But genomes aren't like that. The actual number of functional genes in the human genome is not definitively known, but the best estimates are that it consists of around 30,000 genes, some consisting of a few dozen base pairs, and some consisting of thousands and thousands of base pairs, with the rest falling somewhere in that range. But the full human genome consists of about 4 billion base pairs. So it turns out that, IIRC, around 95% of the human genome is not functional - it exists in the form of introns. But these introns are more than just random genetic noise - they are bits and pieces of old genes from precursor organisms that have been retained, which isn't a surprise, I think. Nature is parsimonious, and there's no penalty for retaining them, so why not? And the way that functional genes (extrons) are differentiated from introns is that they have special genetic markers at the beginning and end - "START" and "STOP" codes, essentially.
So there's a critical difference right off the bat - while a bit flip in software is almost uniformly disastrous, it turns out that changes can accumulate within introns with no effect at all, positive or negative. Since the vast majority of the genome is non-functional, changes in those non-functional portions don't harm or help the organism. And mutations can accumulate within those introns without harming the host organism.
So, if you can picture mutations occurring within those introns, and gradually accumulating, then you can have a fully functional gene spring forth after a while by simply mutating the start and stop codes - what was junk becomes a (hopefully) working gene if the markers change to indicate it as a working gene.
And it turns out that this is mostly how new traits are added. Every once in a while, a mutation to an existing, functional gene will result in some improvement, but like software, such changes turn out to be disastrous far more often than not. But changes can accumulate within the introns far more readily, because there's no penalty for mutations within introns. And then the right mutation flips the switch, so to speak, and a new gene (or a non-functioning old gene, of course) is turned on. Of course, it's also likely that this new gene could be dysfunctional itself, and hence will not be an advantage. In which case, if it proves to be a serious disability to the organism, survival becomes less likely in competitive environments, and the new gene is selected against. But every once in a while, such a new gene, or recycled old gene, will prove to be advantageous, and that trait will be selected for, and propagated throughout descendant organisms.
So, that's the first key difference that springs to mind - software doesn't have an analogous genetic "junkyard" from which to draw new parts, and recycle old ones. Virtually all hits to software are fatal, whereas the majority of mutations to genomes make no difference at all, because they land in unused parts anyway.
Let's pick up human intelligence, which somehow got into your/our posts. So far as we can tell, there is not a photomorph increase in human intelligence--our ancestors of 200,000 years ago are thought to have the same brains as us, and therefore the same raw intelligence as us. What the heck did they need that potential brainpower for? How was that intelligence "evolved" by the exigencies of the environment? Put another way, why did 1%, or .01%, of cavemen have the potential for understanding number theory?
Go back further. "Cro magnon" man is thought to have arisen about 120,000 years ago, and is anatomically identical to modern man - it is modern man, essentially. And if you had a time machine, and brought back a representative from back then, and you cleaned him up and gave him a shave and haircut, probably the only thing that might be unusual about him would be his height.
But, if you go back further, to about 1.6 million years ago, you find the first remains of Homo erectus (the so-called "Java man"). And H. erectus had a noticeably smaller cranial capacity. The current average cranial capacity is about 1350 cc worldwide (with a wide range of capacities considered "normal"), as compared to about 1150 cc on average for early Cro magnon specimens, and about 900 cc for early specimens of H. erectus, and around 1150 cc for later specimens.
So now, take one more step back, to the first known specimen of the genus Homo, and you find Homo habilis. Dated to approximately 2 million years ago, H. habilis had an average cranial capacity of around 750 cc.
So what's the point? Well, there's a well-established lineage, in terms of brain size, and thus intellectual capacity. But there's no reason to think that early hominids had the capacity to understand particle physics or number theory. Indeed, the earliest Australopithecines (A. afarensis and A. africanus) had cranial capacities in the 400-500 cc range, only slightly larger than a modern chimpanzee, at around 390 cc.
But the ultimate question is, why intelligence at all? The thing is, though, asking that question is like asking "why wings?" or "why lungs?" There's no metaphysical "why" that science can satisfactorily answer, but the mundane answer is that basic reasoning capacities must have conferred an advantage to creatures that possessed them, versus creatures that didn't. And even the most limited of problem-solving skills is still better than nothing, I think. So as small improvements in intelligence propagated and were selected for, small improvements to the small improvements would also have proven advantageous. And so forth, and so on. If you look at the very earliest hominid specimens, and then compare them to modern humans, it looks like a quantum leap. But when you see the progression, it looks much less miraculous. This is the hominid answer to the creationists argument of "what good is a partially formed trait?" Well, the "half a brain" of H. habilis is better than the comparative nothing of A. afarensis, even though it's not as "good" as what we have today.
So that's a long-winded answer to both hominid intelligence and transitional forms, with the single added point that every organism, unless it goes extinct, is a transitional form. It doesn't make any sense to think of evolution as having an "end", so thinking of what's out there today as the "end product" is misleading, and thinking of everything that came before as "transitional" is equally misleading. Any step in an evolutionary process is an illustration of a transition - no matter what you are right now, natural selection is operating right now to transform your species into something else in the distant future. We are transitional forms, put simply.
So, without too much work, it ought not to be too difficult to imagine a process similar to hominid brain development also operating in the transition from land mammals to whales. And somewhere on this thread (or other threads, I know for sure - if you can't locate it, you can ping him back and I'm sure he'll be happy to point you to it), VadeRetro has posted what evidence there is for whale evolution, and the process and steps present therein.