Sultan

Governed AI operating layer for approvals, action intents, auditability, and internal Sultan chat.

Command

OpenClaw

Continuity Ledger

This is the first governed continuity layer: operational memory, versioned beliefs, contradictions, self-model snapshots, and consequence memory.

Memories

State under governance, not just stored notes.

1

Beliefs

Claims that can strengthen, mutate, or die.

1

Contradictions

Open tensions that stop memory from turning into dogma.

1

Self-Model Snapshots

Versioned operational identity, not mythology.

1

Shared Memory Contract

Continuity should inherit one company-wide doctrine for what becomes memory, how it stays bounded, and what a decision record must contain.

5 memory types / 11 decision fields

Operating Loop

Reality -> Memory -> Governance -> Planning -> Execution -> Evaluation -> Learning

Shared Principles

  • Memory belongs to the organization.
  • Every important decision leaves evidence.
  • Authority is delegated, not assumed.
  • Approval-bound actions must stay visible.

Memory Promotion Rules

  • Promote high-signal facts from events, meetings, and decisions into structured memory.
  • Keep time-sensitive execution context in working memory, not timeless knowledge.
  • Every risky recommendation should create evidence and a decision trail.
  • Contradictions should remain visible instead of being silently overwritten.

Decision Record Fields

  • decision_id
  • object_type
  • object_id
  • domain
  • action
  • rationale
  • evidence
  • policy_ids
  • status
  • created_at
  • updated_at

Operational Memory

What the machine is holding onto and why.

1 entries

Refund approvals need finance evidence

operationalrefund_request

82%

Refunds above threshold keep producing safer outcomes when finance evidence is attached before approval.

Evidence: refund workbench, approval history

Belief Ledger

Versioned claims with evidence and confidence.

1 beliefs

refund_evidence_improves_approval_quality

Financeactive

82%

Finance evidence before approval reduces refund drift and poor approvals.

Contradiction Engine

Open contradictions are healthy. Hidden contradictions create drift.

1 records

Manual follow-through still exists in places where the operating constitution wants machine-enforced workflow.

memory:manual_task_layersgoal:system_manages_work

open

Self-Model

Operational doctrine, limitations, assumptions, and drift from prior versions.

1 versions

Sultan v0 continuity baseline

Moderately capable in governance and operator memory, still early in continuity and eval rigor.

Capabilities

  • Tracks governed approvals and action intents
  • Maintains shared weekly coordination pressure
  • Stores campaign briefs, experiments, and intelligence notes

Limitations

  • Live Google Sheets sync not yet connected
  • No full autonomous consequential execution
  • Belief revision and contradiction handling are early-stage

Consequence Memory

Reality should either strengthen, weaken, or retire the machine’s claims.

1 outcomes

refund_evidence_improves_approval_quality

Structured finance evidence before approval produced cleaner downstream reconciliation.

learning