Personal Context Fabric

A portable life ontology that represents who you are, where you've been, and where you're going — in a form that's private, encrypted, owned by you, and usable across any AI provider.

What is PCF?

PCF is the personal instantiation of the Context Fabric Standard. Where CFS defines the primitives and protocols, PCF defines how those primitives represent a human life.

It's not memory. Memory is just recall of past interactions. It's not preferences. Preferences are just config flags. It's not a profile. Profiles are static snapshots.

PCF is a personal ontology + state machine + temporal graph that represents:

  • Who you are (identity, values, personality substrate)
  • Where you've been (history, learnings, context accumulation)
  • Where you're going (goals, projects, trajectories)
  • What you're constrained by (resources, obligations, limitations)
  • What you're connected to (people, organizations, tools, domains)
  • What state you're in (emotional, situational, energetic, focus)

And critically: how all of that evolves over time.

Why does this matter?

The constraint on AI usefulness right now isn't model capability — it's context poverty.

Models are smart enough to help with almost anything. But they don't know:

So every interaction is either you spending 80% of the time giving context, or the AI giving generic advice that ignores your actual situation.

A portable, rich, structured life context solves this. It's the difference between talking to a stranger vs talking to someone who deeply knows you.

Design principles

01

Local-first is non-negotiable

This is your life data. It can't live primarily on someone else's server. Your context lives on your device, encrypted with your keys.

02

Author, don't extract

Instead of AI building a model of you from your conversations, you author your own representation. You decide what matters.

03

Temporal-native

Everything has timestamps, validity windows, confidence levels. Your goals from 3 years ago aren't as relevant as yesterday's.

04

Graph-structured

Relationships between entities matter as much as entities themselves. "Project X relates to goal Y which conflicts with constraint Z."

05

Selective disclosure

It's not just "keep private." It's "reveal different facets to different agents for different purposes." Identity faceting, not binary access.

06

Portable by default

Connect to Claude today, GPT tomorrow, a local model next week. Your context travels with you because you control it.

Architecture overview

You
Author your context
Personal Vault
Local, encrypted, versioned
↓ selective disclosure
Claude
GPT
Local Model

Input methods

  • Direct editing via vault app
  • Conversation capture (AI suggests updates)
  • Integration imports (calendar, todo, health)
  • Inference (AI notices patterns you haven't stated)

Output/consumption

  • Context injection at conversation start
  • Agent queries ("what are Ryan's active projects?")
  • Proactive suggestions from local agent
  • Cross-tool coordination

Relation to CFS

PCF is a profile of the Context Fabric Standard, not a separate standard. It uses CFS primitives with a specific schema for personal context:

CFS Primitive PCF Application
Records Goals, projects, people, events, skills, resources, states
Links supports, conflicts_with, depends_on, part_of, caused_by
Views Context packages for specific AI tools — scoped projections
Tools MCP servers that expose context to AI providers
Receipts Audit trail of all access — who queried what and when
Policies Access rules, retention settings, sharing constraints
Sync Device-to-device replication, encrypted backup
Gates Validation before updates — schema conformance, bounds

This layering means PCF implementations automatically inherit CFS guarantees: bitemporality, verifiable provenance, portable export.

The deeper insight

The question isn't "is this a useful tool?"

The question is: who owns the data layer of human-AI collaboration?

Option A

Corporations own it, as proprietary moats, optimized for engagement and revenue, with all the manipulation gradients that implies.

Option B

Humans own it, as a portable standard, with all the interoperability and agency that implies.

PCF is Option B made concrete. It's not just infrastructure — it's a statement about who should control the most intimate layer of human-AI interaction.

Next steps