EDIT: New reward algorithm is nSubsidy = log2(difficulty)

This method of money creation to achieve a stable monetary policy has been suggested before but has been shot down quickly on the assumption that it would create hyperinflation. I agree that, because of technological progress, increasing difficulty by X amount will be cheaper in the future than it is now, but I propose that the relative effort of increasing difficulty by X% should have little variance over time. In other words, going from 1,000,000 difficulty to 2,000,000 at some point in the future will be as cheap as going from 1,000 to 2,000 right now. Thus the block reward could be a function of the difficulty growth rate to maintain stability.