I'm not convinced we need a whole query "language" at all. The huge majority of database use is done on behalf of larger applications written in a different language. Why make developers learn another half-functional language just to be able to express their query intentions? I'd be far more interested in mechanisms to expose the querying possibilities in a developer's "native" language, which is why I found Berkus' sentence
"However, it's more likely that we'll be working more on direct language interfaces in the future instead"