sapiens from neanderthals.