There is no selection for reproductive fitness until there is a replicator. Once a random process has stumbled upon a replicator, selection for reproductive fitness is inevitable, so long as some of the replicator's descendants survive. Reproductive fitness, therefore, is a feature of sufficiently large, complex and long-running random processes, not something additional that is imposed from outside. This is what Dennett called "Darwin's dangerous idea".
No. The "dangerous idea" is that design might not require a designer.
As for the first replicator, we simply do not yet know how (or even when) that came about, though random chance does seem likely. So I guess if you add enough layers of "built the machine that..." then you do end up with pure randomness (probably). But I would conjecture that there's a lot less "informational distance" between the cosmos and the first replicator than there is between the first replicator and Adam Yedidia.
It might not be what Dennett meant, but I don't think issues of informational distance refute the claim that we can expect selection for reproductive fitness to be a feature in any random process that gives rise to replicators. After the appearance of the replicator, the random process continues to operate under the same rules (laws of physics) as before.
After the appearance (and sufficent success) of the replicator, it stops being a random process. Edit: or at least not a completely/primarily random process.
