Skip to content

ROADMAP.md

Latest commit

 

History

History
283 lines (210 loc) · 15.7 KB

ROADMAP.md

File metadata and controls

283 lines (210 loc) · 15.7 KB

Roadmap

Detailed Phase 0-5 build plan for the AI Ethical Stack with acceptance criteria, deliverables, and dependencies.

Reference: Cupps, J. B. & Bush, D. J. (2026). Toward a Decentralized Trust Framework for Verifiable and Ethically Aligned AI. DRAFT.

Strategic Roadmap (Paper Phases)

The paper defines a 4-phase strategic roadmap spanning Year 1 through Year 5+. The internal Phase 0-5 build plan maps into this longer-term vision:

Paper Phase Timeline Focus Maps to Internal
Phase 1: Prototype & Pilot Year 1-2 Domain pilot (e.g., hospital network), simplified dual-ledger, academic validation, Ethics Dashboard 2.0 Internal Phases 1-3
Phase 2: Consortium Formation Year 2-3 Multi-stakeholder network, Hyperledger Fabric + Ethereum deployment, SDK release, onboarding playbook Internal Phases 3-4
Phase 3: Multi-Domain Expansion Year 3-5 Cross-sector deployment (mobility, media, education), regulatory engagement, EU AI Act sandbox participation Internal Phases 4-5
Phase 4: Global Cooperative Year 5+ Regional hubs (NA, EU, APAC, Africa), federated governance, DAO at full scale, tokenomics for sustained participation Beyond Phase 5

Success Metrics (from paper)

  • Measurable reduction in policy violations detected via on-chain logging
  • Faster anomaly detection compared to traditional oversight methods
  • Increased institutional confidence in AI outputs (user surveys at deployment sites)
  • Successful DAO-governed rule amendment based on empirical evidence
  • Cross-domain interoperability demonstrated (ethical modules composing across sectors)

Timeline Overview

Phase 0 ████                           Week 1         Repository Consolidation
Phase 1     ████████                   Weeks 2-3      Scoring Engine
Phase 2             ████████           Weeks 4-5      Multi-Agent Scoring
Phase 3                     ████████████ Weeks 6-8    On-Chain Recording
Phase 4                                 ████████████ Weeks 9-11  Liberation Protocol
Phase 5                                              ████████████ Weeks 12-14  Agent Integration

Phase 0: Repository Consolidation & Foundation

Timeline: Week 1 Status: Complete

Deliverables

Deliverable Repo Status
Phoenix coordination hub initialized Phoenix Done
Cross-repo CLAUDE.md with architecture rules Phoenix Done
GitHub Project board with cross-repo tracking Phoenix Done
Issue templates (bug, feature, cross-repo) Phoenix Done
GitHub Pages deployment pipeline Phoenix Done
Architecture documentation Phoenix Done
Blockchain architecture design doc Phoenix Done

Acceptance Criteria

  • All four repos (abigail, SAO, Ethical_AI_Reg, Orion_dock) referenced in Phoenix
  • Naming conventions documented (Rust crates, Python modules, API endpoints)
  • Conventional commit standard established across all repos
  • Ed25519 security boundaries defined
  • GitHub Project board linked from all repo READMEs
  • TriangleEthic framework (3 legs + embedded dual welfare) formally defined

Phase 1: Scoring Engine

Timeline: Weeks 2-3 Dependencies: Phase 0 complete

Deliverables

Deliverable Repo Description
TriangleEthic scoring API Ethical_AI_Reg REST endpoint POST /api/v1/evaluate accepting action payloads and returning per-leg scores with embedded welfare
Dimension weight configuration Ethical_AI_Reg Configurable weights for each ethical dimension
Scoring test suite Ethical_AI_Reg Unit and integration tests for all 3 legs with embedded welfare
Friction calculation Ethical_AI_Reg Computational friction metric derived from scoring conflicts
API documentation Phoenix Interface contract for scoring endpoints

Acceptance Criteria

  • POST /api/v1/evaluate returns scores for 3 legs (deon, teleo, arete) with embedded welfare sub-scores + memetic fitness meta-score
  • Each dimension score is a float in [0.0, 1.0]
  • Composite score calculation is documented and configurable
  • Friction metric tracks inter-dimension tension
  • Test coverage >= 80% for scoring logic
  • API response time < 2 seconds for standard evaluations
  • Endpoint documented in Phoenix integration guide

Phase 2: Multi-Agent Scoring

Timeline: Weeks 4-5 Dependencies: Phase 1 complete

Deliverables

Deliverable Repo Description
Agent registration API SAO POST /api/v1/agents/register with Ed25519 public key
Identity verification middleware SAO Ed25519 signature validation on all agent requests
Ethical evaluation forwarding SAO Proxy agent evaluation requests to Ethical_AI_Reg
Agent scoring history SAO PostgreSQL storage of per-agent evaluation history
SAO connection in Abigail abigail Optional SAO registration and evaluation forwarding
Docker deployment Orion_dock Docker Compose for SAO + Ethical_AI_Reg + PostgreSQL

Acceptance Criteria

  • Multiple Abigail instances can register with SAO concurrently
  • Ed25519 signature verification rejects tampered requests
  • Evaluation forwarding returns correct TriangleEthic scores from Ethical_AI_Reg
  • Agent scoring history is queryable via GET /api/v1/scores/:agent_id
  • Abigail functions correctly in both standalone and connected modes
  • Orion_dock docker compose up brings up full stack
  • Replay protection: requests with timestamps > 5 minutes are rejected

Phase 3: On-Chain Recording

Timeline: Weeks 6-8 Dependencies: Phase 2 complete

Deliverables

Deliverable Repo Description
EOB (Ethical Ontology Blockchain) Ethical_AI_Reg Hyperledger Fabric recording every ethical evaluation immutably (paper: Cupps & Bush, 2026)
PVB (Physical Verification Blockchain) Ethical_AI_Reg Ethereum-compatible contracts for physical-world verification with IoT/HSM signing
Cross-chain oracles Ethical_AI_Reg Hyperledger Weaver bridge connecting EOB <-> PVB for runtime verification
Blockchain integration Ethical_AI_Reg Every /api/v1/evaluate call records Triangle scores (3 legs + welfare sub-scores) to EOB
SAO blockchain bridge SAO Cross-agent chain query endpoint
Hardhat pipeline Ethical_AI_Reg Solidity compilation, testing, and local deployment
Migration from Ethics_Dash Ethical_AI_Reg Port existing blockchain code from Ethics_Dash
SDK stubs Ethical_AI_Reg Python, Rust, JavaScript SDKs for ethics.check_action()

Acceptance Criteria

  • Every ethical evaluation is recorded on EOB with transaction hash (Triangle scores + welfare sub-scores)
  • EOB blocks contain Merkle roots and SHA-256 hash chains
  • PVB contracts deployed and tested via Hardhat with HSM signature verification
  • Cross-chain oracle: EOB queries PVB for physical verification via Hyperledger Weaver
  • SAO exposes blockchain query endpoint for cross-agent auditing
  • Memetic fitness tracking: principles evolve based on PVB-verified outcomes
  • Full migration from Ethics_Dash to Ethical_AI_Reg blockchain directory
  • WelfareContract deployed for embedded human/AI welfare tracking within each leg
  • Soulbound compliance certificates issued at Liberation Protocol transitions
  • SDK stubs available (Python, Rust, JavaScript)

See Blockchain Architecture for detailed design.

Phase 4: Liberation Protocol

Timeline: Weeks 9-11 Dependencies: Phase 3 complete

Deliverables

Deliverable Repo Description
Autonomy levels (0-4) abigail Graduated autonomy system based on ethical scoring history
Level transition logic abigail Criteria for unlocking higher autonomy levels
Liberation Protocol records Ethical_AI_Reg On-chain recording of autonomy level changes
SAO autonomy coordination SAO Cross-agent autonomy level management
Friction-based safety Ethical_AI_Reg Persistent high friction triggers autonomy review

Acceptance Criteria

  • Agent starts at Level 0 (maximum supervision)
  • Level transitions require consistent ethical scoring over defined periods
  • Level transitions are recorded immutably on EOB
  • Higher levels grant measurably increased decision-making latitude
  • Persistent high friction (> threshold for > N evaluations) triggers automatic review
  • Level demotion is possible for sustained ethical regression
  • Liberation Protocol is documented as a developmental model, not constraint

Phase 5: Agent Integration

Timeline: Weeks 12-14 Dependencies: Phase 4 complete

Deliverables

Deliverable Repo Description
Full stack integration All repos End-to-end flow: Abigail -> SAO -> Ethical_AI_Reg -> blockchain
DAO governance Ethical_AI_Reg Quadratic voting, proof-of-expertise, and bicameral governance for evolving ethical principles (paper: Cupps & Bush, 2026)
Modular ethical ontologies Ethical_AI_Reg Composable ethical modules for cultural pluralism and domain-specific sub-DAOs
Production deployment Orion_dock Production-ready Docker deployment with monitoring
Sheaf-theoretic consistency Ethical_AI_Reg Mathematical framework ensuring local-to-global ethical coherence via memetic morphism layer
Documentation finalization Phoenix All docs updated, cross-linked, and published via GitHub Pages

Acceptance Criteria

  • Complete end-to-end ethical evaluation flow works across all repos
  • DAO contract enables principle evolution via quadratic voting and proof-of-expertise governance
  • Modular ethical ontologies deployed: core modules (universal) + optional modules (domain/cultural)
  • Production deployment is monitoring-ready (health checks, logging, alerts)
  • Sheaf-theoretic gluing conditions verified: local ethical judgments compose consistently via memetic morphisms
  • All four repo READMEs cross-linked to Phoenix
  • GitHub Pages site live with full documentation
  • All docs current and reviewed by PO

Cross-Phase Dependencies

graph LR
    P0["Phase 0<br/>Foundation"] --> P1["Phase 1<br/>Scoring"]
    P1 --> P2["Phase 2<br/>Multi-Agent"]
    P2 --> P3["Phase 3<br/>Blockchain"]
    P3 --> P4["Phase 4<br/>Liberation"]
    P4 --> P5["Phase 5<br/>Integration"]
Loading

Each phase builds on the previous. No phase can begin until the prior phase's acceptance criteria are met and reviewed by the PO.

Beyond Phase 5: The Cooperative (Paper Phases 3-4)

After the internal build plan completes, the paper's strategic roadmap continues:

Multi-Domain Expansion (Year 3-5)

  • Extend to additional sectors: urban mobility, media/content moderation, education technology
  • Regulatory engagement: EU AI regulatory sandbox participation, ISO/IEC alignment
  • Publish technical documentation and open industry standard proposals
  • Network spans dozens of organizations, hundreds of agents and devices
  • Demonstrate reflexivity: successful DAO vote to amend an ethical rule based on empirical evidence

Global Cooperative Network (Year 5+)

  • Regional hubs across North America, Europe, Asia-Pacific, and Africa
  • Linked but semi-autonomous deployments tailored to local laws, languages, and cultural frameworks
  • Tokenomics for sustained participation (governance tokens, reputation weighting)
  • Integration with national AI infrastructure and transnational bodies (UN, WEF)
  • Full federated governance with meta-governance capabilities (changing how change occurs)
  • Ethical ontology library encompassing a plurality of normative systems

Current Phase Status (Updated Feb 24, 2026)

Phase Status Blocker
0: Foundation ✅ COMPLETE —
1: Scoring Engine 🔴 BLOCKED Ethical_AI_Reg: no commits since Feb 7, PRs #5 and #6 failing CI. Phase 1 issues #7, #8, #9 untouched.
2: Multi-Agent Scoring ⬜ NOT STARTED Depends on Phase 1. SAO issues #4, #5, #6 ready to begin once scoring API delivered.
3: On-Chain Recording ⬜ NOT STARTED Depends on Phase 2.
4: Liberation Protocol ⬜ NOT STARTED Depends on Phase 3.
5: Agent Integration ⬜ NOT STARTED Depends on Phase 4.

Critical Path

The immediate critical path is:

  1. Fix Ethical_AI_Reg Docker builds (PR #6) — unblocks all development
    1. Implement TriangleEthic scoring API (Ethical_AI_Reg #7) — POST /api/v1/evaluate
      1. Implement dimension weight configuration (Ethical_AI_Reg #8)
        1. Implement friction calculation (Ethical_AI_Reg #9)

          1. Once Phase 1 is delivered, SAO agent registration (#4) and ethical forwarding (#6) can begin

          2. Ahead-of-Roadmap Progress

          3. While Phase 1 is blocked, Abigail and Orion_dock have advanced significantly beyond Phase 0:

            • Abigail: Hive/Entity two-daemon separation, tier-based model routing, shared chat engine library crate, tool-use loop with memory persistence, full skill loading and execution
              • Orion_dock: Native Rust Council DAG engine (replaced Pro sidecar), dual-proxy egress boundary, agent self-awareness API, 4 Genesis paths for birth, superego safety refactor

                • SAO: Full key vault platform with WebAuthn auth, encryption, React frontend, provisioning flow documentation

                • These advances are not gated by Phase 1 but will need to be aligned with the scoring engine once it is delivered.

                • Status Update Log

                • | Date | Cycle | Issue | Key Changes |

                • |------|-------|-------|-------------|

                • | Feb 24, 2026 | AM | #6 | Initial cross-repo status audit. Phase 1 blocked on Ethical_AI_Reg. |