Dyad is built by a two-person team in Berlin. It’s been in private beta since spring 2026. The cycle: a member writes a conversation prompt (a poem, a question, an essay) and proposes one to three times and places in the coming week when they would be open to meet. Another member writes a single response. After responding, they can invite the author to one of the times. If the author accepts, the exact location is shared and they meet. After the meeting (or not) both members leave feedback before either can use the platform again.

The platform doesn’t have messaging, recommenders, a follower graph, engagement metrics, or profile bios. Most of those absences are enforced at the database layer; the rest live in DESIGN.md and team discipline. The app is the coordination layer that gets two people to the same place at the same time. The conversation happens in person.

Invitations and waitlists
Private beta means specific people get invited specifically. Off-the-shelf waitlist tools collect email, name, source, and a behavioural profile by default.

We’re working toward steward ownership. That doesn’t preclude charging members, but it shapes how. Stripe Elements embeds deeply enough to require a consent banner; Stripe Checkout doesn’t but moves the user off-site. Each option ships with a conversion-tracking default that has to be actively turned off.

Community safeguarding
Stranger-meeting carries real safety considerations. No-shows. Reports of feeling unsafe. Harassment. What counts as the platform’s responsibility after the meeting ends. Most moderation tooling assumes content moderation, and we have almost no content. What we have instead is a post-meeting feedback gate and a small operations team reading every report. The design problem is making that scale without becoming surveillance, and negotiating the tension of commons built on trust against our real responsibilities as platform providers

The talk walks the tensions and out journey, warts and all, with the concrete decision, the cost we’re paying for it and the parts where we’re still working it out. The pattern across them is what I most want to surface for the room: building a values-aligned product on retention-shaped substrate is daily friction, every release, and most of that work is invisible from outside, including, often, to funders and to other builders working on adjacent problems. And, hardest of all.. to one another as founders, which is a tension we'd also love to talk about.

I’ll close with what’s next. The path toward steward ownership. The open question of how/whether to expand, federate, cooperating ownership..!