Yeah, I honestly couldn't be arsed to search in the twitter hose but he specifically corrected me on the topic.

edit: and to clarify his comments were mostly that the data is not realtime, but the idea that audio recorded is then processed and within a day or two was much more realistic.