I'm building the preference layer first. The rest is hypothesis.

Remembering you inside one app helps that app. It doesn’t help the next company’s chat. What’s “portable” today is usually a messy paragraph you copy by hand. I’m testing whether small, clear pieces can cross a brand boundary without moving your whole history.

Advice from other people is thin when few people have shared anything. Does keeping your own preferences still help before a circle fills in?

Will people permit pulling useful bits from their chats—and what does “I can take it back” have to feel like to count?

I’m starting with AI assistant chats (where access exists). Human texts (iMessage, WhatsApp) may hold richer judgment—and are much harder to reach.

Load-bearing and unsolved. Without a novel answer, any “network” stays decorative. For now trust should be something you declare (Maya marks Leo), not something an algorithm invents.

Once shared judgment has commercial value, expert voices can drift toward promotion.

Hotels pay for a guest ready to book—not for better guesses inside someone else’s chat. Si as a layer may improve the stay without capturing Booking-style economics unless it owns a moment worth paying for.

Loyalty and platform spend stories are plausible on paper. I have not validated that brands will fund this.