@PaulBacon Yes, this could work with pre defined times, but the system could also plan and write its own prompts based on the user actions.
To not have the rabbit just start talking it could work like some sort of notification / voicemail system to engage with what was being ‘created’ for you.
The rabbit agent in the cloud would be able to look at the user profile, recent news, and recent interactions. Based on that it would plan to send a ‘voicemail’ like
‘Hi, remember the thing we talked about yesterday? The assassination attempt, there is an update on it! Pretty unexpected developments. Want to listen to it?’