marakaci

We do both fiction (romance, fantasy, mystery, erotica, paranormal) and non-fiction (memoirs, self-help, programming, spiritual) with separate flows, and visual books like colouring books are on the way.

thanks, it's a good suggestion, we definitely need to play around a little bit to showcase what we do better!

yep, it would, there is nothing magical though. We collect a lot of context about the couple, and GPT-3 can provide more precise suggestions then. We get some decent bits of info from their diary, feedback and onboarding process

And we also played with dozens of different prompts to choose the one that performs the best.

Like, when I write something as "I felt unpleasant when I noticed that my partner did not provide me enough attention when I told her that I have an issue in my family, I feel that she does not understand well how to support me"

and gave an answer in the feedback like "I want to talk more about how we can support each other in situations we feel down", the GPT-3 was able to pick the context with our guidance and give a question "What steps will you take to ensure that you both feel supported and heard in frustrating situation" for the next date.

And I also think we can bring it to another level with fine-tuning

does it work on some pre-made role-based text-davinci prompts? Like "answer all the questions in this chat is if you were a marketing assinstant user_input: *" or is there more to it?