Intelligence can change
Commercial AI, open-source models, specialist agents and partner-hosted intelligence can work with the same permissioned personal foundation. Changing the intelligence should not require rebuilding the person.
TECHNOLOGY
The personal intelligence layer.
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.
THE PRIFINA STACK
The personal layer stays. Everything around it can change.
Prifina puts a persistent personal intelligence layer between a person's information and the applications, AI models and partner services that work with it. The context, memory, permissions and relationships remain around the person while the intelligence and experiences above them can evolve.
(opens the full size diagram in a new tab)The durable core is the personal intelligence layer. Souvéa, AI Twin, Spaces and third-party applications can use it; commercial, open-source and partner-hosted AI can provide intelligence; and Zones can connect trusted organizational services, data and policies.
Applications can change
Souvéa, AI Twin and Spaces are current Prifina experiences. The same layer can also support third-party applications that receive only the context and permissions needed for their purpose.
Organizations can connect without owning the person
Zones allow selected partners to connect trusted information, services, AI and governance to people and Spaces while the individual's private context remains separately controlled.
The goal is a personal layer that supports an open ecosystem around the individual. Over time that ecosystem can include third-party applications, commercial and open-source AI models, AI hosted by Prifina or by universities, companies and infrastructure partners, specialist agents, and Zones with partner services.
BUILD ON THE PERSONAL LAYER
We work with selected organizations, application builders, AI and infrastructure partners, and research groups exploring new uses of persistent personal intelligence.
ARCHITECTURE
Seven layers, from a person's existing information to approved learning.
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.
The layers
AVAILABLE TODAY: In use in current Prifina products or deployments.
Personal data and intelligence space
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
- Personal information sources
- Personal data and intelligence space, currently selected
- Permission and governance layer
- Interchangeable AI capabilities
- Storage and processing choices
- Governed shared contexts
- Approved learning signals
CLAIM STATUS
What exists today, what runs with partners, what is designed, and what we are researching.
Every capability on this site carries one of four labels, matched to its actual current state.
- AVAILABLE TODAY: In use in current Prifina products or deployments.
- In use in current Prifina products or deployments.Souvéa and AI Twin are in market. More than 2,000 AI Twins have been created, and the product has paying customers.
- AVAILABLE WITH SELECTED PARTNERS: Available through joint implementation or partner-specific deployment.
- Available through joint implementation or partner-specific deployment.
- DESIGNED TO ENABLE: Architected, but not yet available as a working capability.
- Architected, but not yet available as a working capability.
- WE ARE RESEARCHING: Exploratory research direction; no current availability implied.
- Exploratory research direction; no current availability implied.
CAPABILITIES
Every capability, labelled honestly.
Available today, available with selected partners, designed to enable, or under research. Each label reflects the capability's actual current state.
Personal data spaces
A persistent space holding personal information, history, permissions and relationships. It is the unit that survives when products, models or providers change.
- Connect existing sources instead of migrating everything once
- Keep history in one place across services
- Carry relationships and permissions, not only records
AI model and capability choice
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.
- Capabilities receive access; they do not acquire the data relationship
- Neutral capability nodes, never locked to one provider
- Replacement does not reset a person's context
Storage, processing and jurisdiction
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.
- Where personal data is stored is an architectural choice, not a marketing setting
- Not every capability can be processed everywhere today
- Nothing here requires everything to run on a phone
Permissions
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.
- Purpose bound
- Time bound
- Reviewable
- Revocable
Provenance
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.
Portability
A personal space is meant to be movable. Changing product, provider or storage environment should be an operation, not a migration project.
Deletion
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.
Agent governance
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.
Distributed learning research
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.
Shared contexts
AVAILABLE TODAY: In use in current Prifina products or deployments.
Souvéa Spaces
People can create shared contexts around selected people, information and AI. Only what participants contribute to the Space becomes shared; the rest of their personal context remains separate.
Explore Spaces in Souvéa(opens in a new tab)- Person led SpaceSelected people, information and AI around a shared purpose.
- Professional or member environmentShared expertise and knowledge with personal boundaries preserved.
- Organization provided ZoneGoverned Spaces, information, AI services and policies.
NEXT