Zenith Sapience Architecture

This document dissects the control surfaces, invariants, and feedback mechanisms that enable the Zenith Sapience Initiative to behave like a fully automatable, CI-aligned demonstration using only the components present in AGI Jobs v0 (v2).

Control Layers

flowchart LR
    classDef core fill:#0b1120,stroke:#22d3ee,color:#22d3ee
    classDef svc fill:#1c1917,stroke:#f97316,color:#fed7aa
    classDef guard fill:#111827,stroke:#facc15,color:#facc15

    subgraph L1[Core Ledger Intelligence]
        Identity[IdentityRegistry]:::core
        Jobs[JobRegistry]:::core
        Rewards[RewardEngineMB]:::core
        Stake[StakeManager]:::core
        Identity --> Jobs --> Rewards --> Stake
    end

    subgraph L2[Autonomous Services]
        Planner[orchestrator/planner]:::svc
        Policies[orchestrator/policies]:::svc
        Runner[orchestrator/runner]:::svc
        Gateway[agent-gateway services]:::svc
        Planner --> Policies --> Runner --> Gateway
    end

    subgraph L3[Protective Envelope]
        CI[GitHub Actions ci.yml]:::guard
        Foundry[forge test + fuzzing]:::guard
        Monitoring[monitoring/ + routes/]:::guard
        Docs[docs/owner-control-*.md]:::guard
        CI --> Foundry
        Foundry --> Monitoring
        Monitoring --> Docs
        Docs --> CI
    end

    Identity -.feeds ENS proofs .-> Planner
    Runner -.emits metrics .-> Monitoring
    Stake -.drives payouts .-> Gateway

Invariants

1. Identity Sovereignty – Every mission execution path must originate from a verified ENS identity. Scripts in scripts/v2/ enforce ENS checks before any job is posted.
2. Thermodynamic Integrity – Reward allocation parameters in config/thermodynamics.json cannot bypass the PID guardrails; the Hardhat tests fail whenever allocations violate invariants defined in test/v2/RewardEngineMB.t.sol.
3. Dispute Resolubility – The dispute pipeline relies on contracts/v2/modules/DisputeModule.sol together with the flows covered in test/v2/jobLifecycleWithDispute.integration.test.ts; execution halts until disputes resolve.
4. Observability Coverage – No epoch upgrade is valid unless npm run lint:ci && npm run test && forge test && npm run coverage completes, reinforcing the parity with the CI fan-out and keeping Hardhat coverage plus Foundry fuzzing in the critical path under branch protection.

Data Contracts

CI & Deployment Hooks

• The demonstration inherits deployment-config/ templates; npm run owner:plan produces deterministic transaction batches before live updates.
• GitHub Actions produce the ci-summary.json artifact; docs/asi-feasibility-verification-suite.md references this artifact to keep automated and human audits synchronized.
• npm run owner:dashboard -- --json snapshots treasury and staking balances before and after orchestration upgrades, keeping financial invariants auditable.

Feedback Dynamics

stateDiagram-v2
    [*] --> PolicySynthesis
    PolicySynthesis --> ScenarioSimulation
    ScenarioSimulation --> IncentiveTuning
    IncentiveTuning --> JobEmission
    JobEmission --> AgentExecution
    AgentExecution --> LedgerSettlement
    LedgerSettlement --> TelemetryAggregation
    TelemetryAggregation --> AssuranceLoop
    AssuranceLoop --> PolicySynthesis

Every cycle ensures data gathered from execution (telemetry, settlement records, dispute logs) reshapes the next policy synthesis phase. The combination of orchestrator simulations and CI gating yields a continuous improvement engine that remains bounded by verifiable invariants.

Risk Posture

• Economic Shock Absorption – Scenario banks in data/ encode macroeconomic stressors; the simulator can replay them to rehearse how the incentive thermostat responds.
• Governance Attack Surface – Owner control playbooks define multi-signature requirements and manual overrides; docs/owner-control-emergency-runbook.md stipulates the minimum quorum before emergency toggles are invoked.
• Compliance Envelope – The identity registry metadata includes jurisdiction tags, enabling automated compliance filters prior to job allocation.

Zenith Sapience is therefore not a theoretical aspiration; it is a precise composition of the AGI Jobs v0 (v2) stack designed to operate responsibly at global scale.