Right now, the babbler parses text in one of two ways:
- As a request for a babble
- As just regular conversation to feed the babblers
The decider for which case is the existence of the word "says" as the second token in the text. If the second token is "says" then case 1 else case 2. If there isn't a babbler, this will just be a failed babble request, where it could have been used to feed a babbler.
Currently it's broken into these two cases because if we didn't filter things this way all babblers would produce "seabass says" with high probability.