Adaptation is a variation within a species. The extrapolation is that this variation will also produce new species with completely different genetic information given enough time.
The extrapolation relies on inference from indirect evidence.
The extrapolation is challenged by the up and coming mathematics of information theory. One of the fascinating tenets of that theory is that information is independent of the carrier of that information. In other words, DNA, in and of itself is nothing magical or special; it's simply the carrier of the information.
You can do some interesting experiments based on this. By treating evolution in terms of information, you can create computer simulations that can test to see if information can be added to an information matrix through random copying errors. This is the fundamental test of evolution (because a mutation doesn't count unless it is passed on to the offspring). If you cannot add information, then the creation of species with entirely new morphologies and biochemistries is impossible, and the extrapolation that variation within an information matrix equals creation of new information matrices (the creation of new species by an extrapolation of variation within a species) is proven wrong.