To be precise, Matt said:
Different algorithms for different queries? That's an entirely new thread :-) Let's just go with the obvious: It was non techie summarization for the public.
There are a lot of ways to read Matt's statement, but every interpretation I can think of involves some level of manual human intervention.
I would guess that what they did is increase the weight of previously recorded human reviewer data based on brands. Again, humans are being paid specifically to decide which brand should rank on certain searches. How that has been ignored in this thread is a mystery to me.
It's definitely possible that they added in some new automated logic too but that does not need to be true in order to explain what people are seeing.