David Fenton wrote yesterday with a puzzle:
This post, which is just a link to a video, has as its comment this phrase:
 Is there nothing the iPhone can’t do?
Reading this caused a mental stumble — seems like double negative problem, and more naturally expressed as:
 Is there anything the iPhone can’t do?
But then I thought about it, and it seems to me that the two phrases mean exactly the same thing. Normally nothing/anything would be something of a loose antonym pair, but in this case they are interchangeable.
I replied (again, slightly edited):
Seems to me that  is naturally read as a rhetorical question, conveying
There is nothing the IPhone can’t do. [3a]
There isn’t anything the iPhone can’t do. [3b]
while  can be read as an information question, asking about the abilities of the iPhone. But, in the right context, information questions can be deployed rhetorically, so  and [2 ] can end up having very similar effects.
I consulted negation maven Larry Horn, who agreed with my take on the question, and expanded on it:
I don’t see the two questions as being at all the same in terms of what kind of background they presuppose and what kind of answer (if any) they anticipate.
 is most naturally uttered as a surprising empirical conclusion to a repeated demonstration of the iPhone’s versatility and capability, and is indeed (as are many negative questions) normally intended rhetorically, roughly equivalent to a (hedged) assertion like “(So) it seems there’s nothing the iPhone can’t do”, i.e.”The iPhone can evidently do anything, right?”
, on the other hand, is naturally asked when one is uncertain about the limitations of the iPhone, and might be asked of a salesperson in the Apple Store as a legitimate query: “You’re asking me to pay all this money for a device–what are its limits? What can’t it do?” But  can also be used rhetorically in a way quite similar to , building in an inference that apparently there isn’t anything it can’t do.
So  is a bit more versatile than .
Note that Larry’s and my responses go well beyond simple-minded truth-functional semantics (not that you couldn’t try to devise a complex-minded truth-functional semantics, though that takes a considerable amount of work). In particular, both of us replied with references to concepts that are ordinarily understood as pragmatic rather than semantic: see the references to questions understood rhetorically and Larry’s more general reference to “what kind of background [the questions] presuppose and what kind of answer (if any) they anticipate”.
So  and  are subtly different pragmatically.