TECHNOLOGY
A persistent layer for data, permissions, relationships and learning. Where personal information is stored and where an AI request is processed are different choices. Prifina is building a foundation that can keep both visible and controllable.
What is different: capabilities attach to a person's foundation with explicit permission instead of absorbing it. What to do next: walk the layers, then tell us where you would connect.
ARCHITECTURE
Select any layer to read a plain language explanation. The whole stack is listed in order, so the diagram is never the only way to understand it.
A durable place for personal information, history, permissions and relationships. It persists across products, models and services, so a person does not rebuild their context every time they change something above it.
Position in the stack
CLAIM STATUS
Every capability on this site carries one of three labels. Available products, architecture direction and research vision stay separated.
CAPABILITIES
Available today, designed to enable, or under research. Nothing on this page converts a design goal or a research direction into a present tense product claim.
A persistent space holding personal information, history, permissions and relationships. It is the unit that survives when products, models or providers change.
Different tasks deserve different capabilities. Because no capability holds the foundation, planning, research, health and communication can each use whatever is best, and any of them can be replaced.
Personal data can be stored and an AI request can be processed on a personal device, in a selected cloud environment, on private infrastructure or inside a chosen country or region. The choice follows sensitivity, latency, cost and jurisdiction.
Access is expressed as a scope: a purpose, a data class, an audience and a duration. Grants can be reviewed at any time, narrowed, or withdrawn.
When several models and processing environments contribute to one answer, a person should still be able to see which source produced what. Provenance belongs to the personal layer rather than to any single provider.
A personal space is meant to be movable. Changing product, provider or storage environment should be an operation, not a migration project.
Removal is a primary operation. Withdrawing access, deleting a record and ending a learning permission are distinct actions, and each is available to the person.
A governed context that a family, a team, a network, a local area or a service created by a partner can hold together. Members contribute selected information; the rest of each personal space stays personal.
An agent acting on someone's behalf needs a scope that both the person and the system can inspect, and a record of what it did. We are studying how those permissions should be expressed and enforced.
If the richest personal context should not be pooled, learning has to work where the context lives. Federated learning is one possible method among several. It is not assumed to be the answer, and no distributed learning capability is claimed as available today.
NEXT