A template tells 35M people the same thing. This writes each saver's message from their own streak, goal and reward, in their language: the kind of message that gets a save today and brings a lapsed user back. Jar's rules still pick who hears what; the model writes it, within Jar's approved claims.
Six checks run on every message: valid output, fits the channel, right language, no invented numbers, approved claims only, and the key fact included. A draft that fails a check goes back to the model once, with the failures, for both models.
Six checks run on every message: valid output, fits the channel, right language, no invented numbers, approved claims only, and the key fact included. A draft that fails a check goes back to the model once, with the failures, for both models.
