"Both of these companies can easily be wiped out by a company that's driven by programmers and organized to put programmers in the driver's seat, but which have an excellent abstraction that does all the hard work to convert code into products below the decks."
Notice the emphasis on the difference between the code and how it is packaged into a whole experience. The "product" is not just something which developers contribute to in Joel's world, but something which lots of people are involved in.
I try to tell people about the truly mad ideas, as I see them, in science. One of them occurs in electronics and nuclear chain reactions and biological evolution: it's the idea of "feedback". Here's the nuclear version: you've got all of these nuclei which are too heavy, they "want" to fall apart, they just need a little disturbance. (It helps to think of mousetraps: they're in a metastable state and they just need a good jostle to collapse to a new equilibrium, releasing a bunch of energy.) This disturbance usually has to come from a free neutron which smacks into the nucleus. Nuclei are small, and so you might think this is excessively unlikely, and that there would be complex probability arguments involved in understanding the procedure. Actually, it's much more simple.
To get a chain reaction, the neutrons formed by one fissioning nucleus need to cause other nuclei to fission. If each fission creates, on average, even a little more than one other fission, then you have positive feedback, and an exponential growth -- unbounded energy. If each fission creates on average less than one, then you will have exponential die-off of any reaction, with finite energy. (The "special case" of exactly one is what you need to simulate if you want to create a stable nuclear reactor; you need another feedback system sitting on top of your existing one which lowers the expected value if it gets too high, or raises it as it gets low.) The same principle viewed different ways is evolution by natural selection (the genes which characterize the population are precisely those which cause themselves to be more likely), or the psychology of neurosis (paranoia is defined by the fact that it causes those you love to fail you, thus feeding the paranoia). In the language of probability, if N is the number of events caused immediately by one event, you need E(N) > 1. That's the entire mathematics for positive feedback.
The reason that you want to focus on product is, I think, because you are struggling and struggling to get this value as high as possible. You need the expected value of converts, given one convert, to get greater than one. If you can do that, then traction builds automatically, as people who hear about it contact others to join further. In the long run, the people with the best product have the exponential advantage.
This is not to invalidate the problem-in-the-street of "how do I survive the times when E(N) < 1?". I am often reminded of the Chinese formulation of Buddhism, where you have an abstract goal (realize that all things are empty), but a concrete goal (there is a lot of meaning here and now), and you must form a synthesis of these two opposites. Your long-run abstract goal is certainly "chain reaction"; your short-run concrete goal is simply to stay afloat.