You're right that they don't solve any open problem in computer science or mathematics, but neither did Facebook, Amazon, Apple, Google (I think), etc., and they seem to have done ok. :)
Not to take away from your main point, but Google did need to solve a fundamental computer science problem: how to compute eigenvector centrality efficiently. It would have been impossible to calculate PageRank at web scale without the Power Method (and applying EC to the link structure of the web in the first place involves lots of nontrivial math).
What makes selling live animals trivial?
I think he means the "change the world" vibe we get from a lot of startups. I think its great YC is broadening its scope into more traditional businesses.
Well, you got it right - YC is broadening its scope - that means that this investment is really kind of new type for Graham and co.
I guess, I should not ask what was so cool in AnyVivo but try to say in another words that it looks like a turn to traditional spheres to me.
It's probably true that YC companies are more willing to have a physical component (producing hardware hasn't worked out very well, but service, inventory, etc. might) now than in the past, but that is probably true of startups in general (Square is a great example; a tiny loss-leader physical product which is the wedge into a trillion dollar business).
I guess, online shops were kind of innovation when Amazon emerged but now they are rather common phenomenon and making them sell live animals is more about right organizational process than about building strong online product.