Every company is building an AI agent right now. Almost nobody is building the other side of the conversation.

An agent can verify another agent. It cannot verify a human, and it cannot get consent from one. So when an agent needs to reach a real person, it does what software has always done: it buys a scraped list, guesses at intent, and contacts people who never agreed to be contacted. That was tolerable when a human sent the email. It breaks when a million agents do it continuously, and it breaks faster as the open web fills with machine-generated text that reads like human signal and is not.

twin3 builds the human layer those agents are missing.

Each person holds a Twin Matrix: a 256-dimension profile of their real-world experience, issued on-chain as a credential that cannot be transferred or sold. An agent can query it and receives only the dimensions that person authorized, for as long as they authorized them. The person decides what is visible, to whom, and for how long.

On top of that sits an exchange. An enterprise agent finds a person's personal agent, negotiates with it, and buys one specific slice of their experience: what they intend, what they did, what they value. The person is never exposed, because the profile is an encrypted classifier. The person keeps 70% of what their data earns.

More than 143,000 credentials have been issued on BNB Chain with no paid acquisition, across 122 countries.

We are building from Taipei, on twenty years of human-computer interaction research into a single question: how do you represent a person to a machine in a form the machine can act on.

twin3.ai

Org chart

Teams

Offices

twin3.ai

Org chart

Teams

Offices

Related companies