Neuruh | Master-Codex-001 | Canonical Thread Dossier

Neuruh Local AI Estate: Master Canonical Codex

Recovery, custody, graph integration, productization, and controlled release architecture. A faithful standalone demo of the 13-page source codex dated 2026-08-01. This page reproduces its full section structure and makes its systems explorable. It is a demo built from the document, not the live estate registry itself.

DocumentMASTER-CODEX-001
Source date2026-08-01
OwnerJeramie Hicks | Neuruh LLC
StatusCanonical working reference
Source basisConversation thread, terminal transcripts, screenshots, DATUM Codex, FHUND operator manual, generated control-plane packages
Observed hub227 artifact links (source claim, time-specific)
Question askedWhere was the document and collection containing roughly 232 HTML sites and other local builds?
This demoBuilt 2026-11-04 | Single file | No live registry connection
Reading rule (from the source): factual statements in these documents are limited to the thread evidence. Proposed architecture, commercialization, and sequencing are explicitly framed as recommended operating decisions, not as completed disk-state proof.

Neuruh Hub | Registry explorer | Estate calculators | Try-it walkthrough | Share link

1. Mandate and executive finding

The thread began with a practical question: where was the document and collection containing the roughly 232 HTML sites and other local builds? The search exposed a larger reality. The local machine contains a significant estate of prototypes, engines, connectors, agent systems, harnesses, code packs, specifications, receipts, and deployment work that is not represented by a reliable registry or a controlled GitHub release pipeline.

Operator note (source): the core finding is operational. This is not a folder-cleanup project. It is a software-estate recovery, identity, custody, governance, and monetization program.

The observed ~/neuruh-live hub had 227 artifact links at the time of the terminal check (source claim). Those entries are primarily symbolic links to original files spread across multiple directories. The hub is a viewing layer, not a canonical repository, not a deployment catalog, and not proof that each entry is unique or publishable.

Executive outcome required

Count check: 227 observed links against roughly 232 sites named in the opening question. That is a gap of 5, or 97.8 percent, using the calculator in section 13. Both numbers are source claims, not verified by this demo.

2. Evidence and observed state

Evidence itemWhat it proves (source claim)
Terminal count: 227 linksThe local HTML hub is real and populated; the count is time-specific.
Symlink targetsAssets originate in Downloads, Desktop trees, workspace audit outputs, prompter, repository trees, and backups.
Permission-denied transcriptThe artifacts path is a directory; executing the directory was the wrong operation. cd or open is correct.
Port 8091 transcriptThe hub was already being served; starting a second server produced Address already in use.
Artifact inventoryThe collection mixes original builds, third-party saved pages, copies, research, prototypes, dashboards, and receipts.
ScreenshotsJARVIS, Field Pins, plat views, SPOT-CHECK, Plat-Lab, Continuity Cockpit, and Galaxy were not conceptual only; visual surfaces existed locally.
DATUM CodexA deeper deterministic/probabilistic valuation architecture exists beyond the visible calculator surface.
FHUND operator manualThe broader Neuruh lane/layer model, boot services, governance, receipts, and human-fire doctrine already exist.
Control Plane v0.1 / v0.2 packagesA read-only registry and post-census consolidation layer were generated and fixture-tested in the thread.
Control warning (source): the thread does not include the completed deep census result. Therefore it does not prove the final number of repositories, unique projects, duplicates, or ship-ready candidates. Those counts must come from the generated CSV and SQLite outputs after the scans finish.

3. The actual problem

3.1 Fragmentation

The estate is distributed across competing roots. Different roots may contain active repositories, copies, exports, generated builds, worktrees, backups, and stale forks. Modification time alone is not enough to determine the source of truth.

~/neuruh-repos~/neuruh-workspace/repos~/repos~/neuruh-vault~/Downloads~/Desktop and Desktop/Neuruh MasterMounted external volumesLocally mirrored cloud folders

3.2 Mixed asset classes

  • Git repositories and worktrees
  • Standalone source code and HTML prototypes
  • MCP servers, connectors, adapters, plugins, and skills
  • Agent orchestration and autonomous workflow systems
  • Harnesses, evaluations, benchmarks, governance, and receipts
  • RAG, Qdrant, memory, Obsidian, and knowledge systems
  • Local model and inference infrastructure
  • Deployment, Docker, Vercel, Supabase, and runtime assets
  • Prompts, specifications, Codexes, SOPs, and IP documents
  • Vertical systems: DeedSonar, LandOS, REI-X, DATUM, TaxShield, Bookeit, StudioGPT, resale and other wedges
  • ZIP/code packs and archives
  • Exact duplicates and renamed duplicates
  • Version families marked v1/v2/final/copy/numbered copy
  • Secret-risk, private-IP, defense, customer-data, and patent-sensitive material; third-party saved webpages and research captures

3.3 Canonicity failure

A filename such as final, latest, or v6 does not establish authority. A folder may be newer because it was copied, exported, or touched, while the real Git history and deployment origin live elsewhere. Canonical selection must use repository history, content hashes, manifests, tests, deployment references, and operator decisions.

3.4 Release failure

The estate has accumulated deployable or demonstrable work without a consistent path from local disk to governed release. The missing bridge is not raw coding capability. It is repeatable identity, packaging, decision, and publication infrastructure.

4. Estate taxonomy and topology

Eight axes. Use the filter to inspect one axis at a time. Every axis matters because it changes how an asset is inspected, packaged, released, or priced.

5. Existing Neuruh organs

The thread did not invent an entirely new organism. It identified that the organism already had many organs, but lacked a reliable estate registry between disk and execution.

GALAXYHealthPortsCOCKPITOrgansEventsJARVIS
Figure 1. Observed Galaxy/Cockpit surface: a dense local knowledge graph with health, port, organ, and event context. Diagram is a schematic of the source description, not a screenshot of the live graph.

6. The missing custody layer

The missing layer establishes identity and custody before an asset enters the graph, factory, or release process. Without it, the graph knows concepts but cannot reliably answer which exact file, repository, version, hash, or deployment is authoritative.

CUSTODY CHAIN: DISK TO RELEASEHard-drive asset > resolved path + content hash > canonical registry recordProduct / capability / role / version / risk / shipping state > append-only eventGalaxy node and edges > JARVIS retrieval > AXON proposed actionHuman approval > World Emitter package > NIMDP gateGitHub or deploymentIDENTITYHUMAN GATERELEASE
Figure 2. Custody chain from the source, section 6. Identity comes first. The human gate sits between proposal and release.

Responsibility separation

LayerQuestion answered
RegistryWhat exists, where is it, and what is its current custody state?
GraphHow is it related to products, capabilities, repositories, decisions, runs, and deployments?
LedgerWhat changed, when, by whom, and with what receipt?
OrchestratorWhat should happen next?
ExecutorWhat action was authorized and actually ran?
Release planeWhat became externally available and how is it monitored?

7. Canonical target architecture

CANONICAL TARGET ARCHITECTURELOCAL DISK + EXTERNAL VOLUMES + CLOUD MIRRORSREAD-ONLY ESTATE CENSUSARTIFACT CONTROL PLANE v0.2GALAXY / COCKPITJARVIS SEARCH + IARAXON HUMAN-GATED ACTIONWORLD EMITTER > NIMDP > GITHUB / DEPLOY / LICENSE > MONITORING
Figure 3. Canonical target architecture, section 7. The control plane sits after the census and before the graph. Nothing is published before the human gate and the NIMDP gate.

v0.1 to v0.2 evolution

v0.1 indexed the current HTML symlink hub into SQLite, CSV, version-family reports, events, and a graph feed. v0.2 correctly moved the control plane after the broader census: it merges every complete census run under ~/neuruh-estate-census/ with the HTML hub, deduplicates by resolved canonical path, retains the newest record for each path, and serves a combined project/artifact dashboard.

8. State, identity, and evidence model

8.1 Lifecycle

LIFECYCLEDISCOVERED > CLASSIFIED > CANONICALIZED > PACKAGED > TESTED > STAGEDSHIPPED > MONITORED > DEPRECATED
Figure 4. Lifecycle, section 8.1. Nine states, in order. Deprecated is a terminal state, not deletion.

8.2 Core node types (12)

project, artifact, product, capability, version_family, duplicate_group, repository, deployment, decision, run, receipt, source.

8.3 Core edge types (13)

belongs_to, implements, capability_of, version_of, contained_in, duplicate_member, supersedes, depends_on, generated_by, deployed_as, blocked_by, validated_by, produced.

8.4 Human gates (6)

  • Canonical-version selection
  • Secret review
  • Ownership and license review
  • Public/private repository decision
  • GitHub publication
  • Production deployment

8.5 Evidence rules

  • Resolved paths and content hashes identify file instances.
  • Git branch, HEAD, origin, and dirty state identify repository instances.
  • Missing files are marked broken; absence must never be interpreted as authorization to delete historical graph nodes.
  • Events are append-only. Existing event history is not rewritten.
  • A registry refresh may reclassify state but must preserve source evidence.
  • The graph may represent relationships; it must not silently move or delete files.

Registry explorer and completeness meter

Build a canonical registry record the way the codex requires it: identity first, then classification, then a human decision. Fill the fields by hand or load a sample. The meter counts the 12 required fields below.

Sample loaders

Three samples drawn from section 9. Each one is a source-described asset family, not a live registry row.

Canonical record

Completeness

0 of 12 fields complete (0 percent).

Missing fields:

    No record loaded yet. Load a sample or fill the form.

    Registry truth check

    A record is not canonical because its name says final or v6. It is canonical only when path, hash, repository evidence, and an operator decision agree.

    Share link

    Share encodes the current record into the URL hash as base64 JSON. Anyone opening the link restores the same record in this demo. It shares only what you entered here.

    9. Recovered asset families

    Fifteen families recovered in the thread. Filter by treatment or search by name. Canonical treatment is the source recommendation, not a completed action.

    10. Commercialization architecture

    The recovered estate can support multiple commercial forms. The business error would be forcing every artifact into a hosted SaaS product. Many tools can be valuable as static demonstrations, downloadable utilities, white-label components, or sales funnels.

    Commercial formBest fitOperating cost profilePrimary gate
    Free lead magnetStatic HTML calculators, diagnostics, samplesNear-zero if browser-only and no paid callsOwnership, claims, dependency audit
    Paid static downloadOffline calculators, reports, checklistsNear-zero marginal hosting costPackaging, license, support boundary
    Hosted micro-toolForm + API + stored resultPer-run API/storage costUnit economics, abuse control, monitoring
    White-label licenseTenant-branded CRM or intelligence surfacesSetup plus support; engine may remain centralizedBrand seam, data segregation, contract
    API/engineDATUM, valuation, scoring, evidence retrievalUsage-based compute/data costReproducibility, versioning, SLA
    Internal engine surfaceCore scoring, discovery, defense/private methodsNot sold directlyFirewall and private IP review
    Sales demo / proof artifactJARVIS, Galaxy, visual cortex, prototype screensLow; used to convert prospectsTruthful capability labeling
    Operator note (source): a static HTML file is not automatically zero-maintenance. External map tiles, paid APIs, model calls, storage, authentication, email capture, analytics, and support obligations create real cost. The control plane dependency audit exists to distinguish genuinely static assets from disguised hosted services.
    Select a form to see its best fit, cost profile, and gate.

    11. Governance constitution

    DoctrineBinding rule
    census-firstRead actual disk/repository state before any mutation.
    single-writerExactly one authorized writer per table or record domain.
    AI-never-deletesAutomated jobs append or overwrite their own outputs; reorganization is move-not-delete and operator-gated.
    human-fires-writesMint, merge, deploy, publish, and money-moving actions require an operator keystroke.
    firewallEngine IP, tenant data, credentials, defense work, and statutory/scoring methods do not leave the Mac without explicit review.
    subdivide-the-brandsNeuruh engine, platform/terminal products, and tenant brands remain separable.
    receipts-firstCore reports and proof depend on deterministic receipts, not optional LLM narrative.
    keychain doctrineSecrets migrate to macOS Keychain and are fetched at runtime; no plaintext secrets in published repositories.
    dual-key namingCanonical repository name plus pinned lane alias; every work item lands on the board with a stable identifier.
    revenue outranks buildRecovered assets move toward validated commercial release instead of endless reinvention.

    12. Execution program

    PhaseObjectiveExit condition
    P0 Freeze writesAllow census and registry operations only; no bulk moves or publication.Census processes complete; outputs intact.
    P1 ConsolidateRun v0.2 and merge complete census runs plus HTML hub.Canonical project/artifact registries generated.
    P2 CanonicalizeResolve exact duplicates, version families, competing roots, and repository lineage.Every priority asset has a declared canonical path and evidence.
    P3 ProtectRun secret, ownership, license, private-IP, customer, and defense reviews.No outbound candidate carries unresolved red flags.
    P4 PackageCreate clean repositories, READMEs, manifests, tests, screenshots, and sample data.Release candidates pass local packaging gates.
    P5 StageCreate private/public GitHub staging, preview deployments, and acceptance receipts.Operator can review real links and diffs.
    P6 ShipHuman-approved publication, release tags, deployment, and monitoring.External asset is live and registered back into Galaxy.
    P7 MonetizeConnect assets to funnels, sales, licenses, and usage measurement.Revenue/lead evidence is captured per asset.

    13. Measurement system and estate calculators

    All ten metrics from section 13, with the share math the source defines. A share is: part divided by total, times 100. Enter census numbers to compute the first four. The other six are defined here and scored by hand in the registry, because the source gives definitions, not numeric inputs.

    Coverage

    Share of eligible roots scanned and represented in registry.

    Uniqueness

    Unique content hashes versus duplicate instances.

    Canonicalization debt

    Version/duplicate families lacking an operator decision.

    Packaging readiness

    Candidates with README, license, source, tests, manifest, and screenshots.

    Hub count check

    227 observed links against roughly 232 sites named in the opening question.

    Other metrics (defined, not computed here)

    • Protection debt: secret, license, ownership, or private-IP holds unresolved.
    • Release throughput: assets moved from packaged to shipped per week.
    • Time to locate: median time to find the canonical asset and current status.
    • Graph fidelity: registry entities and state changes represented accurately in Galaxy.
    • Commercial activation: shipped assets with an active funnel, prospect, license, or usage signal.
    • Reuse ratio: products/tenants reusing canonical components rather than copied forks.

    14. Decision register

    DecisionStatusReason
    Treat neuruh-live as a viewing layer, not source of truth.LOCKEDEntries are symlinks to mixed origins.
    Use read-only census before any move or publication.LOCKEDDisk is truth; memory and names drift.
    Do not bulk-create GitHub repositories from raw folders.LOCKEDWould publish duplicates, secrets, vendor pages, generated output, and unclear lineage.
    Run v0.2 only after census processes finish.LOCKEDIt merges completed runs and refuses to race the census.
    Keep registry, graph, ledger, orchestrator, executor, and release responsibilities separate.LOCKEDPrevents another opaque monolith.
    Do not write directly into live Galaxy until its schema is read and mapped.LOCKEDGraph mutation without schema proof risks corruption.
    Keep JARVIS as the operator experience, not the authority layer.RECOMMENDEDRegistry and receipts hold truth; JARVIS retrieves and proposes.
    Select canonical APG prototype before deriving white-label editions.PENDING OPERATORMultiple versions exist and each may contain useful modules.
    Keep QFA experimental/private until validated.RECOMMENDEDVisual and conceptual value is not proof of scientific or production claims.

    15. Risk register

    RiskFailure modeMitigation
    Duplicate publicationMultiple versions appear as separate products.Hash grouping, version families, canonical decision receipt.
    Secret leakageAPI keys or local env files reach public GitHub.Path/content scanning, Keychain migration, explicit secret gate.
    IP leakageCore scoring or defense logic is published with a demo.Private-IP classification and component firewall.
    Vendor-content publicationSaved webpages are misrepresented as Neuruh software.Third-party reference classification and do-not-publish state.
    Generated-build publicationCompiled HTML/dist is shipped without source.Source/package manifest gate.
    Graph corruptionAutomated import deletes or overwrites historical nodes.Upsert by immutable IDs; never infer deletion from absence.
    False capability claimsA local frontend is described as a functioning backend.Separate observed UI state from verified runtime/API state.
    Cost blindnessA static tool calls paid APIs or maps.Dependency and per-run unit-economic audit.
    Operator overloadHundreds of discoveries become another unranked backlog.Priority scoring by revenue, uniqueness, maturity, and strategic reuse.
    Endless rebuildingNew versions are created instead of shipping canonical assets.Freeze version families and impose weekly release throughput.

    16. Final operating doctrine

    Operator note (source): the estate is not missing invention. It is missing identity, custody, packaging, and controlled distribution. The recovery program must convert scattered work into canonical assets, canonical assets into governed releases, and governed releases into measurable commercial channels.
    1. Finish the census.
    2. Consolidate with v0.2.
    3. Choose canonical paths and versions.
    4. Protect secrets and proprietary material.
    5. Package instead of copying.
    6. Test and stage.
    7. Human-authorize publication.
    8. Register every release back into the graph.
    9. Measure usage, leads, licenses, and revenue.
    10. Stop building replacement versions when a canonical asset can be shipped.

    Try-it walkthrough

    Five steps, in the order the codex requires. Use the buttons to move through them.

    Step 1: Census first

    Run a read-only census before any move. In this demo, load Sample B (DATUM) in the explorer above and note that its path and hash are placeholders, not census output. A real census would supply the resolved path and SHA-256 hash.

    Step 2: Classify on eight axes

    Use the taxonomy tabs in section 4. DATUM is a valuation capability, an engine role, and an API/engine commercial form. Classification is what makes the estate searchable by function instead of by folder name.

    Step 3: Check completeness

    Press Update completeness in the explorer. Sample B fills 10 of 12 fields. The missing fields are the operator decision and one classification gap, depending on what you edited. Nothing ships while required fields are missing.

    Step 4: Apply the human gates

    Check section 8.4. Canonical-version selection, secret review, ownership and license review, public/private decision, GitHub publication, and production deployment all require a human. JARVIS proposes. AXON proposes. The operator fires.

    Step 5: Share the record, then stage it

    Press Copy share link in the explorer. That link restores this exact record. In the real program the next step is P5 staging: a private or public GitHub staging area, a preview deployment, and an acceptance receipt, then registration back into Galaxy.

    Step 1 of 5

    How it works

    How this demo works

    • All 16 source sections are reproduced in order, with their tables, lists, and doctrine intact.
    • The registry explorer is a form, not a live database. It stores one record in the page and in the share hash.
    • Calculators use only the share math the source defines: part divided by total, times 100, rounded to one decimal.
    • Sample loaders fill the form from section 9 families. They are starting points for editing, not census results.
    • Share links use base64 JSON in the URL hash. Loading a shared link restores the record and recomputes completeness.

    What this is not

    • Not the live estate census. No disk was scanned to build this page.
    • Not the Artifact Control Plane v0.1 or v0.2. No SQLite, CSV, or graph feed is generated here.
    • Not a Galaxy or JARVIS backend. No search, retrieval, or AXON action runs here.
    • Not proof that any asset is unique, ship-ready, or publishable. Only a completed census plus operator decisions can prove that.
    • Not a release. Nothing on this page publishes a repository or deploys an asset.

    Honest gaps

    • No completed deep census in the source. The source states this directly. Final counts of repositories, unique projects, duplicates, and ship-ready candidates are unknown. This demo repeats that gap instead of filling it with an invented total.
    • 227 vs roughly 232. Both numbers are source claims from different parts of the thread. The 5-item gap is unexplained in the source and is not resolved here.
    • No DATUM math in the source. The DATUM Codex is cited as a deeper valuation architecture, but this PDF contains no DATUM equations, inputs, or worked values. No DATUM calculator is built here, because inventing one would misrepresent the source.
    • No live Galaxy schema. The source orders a read-only schema census before any graph write. This demo has no graph connection and makes no graph write.
    • Sample hashes are placeholders. Sample records use illustrative hash strings, not measured SHA-256 values from a real census.
    • APG canonical version is still pending operator decision in the source decision register. This demo does not choose it.
    • Layout not eyeballed in a real browser. Controls were tested in jsdom only, as specified for this build.
    • Source claims are labeled as source claims throughout. Counts, screenshots, and observed states are what the thread reported, not facts verified by this demo.

    Related: Neuruh Hub (master map of Neuruh sites). Sibling codex and system demos are listed on the hub. No other sibling URL is asserted here, because the source PDF does not name live sibling demo URLs.

    Source: NEURUH LOCAL AI ESTATE - MASTER CANONICAL CODEX, MASTER-CODEX-001, 2026-08-01, 13 pages. Owner: Jeramie Hicks | Neuruh LLC. Status: Canonical working reference. Private operator working document.