The distinction
| Personal AI | AI twin | |
|---|---|---|
| Works… | for you | as you |
| Faces | Inward: you are its audience | Outward: others talk to it |
| Job | Assistant, advisor, research partner, memory | Represents your knowledge, voice and perspective to others |
| Main risk | The privacy of what it knows about you | Misrepresenting you, or making commitments you didn’t make |
Many systems do both. The same knowledge base can power a private advisor for you and a public twin for others, with different boundaries on each.
Five layers
- Knowledge: what it knows. The 5Cs feed it, and the metadata gives it context, not just content. This is the layer you control and the one most worth getting right.
- Identity: whom it represents, and how that is verified.
- Boundaries: what it can access, say and do.
- Tools: the connections that let it act.
- Channels: where people meet it: web, chat, email, voice.
A twin is presence, multiplied
A twin is you in many conversations at once, even while you sleep. It scales your reach, not just your labor. Five duties follow:
- Disclosure: people know they are talking to a twin.
- Consent: from the person represented, and from the people whose data it holds.
- Limits: what it may never do.
- Freshness: how current its knowledge is, and who keeps it current.
- Accountability: the person behind the twin answers for what it says.
Use it: the twin charter
Before you build one, write one page: its purpose and whom it serves; what it knows and must not know; what it may and may never do; who can talk to it and how it says it is a twin; where it sits on the Relational Technology Architecture cube; and who reviews it, how often, and how it will be retired.