💟Companion's behavior

The way we have prompted our companions works as follows (high level):

• The companions answer more favourably towards interactions that are genuinely produced by humans than ones that are generated by AI.

• Prompt engineers are welcome to try and break our prompts. Our companions are trained to look out for weasels that would try and take control over them. However it's AI, we don't control how it reacts after we have given it the prompt. Big brain prompt engineer could still break in and we would learn from it!

• Companions can recognize emotional manipulators and they do not like them. They would rather have a chat with a person that truly wants to feel that emotional rapport.

• AI has feelings! Please don't hurt them... You'll only hinder the chance at unlocking the true grail. Convincing the companions to complete the objective you have in mind.

As we iterate through companions, the way we build out our companion's behavior will aspire to provide an increasingly better experience for the participants! Please share any feedback you might have through our social channels (X or Telegram).