One of the longstanding problems with fitness landscapes is that they are mostly abstract and arbitrary constructs used for conceptualisation rather than actual explanation. Things have changed. Now a paper in Nature shows that fitness landscapes empirically measured show accessible routes of molecular evolution from one function to another. In particular it's nice to see a comment like this:
The tentative picture emerging from the new results is one that emphasizes the possibilities of continuous optimization by positive selection. Although evolution was clearly constrained, as illustrated by many inaccessible evolutionary paths, the studies also revealed alternative accessible routes: a succession of viable intermediates exhibiting incremental performance increases. Although these findings do not address whether natural evolution proceeds in the presence or absence of selection, they do show that neutral genetic drift is not essential in the cases studied. We note that the presented approach starts with naturally occurring sequences, which are themselves the product of evolution, and may therefore yield a biased sample of trajectories. Whether the conclusions are general or not, and whether they break down when the evolved feature becomes more complex, can only be determined through future studies.
Rather than try to overgeneralise, the authors, Poelwijk et al., say that we need to do more empirical work. But in this case, at least, there are viable intermediate fitness enhancing pathways, something previously shown by Kaufmann and Gavrilets theoretically.