Personal Intelligence
Intelligence,
made personal.
PI is a new kind of computer designed around intelligence rather than applications.
Your memory. Your context. Your permissions. Your intelligence.
The transition
We are still living in the mainframe era of AI.
Early computers were too large and expensive for individuals to own. The personal computer changed that. AI may be approaching a similar transition.
Mainframe
Intelligence lives inside institutions.
Access is mediated. Compute is remote and shared.
PC
Computing becomes personal.
The machine sits on your desk. The work is yours.
PI
Intelligence becomes personal.
Memory, context and reasoning belong to you.
The current architecture
Today, intelligence
is rented.
The model provider controls the infrastructure, model, pricing, limits and availability. The user receives intelligence, but does not possess it.
The PI architecture
The cloud doesn't disappear. It moves to the edge of the system.
Data stays local by default. Only selected requests travel outward to frontier cloud intelligence.
The cloud becomes an extension of the personal machine rather than the owner of its intelligence.
Intelligence instead of applications
Applications wait for instructions. Agents pursue goals.
Instead of opening an application and performing a task manually, the user describes an objective. PI determines which tools, information and actions are required.
Traditional computer
A grid of tools. You choose one, then do the work by hand.
PI
One objective, resolved into the tools, information and actions it requires.
“Prepare everything I need for tomorrow's meeting.”
- understands the meeting
- retrieves relevant context
- identifies previous decisions
- gathers documents
- prepares the required information
Your context is the moat
A model can be replaced.
Your context cannot.
A generic model may know more facts than any human being. It still does not know which facts matter to you.
Who owns your intelligence?
- —Intelligence
- —Infrastructure
- —Model availability
- —Usage limits
- —Much of the surrounding context
- Memory
- Models
- Permissions
- Identity
- Local intelligence
- When the cloud is used
Privacy isn't a setting.
It's an architecture.
The hardware problem
Personal intelligence changes the hardware problem.
Continuous intelligence requires compute, memory and energy. The objective is no longer simply maximum computational performance. It is useful intelligence per watt, per dollar and per unit of memory.
- Maximum compute
- Huge power
- Huge cooling
- Centralized
- Continuous inference
- Low power
- Local memory
- Personal
Concept hardware
The PI Brain
Compute designed for continuous personal intelligence.
- Local Inference
- Memory
- Power Efficiency
- Optical Computing
- Agent Runtime
The PI Brain is the computing architecture envisioned to make persistent local intelligence physically practical. It describes a direction, not a shipping product.
Optical computing
When intelligence becomes continuous, efficiency becomes the constraint.
Moving and processing information with light could eventually provide significant improvements in performance per watt for certain AI operations.
Optical computing is a potential underlying technology, not a proven PI product specification.
Not cloud.
Not local.
Both.
- Private
- Frequent
- Predictable
- Low latency
- —Difficult
- —Frontier
- —Large-scale
- —Exceptional capability
The future is unlikely to be entirely cloud or entirely local. It will be hybrid.
The new computer
The computer is changing.
Mainframe
Shared
Personal Computer
Owned
Smartphone
Everywhere
Personal Intelligencer
Intelligent
The personal computer was designed around applications. PI is designed around intelligence.
The cloud may contain the world's general intelligence.
Your machine should contain the intelligence that understands you.
The Personal Intelligencer