If you hold a position that affects only your own actions and your opinions of them, I don't see much reason for calling it an ethical system. I read the blog post; that second section defends acting on non-universalizable principles, but doesn't so far as I can see defend thinking of them as moral principles and that's what I'm casting doubt on.

For clarity: I am also a moral nonrealist, and my doubts about the moral-theory-ness of fitnessism aren't because it doesn't involve a claim that its values are Objectively Right And Good. Rather: I think a moral theory is something that guides moral judgements by an adherent, and one feature of moral judgements is that they are applied to other people as well as to oneself. Something that affects only its bearer's own behaviour I would call a "preference" or a "motivation" or something of the kind, even if it gets expressed using the word "should".

I agree that if the principle is "maximize long-term number of copies of genes that influence my behaviour" then the counterintuitive consequences I described don't clearly follow. (I'm not sure they don't follow, though. The answer may depend on exactly what you're prepared to count as a "behavioural gene".)

It's true that most of my genes are shared with other human beings, even ones I wouldn't normally think of as related to me. But it's also true that a lot of my genes are shared with, say, pomegranates. Your restriction to "behavioural" genes doesn't (I think) make that problem go away; only in popular magazine articles are there genes for behaviours in any very strong sense; how sure are you that there are no genes you share with (some or all) pomegranates that have an effect on your behaviour? If it turns out that there are some, would you start regarding it as an important obligation to increase the number of pomegranates (at a rate, perhaps, of 1000 pomegranates per human life)?

I suspect that if we pay attention (as you do) to the very long term, it actually matters rather little in practice what we care about there -- in particular, it's likely that much the same actions now maximize long-term human happiness, long-term human numbers, long-term number of books-or-equivalent written, etc. (A similar thing happens in computer game-playing: the further ahead you look, the less the details of your evaluation function matter.) So it may be hard to distinguish between fitnessism and almost anything else, in terms of the actual decisions it provokes...