I've been building an AI companion that calculates your Human Design, astrology, and Gene Keys chart from your exact birth minute, then uses that as context for every conversation instead of answering like you're nobody in particular.
The reaction splits people hard. Half say "finally, advice that isn't generic." The other half say "I don't want an AI having a model of who I am, even one I confirmed myself."
We made confirmation explicit because of this. Eva reads your chart back claim by claim and only what you agree with enters her memory. But I keep wondering if that's enough, or if personalization this deep has a trust ceiling no UX can fix.
So, honestly, where's your line? What would an AI need to do (or never do) for you to let it actually know you?
(For context, this is Primal Eva, we launch here Wednesday. I'm asking because the answers change what we build)
The no-card-to-start move is such a smart trust builder, especially for a tool asking people to explore something as personal as their birth chart. Love that Eva actually remembers context between sessions instead of treating every question like a stranger walking in.