04 · Human + AI Experience View

Shared Context, Separate Continuity

              BUSINESS CONTEXT
                     │
          ┌──────────┴──────────┐
          ▼                     ▼
        HUMAN                   AI
          │                     │
    Human Workbench       AI Workbench
          │                     │
 Personal Contextual      AI Continuity
      Memory                 Memory
          │                     │
          └──────────┬──────────┘
                     ▼
                Shared Work
                     │
                     ▼
              Business Decision
              (human-owned)
                     │
                     ▼
              Governed Execution

Humans govern. AI helps. Spine preserves continuity.

Who Does What

Responsibility	Human	Twin	Spine Agent
Own the decision	✅	—	—
Grant authority / approve	✅	—	—
Provide business context	Spine (via projection)	Spine (via projection)	Spine (via projection)
Assemble active context	—	✅	✅
Reason and plan	Reviews	✅	✅
Select capability	—	Proposes	Proposes
Execute externally	—	Via Hermes, if authorized	Via Hermes, if authorized
Verify outcome	Reviews	Assists	Assists
Continuity	Personal Contextual Memory	AI Continuity Memory	Agent Continuity

Twin

The Twin is the human-facing AI coworker, not a separate Workbench.

• Identity: twin:{userId} — one Twin per human, bound to that human's identity and tenant.
• Relationship to user: Knows personal context, current Work, and what the user is authorized to see — via the AI Workbench projection, not via unrestricted Spine access.
• What Twin does: Context assembly, summarization, planning, capability selection, drafting approvals — always in the user's Work context.
• What Twin does not do: Own decisions, bypass governance, or silently promote AI inference to organizational truth.
• Memory: AI Continuity Memory — Twin's working continuity that serves Business Context, tracked for rehydration and audit.

Twin lives inside the AI Workbench:

BUSINESS CONTEXT SPINE
        │
        ▼
  AI WORKBENCH (projection over BC)
        │
        ├── Twin (human-facing coworker)
        ├── Agent runtime
        ├── Model routing / ADK
        └── Capability access (governed)
        │
        ▼
  AI CONTINUITY MEMORY (serves BC)

See Twin Execution for the full execution flow.

Spine Agents

A Spine Agent is a governed organizational actor with identity, role, authority, capability access, Work state, evidence, and continuity.

Identity → Role → Authority → AI Continuity → Business Context
  → Observe → Orient → Plan → Capability → Governance
    → Execute → Observe Outcome → Verify → Continue

• Not a chat persona. Not a tool wrapper.
• Can own a subtask, hand off via Shared Work, and leave evidence.
• Every agent action is authorized, traced, and reconcilable back to the Spine.

See Spine Agent Execution and Agent Collaboration.

Human–AI Collaboration

Cheap handoff, not repeated explanation.

• Human and Twin see the same Active Context Bundle (Business Context slice + Work + evidence + authority).
• When a Spine Agent joins, it hydrates from the same Work and Business Context — it never needs "the whole conversation pasted again."
• Decisions remain human-owned. Twin proposes; humans approve. Agents execute only what is authorized.

Human opens Work → Twin has Active Context → Twin proposes plan
  → Human approves → Agent takes subtask → Hermes executes
    → Outcome verified → Spine updated → Both workbenches rehydrate

Decision Ownership

• Every decision has an owner (human identity), a Work context, and evidence.
• Every execution has an authorization (governance verdict), not just a trigger.
• Every outcome has verification before it can update organizational truth.

Where to Go Next

<CardGroup cols={2}>

The common unit everything coordinates through.


What happens when AI receives work — the OODA-derived internal flow.


Why two working memories serve one organizational memory — and G-10.


How "may it be done?" is decided before Hermes ever executes.

</CardGroup>