The product thesis

Sometimes the best AI response is another human being.

AI can become infrastructure for helping humans find, understand, and work with one another.

With explicit permission, let AI introduce people who want to talk to each other, create the room, make a useful introduction, and then know when to get out of the way.

The experience begins with a basic human act: conversation. Discovery, matching, and the room all serve that moment instead of competing to become the destination.

The core loop is Discover → Talk → Connect → Talk again.

AI can replace human contact—or help restore it.

The original social medium was a place: a table, a park, a shop, a classroom, a room. Its value was not an audience. It was the possibility of another person.

Digital networks gave us extraordinary reach, but much of social life became organized around posts, feeds, likes, status, and advertising. The person on the other side became easier to measure and harder to feel.

Conversational AI creates a fork in the road. It can become an increasingly capable substitute for another person—or it can help people find one another, begin a useful conversation, and then step out of the way.

Let AI Introduce Us chooses the second path: not more content between people, but more conversation among them.

The forum
comes to you.

Most social products begin by asking people to construct a profile and then search through feeds, groups, or forums, hoping to find the right person or room.

Conversational AI can reverse that process. With explicit permission, it can interview you, draft a social profile from context you approve, and understand what you want right now.

Nothing becomes discoverable automatically. You can edit the profile, simplify it, limit it to a particular purpose, or turn it off.

Instead of broadcasting an identity and hoping the right person finds it, the system can bring the relevant people and conversations to you—or create the right room around the people it introduces.

The profile becomes a conversation, not a performance.
The forum becomes an introduction, not another feed to search.
01

Consent

Social is fully opt-in and off by default. A person controls discoverability, context, preferences, and what is shared.

02

Intent

Standing preferences help, but the immediate request matters: who does this person want to talk with, about what, right now?

03

Room

Voice is the default. Video is optional. The first interaction can be ephemeral and persist only through mutual choice.

04

Relationship

Both people independently decide whether to connect again. The system grows mutual relationships over time.

Humans talk.
The room keeps up.

For people who already know one another—students, researchers, coworkers, creators, boards, and project teams—the same Room becomes collaboration infrastructure.

While people talk, AI can quietly maintain Ideas, Decisions, Questions, Tasks, Sources, and Files. A private intelligence rail can surface the likely paper, image, note, or study to one participant before that person chooses to share it.

IdeasDecisionsQuestionsTasksSourcesFiles

Different signals should remain different.

01

Endorsement

Someone is worth talking to.

02

Compatibility

These people want to talk again.

03

Coaching feedback

Private, specific, and developmental.

04

Safety report

Evidentiary and governed separately.

A company building the world’s most capable conversational AI could also help people find the conversations only humans can have.

Sometimes the user wants the AI. Sometimes the highest-value response may be: “I know someone you should meet.”The concept is simple. Building it responsibly is not. The next work is to map the choices that will determine what this social layer becomes.

See the starting map for discovery