Many years ago, Francis Crick promoted (attributing it to his long-time collaborator Leslie Orgel) an aphorism that dominates the thinking of most biologists: "Evolution is cleverer than you are." This is often viewed as a more succinct version of Theodosius Dobzhansky's famous dictum: "Nothing in biology makes sense except in the context of evolution." But these two observations, at least in the terms in which they are usually interpreted, are not so synonymous as they first appear.
Most of the difference between them comes down to the concept of maladaptation. A maladaptive trait is one that persists in a population in spite of inflicting a negative influence on the ability of individuals to pass on their genes. Orgel's rule, extrapolated to its logical conclusion that evolution is pretty much infinitely clever, would seem to imply that this can never occur: evolution will always find a way to maximize the evolutionary fitness of a population. It may take time to respond to changed circumstances, yes, but it will not stabilize in an imperfect state. And yet, there are many examples where that is what seems to have occurred. In human health, arguably the most conspicuous case is that the capacity to regenerate wounded tissues is lost in adulthood (sometimes even earlier), even though more primitive vertebrates (and, to a lesser extent, even some other mammals) retain it throughout life.
This defiance of Orgel's rule is not, however, in conflict with Dobzhansky's. That's because of the phenomenon of pleiotropy, or trade-offs. Sometimes, the advantage gained by optimizing one aspect of fitness is outweighed by some downside that results from the same genetic machinery. The stable state to which the species thus gravitates is then a happy—but not perfectly happy—medium between the two extremes that would optimize the corresponding aspects.
Why is this so important to keep in mind? Many reasons, but in particular it's because when we get this wrong, we can end up making very bad evaluations of the most promising way to improve our health with new medicines. Today, the overwhelming majority of ill-health in the industrialized world consists of the diseases of late life, and we spend billions of dollars in the attempt to alleviate them—but our hit rate in developing even very modestly effective interventions has remained pitifully low for decades. Why? It's largely because the diseases of old age, being by definition slowly-progressing chronic conditions, are already being fought by the body to the best of its (evolved) ability throughout life, so that any simplistic attempt to augment those pre-existing defenses is awfully likely to do more harm than good. The example I gave above, of declining regenerative capacity, is a fine example: the body needs to trade better regeneration against preventing cancer, so we will gain nothing by an intervention that merely pushes that trade-off away from its evolved optimum.
Why should it be, though, that evolution accepts these trade-offs? In reality, it doesn't: it is always looking for ways to get closer to the best of both worlds. But that, too, must be considered in the context of how evolution actually works. Some adaptations, even if they may theoretically be possible, just take too long for evolution to find, so what we see is the best that evolution could manage in the time it had. It looks like stability—as if evolution has decided that a particular trade-off is good enough—but that's really just an approximation.
So in summary: follow Orgel when you're coming up with new ideas, but follow Dobzhansky when you're engaging in the essential rigorous evaluation of those ideas.