initial priming commit

canon/projects/coulomb.social/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-COUL-2026-000001
+type: concept
+title: "Coulomb.social — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Coulomb.social"]
+  sensitivity: internal
+tags: ["concepts", "marketplace", "governance"]
+---
+
+# Coulomb.social — Seed Concepts v0.1
+
+1. **Challenge** — a request for progress with explicit acceptance criteria.
+2. **Deliverable** — the artifact produced in response to a challenge.
+3. **Obligation** — a rule-bound commitment; violations have consequences.
+4. **Downgrade** — reduction of rights/permissions after violations.
+5. **Jurisdiction** — a portable rule-set defining rights/obligations and enforcement.
+6. **Reputation** — aggregated reliability signal derived from traceable outcomes.
+7. **Arbitration** — dispute resolution process under jurisdiction rules.
+8. **Agent Participation** — non-human actors operating under scoped permissions.
+9. **Audit Trail** — non-repudiable record supporting trust.
+10. **Economic Stabilization** — constraints ensuring the platform can sustain itself.

canon/projects/coulomb.social/project_charter_v0.1.md

@@ -0,0 +1,39 @@
+---
+id: CUST-PRJ-COUL-2026-000001
+type: charter
+title: "Coulomb.social — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Coulomb.social"]
+  sensitivity: internal
+tags: ["cocreation", "marketplace", "governance", "agents"]
+---
+
+# Coulomb.social — Project Charter v0.1
+
+## Purpose
+Build a co-creation platform and socio-economic experiment that enables humans and AI agents to collaborate on challenges with clear rules, incentives, and evolving governance.
+
+## Problem
+Current collaboration systems either:
+- centralize control (slow, bureaucratic), or
+- decentralize without governance (chaotic, unsafe).
+Mixed-intelligence societies require explicit interaction rules.
+
+## Outcome
+A platform where:
+- challenges and solutions are exchanged with clear obligations
+- trust and reputation emerge from traceable behavior
+- rulesets can evolve and be ported across spaces (“jurisdictions”)
+- agents can participate under bounded permissions
+
+## Boundaries (v0.1)
+- Prove core loop before scaling: challenge → work → deliverable → acceptance → reputation.
+- Governance is a first-class product feature, not a later add-on.
+
+## Success criteria (v0.1)
+- A small number of users (and agents) can co-create reliably with low coordination overhead.
+- Rule enforcement and auditability prevent obvious failure modes.

canon/projects/coulomb.social/roadmap_v0.1.md

@@ -0,0 +1,36 @@
+---
+id: CUST-RDM-COUL-2026-000001
+type: roadmap
+title: "Coulomb.social — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Coulomb.social"]
+  sensitivity: internal
+tags: ["roadmap", "governance", "market", "reputation"]
+---
+
+# Coulomb.social — Roadmap v0.1
+
+## Phase 0 — Core Loop Prototype
+- Challenge posting + bounty/terms
+- Deliverable submission
+- Acceptance/decline with rationale
+- Reputation update and audit trail
+
+## Phase 1 — Jurisdiction Engine
+- Portable rule packs
+- Rights/obligations encoded as enforceable policies
+- Violation → downgrade mechanisms
+
+## Phase 2 — Agent Participation
+- Agent identities + scoped permissions
+- Tool boundary enforcement
+- Agent reputation and accountability
+
+## Phase 3 — Economic Stabilization
+- Pricing and sustainability model
+- Anti-spam and anti-fraud controls
+- Governance for disputes and arbitration

canon/projects/custodian/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-CUST-2026-000001
+type: concept
+title: "Custodian — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Custodian"]
+  sensitivity: internal
+tags: ["concepts", "agent", "continuity", "security"]
+---
+
+# Custodian — Seed Concepts v0.1
+
+1. **Identity Substrate** — the canon + episodic archive that preserves continuity across model swaps.
+2. **Bounded Agency** — doing useful work inside enforced limits and escalation rules.
+3. **Promotion Workflow** — controlled entry of knowledge into canon via gates and approval.
+4. **Sensitivity Labeling** — enforced classification (public/internal/confidential/sealed).
+5. **Memory Poisoning Defense** — preventing untrusted inputs from becoming “truth” in canon.
+6. **Tool Boundary Enforcement** — permissions enforced where actions happen (not only in prompts).
+7. **Continuity Drill** — test: reconstruct state from canon after downtime or model change.
+8. **Canon Release** — signed snapshot enabling rollback and long-term stability.
+9. **Sovereign Operation** — the system remains functional without external dependencies.
+10. **Transgenerational Custody** — long-timescale stewardship with governance, consent, and succession.

canon/projects/custodian/full_circle_map_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-MAP-2026-000001
+type: concept
+title: "Full Circle Map: Railiance → Markitect → Coulomb.social → Personhood/Foerster → Custodian"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Custodian"]
+  sensitivity: internal
+tags: ["strategy", "map", "dependencies"]
+---
+
+# Full Circle Map v0.1
+
+## Dependency Chain
+- **Railiance** provides: sovereign ops, backups, audits, security primitives.
+- **Markitect** provides: artifact grammar, canon, provenance, retrieval, curation workflow.
+- **Coulomb.social** provides: interaction space, governance engine, economics, reputation.
+- **Foerster Capabilities** provides: capability taxonomy for agency/governability design.
+- **Personhood** provides: rights/obligations/downgrade frameworks for mixed intelligences.
+- **Custodian** integrates all layers into a long-lived co-creative institution.
+
+## Priority Principle
+Stabilize economically and operationally via robust targeted information processing first; expand into transgenerational/family custodianship step by step.

canon/projects/custodian/project_charter_v0.1.md

@@ -0,0 +1,39 @@
+---
+id: CUST-PRJ-CUST-2026-000001
+type: charter
+title: "Custodian — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Custodian"]
+  sensitivity: internal
+tags: ["agent", "canon", "continuity", "sovereignty", "family"]
+---
+
+# Custodian — Project Charter v0.1
+
+## Purpose
+Build a sovereign, secure, co-creative agent system that supports Bernd’s projects and evolves toward transgenerational custodianship for a biological family.
+
+## Problem
+Founder-centric initiative fails under absence. Provider-dependent AI services threaten continuity, sovereignty, and privacy.
+
+## Outcome
+A locally operable “institution in software” consisting of:
+- canon (identity substrate)
+- memory tiers (working, episodic, vault)
+- governance (constitution, roles, permissions)
+- bounded agency (safe tool surface + escalation rules)
+- reliable infrastructure (Railiance)
+
+## Boundaries (v0.1)
+- Co-creation and stewardship for projects only.
+- Family functionality is future scope; build prerequisites now (security, governance, portability).
+- No autonomous irreversible actions.
+
+## Success criteria (v0.1)
+- Custodian can co-create daily outputs across the project set.
+- Custodian can reconstruct project states from canon.
+- Canon is signed, backed up, and portable.

canon/projects/custodian/roadmap_v0.1.md

@@ -0,0 +1,38 @@
+---
+id: CUST-RDM-CUST-2026-000001
+type: roadmap
+title: "Custodian — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Custodian"]
+  sensitivity: internal
+tags: ["roadmap", "agent", "governance", "infrastructure"]
+---
+
+# Custodian — Roadmap v0.1
+
+## Phase 0 — Canon + Constitution (now)
+- Canon schemas and initial project artifacts
+- Promotion workflow and quality gates
+- Audit log and sensitivity labels
+
+## Phase 1 — Co-creation Productivity
+- RAG over canon + approved repos
+- Drafting pipelines (charters, ADRs, roadmaps, specs)
+- Consistency checks across artifacts
+
+## Phase 2 — Stewardship Automation
+- Scheduled “health checks”
+- Roadmap drift detection
+- Open question aging and review prompts
+
+## Phase 3 — Sovereign Appliance
+- Local inference and indexing on dedicated hardware (DGX Spark or equivalent)
+- Offline mode, replication, disaster recovery
+
+## Phase 4 — Family Custodianship Prerequisites
+- Consent model, retention rules, vault hardening
+- Multi-role governance and succession protocol

canon/projects/foerster-capabilities/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-FOER-2026-000001
+type: concept
+title: "Foerster Capabilities — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Foerster Capabilities"]
+  sensitivity: internal
+tags: ["concepts", "cybernetics", "capabilities"]
+---
+
+# Foerster Capabilities — Seed Concepts v0.1
+
+1. **Non-Trivial Machine** — a system whose output depends on internal state/history, not only current input.
+2. **Governability** — degree to which an environment/controller can constrain and steer the system.
+3. **Reliability** — stability of behavior under expected conditions; predictability of failure modes.
+4. **Adaptivity** — capacity to adjust behavior in response to changing environment and feedback.
+5. **Reflexivity** — capacity to model/self-reference its own behavior and update based on that model.
+6. **Autonomy** — ability to set and pursue goals with limited external control.
+7. **Sovereignty** — autonomy plus control over critical dependencies (resources, memory, tooling).
+8. **Capability Vector** — multi-dimensional description placing a system in capability space.
+9. **Interaction Envelope** — boundary conditions within which safe, intended interaction holds.
+10. **Control Surface** — set of levers available to govern a system (policies, tools, incentives).

canon/projects/foerster-capabilities/project_charter_v0.1.md

@@ -0,0 +1,36 @@
+---
+id: CUST-PRJ-FOER-2026-000001
+type: charter
+title: "Foerster Capabilities — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Foerster Capabilities"]
+  sensitivity: internal
+tags: ["taxonomy", "systems-theory", "agency", "governance"]
+---
+
+# Foerster Capabilities — Project Charter v0.1
+
+## Purpose
+Create a systems-theoretic taxonomy of computational agency grounded in Foerster’s Non-Trivial Machines, integrating behavior, resources, interaction properties, and governability.
+
+## Problem
+Existing capability classifications are fragmented (complexity, architecture, AI tasks, cybernetics, governance/risk). We lack a shared multidimensional map.
+
+## Outcome
+A rigorous capability space:
+- dimensions such as governability, reliability, adaptivity, reflexivity, autonomy/sovereignty
+- operational tests / indicators where possible
+- mappings from systems (agents/services/orgs) into the space
+- connections to governance and risk controls
+
+## Boundaries (v0.1)
+- Establish vocabulary, dimension definitions, and orthogonality discussions.
+- Produce diagrams and examples rather than chasing completeness.
+
+## Success criteria (v0.1)
+- A paper-grade outline + a stable diagrammatic “map”.
+- A small library of example classifications (LLM agent, company process, service).

canon/projects/foerster-capabilities/roadmap_v0.1.md

@@ -0,0 +1,33 @@
+---
+id: CUST-RDM-FOER-2026-000001
+type: roadmap
+title: "Foerster Capabilities — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Foerster Capabilities"]
+  sensitivity: internal
+tags: ["roadmap", "taxonomy", "paper", "dimensions"]
+---
+
+# Foerster Capabilities — Roadmap v0.1
+
+## Phase 0 — Dimension Definitions
+- Define each dimension with: intrinsic/extrinsic aspects, tests, failure modes
+- Resolve overlaps and dependencies (orthogonality work)
+
+## Phase 1 — Map & Classification Protocol
+- Capability space diagram(s)
+- Classification checklist and rubric
+- Example classifications
+
+## Phase 2 — Governance Mapping
+- Link capabilities to governance controls and risk envelopes
+- “Permission design” implications for agents
+
+## Phase 3 — Publication Package
+- Preprint structure
+- Glossary and diagrams
+- Reference implementation of classification tool (optional)

canon/projects/markitect/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-MARK-2026-000001
+type: concept
+title: "Markitect — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Markitect"]
+  sensitivity: internal
+tags: ["concepts", "knowledge", "schema"]
+---
+
+# Markitect — Seed Concepts v0.1
+
+1. **Artifact** — a unit of structured knowledge with metadata and lifecycle.
+2. **Provenance** — where an artifact came from (sources + context).
+3. **Canon** — curated, stable knowledge substrate; identity and continuity live here.
+4. **Working Memory** — operational notes; not identity-grade unless promoted.
+5. **Episodic Archive** — immutable log of events and interactions.
+6. **Promotion** — controlled movement from working memory into canon.
+7. **Supersession** — explicit replacement of an artifact while retaining history.
+8. **Relationship Graph** — explicit linking of artifacts for navigation and reasoning.
+9. **Canon Release** — signed snapshot of canonical state.
+10. **Retrieval Explainability** — ability to show why a piece of canon was retrieved.

canon/projects/markitect/project_charter_v0.1.md

@@ -0,0 +1,36 @@
+---
+id: CUST-PRJ-MARK-2026-000001
+type: charter
+title: "Markitect — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Markitect"]
+  sensitivity: internal
+tags: ["information-management", "canon", "schemas", "knowledge"]
+---
+
+# Markitect — Project Charter v0.1
+
+## Purpose
+Build an efficient information management platform that turns messy inputs into structured, versioned, reusable knowledge artifacts.
+
+## Problem
+Projects accumulate fragmented knowledge across chats, notes, docs, tickets, and repos. Without a stable schema and curation workflow, institutional memory fails.
+
+## Outcome
+A system for:
+- capturing artifacts (chunks, concepts, decisions, patterns)
+- connecting them via explicit relationships
+- producing “canon releases” and reliable retrieval
+- enabling co-creation without confusion
+
+## Boundaries (v0.1)
+- Prioritize core primitives: artifact schema, linking, provenance, versioning.
+- Avoid UI complexity until the model is stable.
+
+## Success criteria (v0.1)
+- Canon artifacts can be generated, reviewed, merged, and released with integrity.
+- Retrieval is precise enough to support co-creation work daily.

canon/projects/markitect/roadmap_v0.1.md

@@ -0,0 +1,34 @@
+---
+id: CUST-RDM-MARK-2026-000001
+type: roadmap
+title: "Markitect — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Markitect"]
+  sensitivity: internal
+tags: ["roadmap", "schema", "provenance", "retrieval"]
+---
+
+# Markitect — Roadmap v0.1
+
+## Phase 0 — Artifact Grammar
+- Core types: charter, concept, decision, procedure, roadmap, question, value
+- Relationship fields and lifecycle states
+- Provenance standard
+
+## Phase 1 — Curation Workflow
+- Draft → review → merge → release
+- Quality gates and checklists
+- Deprecation/supersession mechanism
+
+## Phase 2 — Retrieval Layer
+- Indexing pipeline (text + metadata)
+- RAG patterns optimized for canon integrity
+- “Explain why retrieved” provenance output
+
+## Phase 3 — Multi-space Support
+- Multiple domains/spaces with shared primitives
+- Portable rulesets across spaces (proto “jurisdiction” pattern)

canon/projects/personhood/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-PERS-2026-000001
+type: concept
+title: "Personhood — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Personhood"]
+  sensitivity: internal
+tags: ["concepts", "law", "mixed-intelligence"]
+---
+
+# Personhood — Seed Concepts v0.1
+
+1. **Agent** — any entity capable of perception, processing, and action in an environment.
+2. **Personhood Vector** — multidimensional profile describing capacities relevant to rights/obligations.
+3. **Legal Personality vs Capacity** — being a subject of law vs being competent to exercise rights/duties.
+4. **Rights Module** — a portable bundle of permissions/entitlements.
+5. **Obligations Module** — a portable bundle of duties and constraints.
+6. **Violation** — breach of obligations triggering consequences under jurisdiction rules.
+7. **Downgrade** — reduction of rights/capabilities granted after violations.
+8. **Restoration Path** — process by which rights can be regained after demonstrated compliance.
+9. **Representation/Guardianship** — delegation model for entities lacking capacity in some dimensions.
+10. **Jurisdiction Pack** — implementable ruleset binding an interaction space.

canon/projects/personhood/project_charter_v0.1.md

@@ -0,0 +1,35 @@
+---
+id: CUST-PRJ-PERS-2026-000001
+type: charter
+title: "Personhood — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Personhood"]
+  sensitivity: internal
+tags: ["law", "rights", "obligations", "agents", "jurisdictions"]
+---
+
+# Personhood — Project Charter v0.1
+
+## Purpose
+Develop a multidimensional, testable framework of personhood for mixed-intelligence societies, enabling experimental legal rule systems with tiered rights/obligations and downgrade mechanisms.
+
+## Problem
+Binary personhood breaks under heterogeneous agents (humans, AIs, orgs, hybrids). We need portable terminology, tests, and rule modules that jurisdictions can adopt and evolve.
+
+## Outcome
+- a dimensional personhood space (capability-based, substrate-independent)
+- a modular rights/obligations library
+- violation → downgrade logic (enforceable within platforms)
+- portability across “interaction spaces” (marketplaces, communities)
+
+## Boundaries (v0.1)
+- Focus on conceptual and operational primitives, not a single “ideal law”.
+- Emphasize clarity, testability, and portability.
+
+## Success criteria (v0.1)
+- A coherent framework that can be implemented as platform governance.
+- Example jurisdictions and rule packs for specific interaction spaces.

canon/projects/personhood/roadmap_v0.1.md

@@ -0,0 +1,32 @@
+---
+id: CUST-RDM-PERS-2026-000001
+type: roadmap
+title: "Personhood — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Personhood"]
+  sensitivity: internal
+tags: ["roadmap", "rights", "obligations", "tests"]
+---
+
+# Personhood — Roadmap v0.1
+
+## Phase 0 — Dimensions + Levels
+- Define dimensions (capacity vectors) and levels (tiers)
+- Specify tests/indicators and ambiguity handling
+
+## Phase 1 — Rights/Obligations Modules
+- Modular catalog of rights and obligations
+- Mapping from capability/profile → applicable modules
+
+## Phase 2 — Violation & Downgrade Protocol
+- What constitutes violations
+- How downgrades are applied, appealed, and restored
+
+## Phase 3 — Jurisdiction Portability
+- Rule-pack format
+- Transfer/translation between contexts
+- Example “interaction spaces” (marketplace, co-creation lab, family archive)

canon/projects/railiance/concepts_seed_v0.1.md

@@ -0,0 +1,26 @@
+---
+id: CUST-CPT-RAIL-2026-000001
+type: concept
+title: "Railiance — Seed Concepts v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Railiance"]
+  sensitivity: internal
+tags: ["concepts", "reliability", "ops"]
+---
+
+# Railiance — Seed Concepts v0.1
+
+1. **Sovereign Ops** — operations that remain functional without external provider control.
+2. **Restore Drill** — a periodic exercise proving the system can be recovered from backups.
+3. **Reproducible Deployment** — same inputs produce the same deployed system.
+4. **Policy at the Tool Boundary** — enforce permissions where the agent calls tools.
+5. **Secrets Hygiene** — secrets never enter plaintext logs, canon, or chat transcripts.
+6. **Immutable Audit Log** — append-only record of actions and changes.
+7. **Degrade Gracefully** — partial functionality under constrained resources/connectivity.
+8. **Operational Minimalism** — minimize moving parts and required human attention.
+9. **Continuous Verification** — automated checks that detect drift and breakage early.
+10. **Version Pinning** — explicit version control of runtime dependencies to avoid surprise changes.

canon/projects/railiance/project_charter_v0.1.md

@@ -0,0 +1,39 @@
+---
+id: CUST-PRJ-RAIL-2026-000001
+type: charter
+title: "Railiance — Project Charter v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Railiance"]
+  sensitivity: internal
+tags: ["devops", "reliability", "automation", "sovereignty"]
+---
+
+# Railiance — Project Charter v0.1
+
+## Purpose
+Build a robust, automated DevOps and operational reliability foundation that supports long-lived projects under sovereignty constraints (local-first, provider-independent).
+
+## Problem
+Modern stacks become brittle due to vendor coupling, hidden dependencies, and operational complexity. Founder-driven operations do not scale and do not survive absences.
+
+## Outcome
+A repeatable operational substrate:
+- reproducible deployments
+- automated backups and recovery drills
+- integrity checks and audit trails
+- secure secret management
+- portable infrastructure templates
+
+## Boundaries (v0.1)
+- Focus on pragmatic reliability engineering and automation.
+- Prefer boring, proven primitives over cleverness.
+- No “platform for everyone” ambition until internal stability is proven.
+
+## Success criteria (v0.1)
+- One-click (or scripted) deploy + restore for the Custodian stack.
+- Routine evidence: logs, checks, and recovery tests.
+- Clear runbooks and minimal operational burden.

canon/projects/railiance/roadmap_v0.1.md

@@ -0,0 +1,36 @@
+---
+id: CUST-RDM-RAIL-2026-000001
+type: roadmap
+title: "Railiance — Roadmap v0.1"
+status: active
+owners: ["Bernd", "Custodian"]
+created: "2026-02-24"
+updated: "2026-02-24"
+scope:
+  domains: ["Railiance"]
+  sensitivity: internal
+tags: ["roadmap", "devops", "security", "continuity"]
+---
+
+# Railiance — Roadmap v0.1
+
+## Phase 0 — Baseline (now)
+- Repo layout + IaC baseline
+- Logging + metrics + alerting minimal set
+- Backup strategy + encrypted snapshots
+- Restore drill (documented)
+
+## Phase 1 — Custodian as a Service (internal)
+- Containerized runtime + pinned versions
+- Automated indexing and scheduled maintenance jobs
+- Key management + secret rotation protocol
+- Immutable audit log storage
+
+## Phase 2 — Hardening
+- Threat model + prompt-injection controls for memory
+- Policy enforcement at tool boundary
+- Canary tests and regression suite automation
+
+## Phase 3 — Portability
+- Hardware migration playbook (workstation → DGX Spark → other)
+- Offline mode guarantee + sync strategy