Chris Phoenix is writing an interesting series of articles over at CRN about the dynamics of the development of molecular engineering. His thesis, as far as I understand it after the first three articles, is that we’re likely to see a fast takeoff, because it’s easy to achieve excellent results after you’ve achieved good enough. One example is error rate - going past an adequate error rate to a superlative one just requires additional purification / error correcting steps. Other things that may improve very quickly once a workable solution is found are reaction rates - which are exponentially dependent on positional accuracy / stiffness.