Unfortunately, it seems a bit awkward to differentiate between empty and NULL strings. That's something to be careful of.
Also, I really think this should be included in the blog post, even if it's simple. Protecting against SQL injection is not optional, so leaving it out only muddies the comparison with more traditional frameworks. Also, there's always going to be someone that copies and pastes it without thinking much about it.
That being said, this is very cool and I hope to see more development and exploration in this area.
This behavior actually is what anyone who has used Oracle is accustomed to (empty string is NULL). I don't recall the default behavior in Postgres but Postgres is "Oracle-ish" in a lot of ways so I would be surprised if this is not the default there as well.
Don't get me wrong -- it's still a pretty neat idea. It's just there's that "other 90%" (i.e. the necessary work to create anything seriously production grade) that always comes to mind when evaluating a new interface.