Neuruh Organism Codex OS · Master Document · Compiled July 15, 2026
Neuruh is Sovereign Opportunity Intelligence and Execution Infrastructure, implemented as a Modulated Agentic Opportunity Organism. It senses fragmented signals across property, businesses, markets, and physical assets, converts them into evidence-backed opportunities, and deploys specialized systems to capture the value. This board renders the master codex as visuals: the operating model, the real-estate organ pack, the contracts, the N-Cell topology protocol, and the governance plane that holds all of it.
Do not build isolated tools. Build organs that plug into one organism.
The organism owns memory, evidence, governance, orchestration, and output. Products own domain workflows and user-facing experiences.
Every claim traceable to evidence. Every calculation deterministic. Every external action passes governance.
The organism at a glance
Products are sensors, organs, recipes, interfaces, or emitters on one shared substrate. Nothing runs outside the loop: every execution returns evidence and outcomes to memory, so the whole system compounds instead of resetting.
Section 2 · Operating model
The model separates core intelligence infrastructure from domain applications, so no product ships its own user table, data layer, prompt logic, reporting format, and half-baked memory system. Select a layer to see its responsibility and how it shows up in two different products.
Section 4 · Product classification
A sensor, an organ, a recipe, an interface, or an emitter. Some products combine roles, but the boundaries stay clean, and that is what lets hundreds of prompts, tools, agents, and apps cooperate without losing provenance or control.
Finds signals in the world.
Performs specialist reasoning or calculation.
Coordinates several organs toward one outcome.
Lets a human operate the organism.
Packages the result for the outside world.
Not one prompt. The contract layer.
The real moat is the contract layer that lets the whole portfolio cooperate: shared identity, evidence, memory, orchestration, governance, and reports underneath every product.
What the customer sees and operates.
What every product shares and none of them rebuild.
Section 4 · Product and service layer map
Each product classified by its primary role and how it plugs into the organism. Filter by role.
Section 5 · Real Estate Organ Pack
The domain package that turns the general organism into a deal acquisition engine: sixteen underwriting organs plus the shared comp, repair, market, risk, and offer infrastructure every one of them draws on.
Section 3 · APG Land CRM on the organism
APG Land CRM is the real-estate cockpit, not the brain. It captures leads and displays deal state; the heavy intelligence lives beneath it in the organs. Step through a single deal's path.
Section 6 · Universal DealOS core
DealOS is the shared real-world opportunity pipeline: real estate first, then resale, tax, bookings, content, and other verticals. Every opportunity moves through the same twelve states, and every state transition is an AXON event.
Deduplicates people, companies, properties, assets, and opportunities. Canonical IDs, merge review, aliases, conflicting identifiers.
Normalizes raw inputs into opportunities and evidence claims: forms, CSV, email, attachments, photos, notes, transcripts.
Stores raw evidence and normalized facts with provenance. Immutable raw payloads, field-level evidence, conflict ledger.
Chooses which organs should run, with a multi-organ routing explanation and confidence.
Coordinates events, retries, pauses, approvals, and receipts on the AXON event bus. Idempotent tasks, run correlation.
Executes deterministic formulas: versioned formulas, decimal math, trace outputs.
Composes exit strategies and alternatives, compared on profit, cash, risk, timeline, and confidence.
Approves, blocks, or escalates actions. Policy-as-code, evidence thresholds, audit log.
Builds outputs from approved results only: PDF, DOCX, CRM note, investor packet, API JSON.
Tests engines against fixtures and expected outputs: formula tests, schema validation, regression cases.
Sections 7–8 · Data contracts and AXON events
Prompts are flexible. Production systems are not. Seven core contracts are shared by every product; eleven real-estate contracts extend them. Every fact references evidence or is explicitly marked as an assumption.
{
"evidence_id": "ev_01JREAL_0001",
"domain": "real_estate",
"source_type": "county_assessor",
"source_name": "St. Joseph County Assessor",
"source_url": "https://example.gov/property/12345",
"retrieved_at": "2026-07-15T15:10:00-04:00",
"raw_storage_uri": "s3://neuruh-evidence/raw/ev_01JREAL_0001.json",
"facts": [
{ "field": "bedrooms", "value": 2, "confidence": 0.92 },
{ "field": "bathrooms", "value": 1, "confidence": 0.91 },
{ "field": "living_area_sf", "value": 984, "confidence": 0.88 }
],
"conflicts": [],
"license_policy": "internal_use_only"
}
{
"event_id": "evt_01JREAL_0001",
"event_type": "realestate.underwriting.completed",
"version": "1.0.0",
"occurred_at": "2026-07-15T15:18:00-04:00",
"correlation_id": "deal_935_s_28th_st",
"run_id": "run_01JREAL_0007",
"actor": { "type": "organ", "id": "residential_arv:v2.0.0" },
"payload_ref": "s3://neuruh-events/payloads/evt_01JREAL_0001.json",
"idempotency_key": "deal_935_s_28th_st:residential_arv:v2.0.0"
}
Engines do not call each other directly outside a local execution context. Events keep the organism replayable, auditable, and upgradeable.
| Namespace | Example events | Owner |
|---|---|---|
| core.identity | identity.created · identity.merged · identity.conflict_detected | Identity Service |
| evidence | evidence.attached · evidence.normalized · evidence.conflict_detected | Evidence Service |
| realestate.deal | received · normalized · staged · dispositioned · outcome_recorded | APG CRM / DealOS |
| realestate.comp | search_requested · candidate_found · candidate_rejected · set_approved | Comparable Intelligence |
| realestate.market | snapshot_requested · snapshot_generated · heat_score_updated | Market Velocity |
| realestate.repair | photos_received · scope_generated · inspection_flagged | Photo Repair Organ |
| realestate.underwriting | requested · completed · failed · superseded | Engine Orchestrator |
| realestate.offer | proposed · approved · rejected · sent · accepted · expired | Offer Architecture + Governance |
| world.emit | report_requested · report_emitted · crm_note_created · packet_sent | World Emitter |
| governance | review_required · approved · blocked · escalated · policy_updated | Governance Core |
N-Cell position paper · Governed Adaptive Topology Protocol
At any moment the organism is a directed, attributed graph: G_t = (V, E_t, X_t, P_t), the registry, the active topology, the state, and the governance constraints. Each node carries an eight-bit routing mask over its Moore neighborhood. Toggle the neighbors of node v to build its route mask.
R_v(t) = Σi=0..7 b_i(t) · 2i, b_i ∈ {0,1} → 256 local activation patterns
The eight bits may represent spatial directions, logical neighbor classes, event channels, or registered capability edges. The mapping must be declared in a versioned schema. A practical organism uses multiple masks rather than forcing all meaning into one byte, which is what the N-Cell/64 header does:
Permitted local event routes or neighbor relationships.
Active specialist organs or capability families.
Workflow stage and completion state.
Observe, recommend, draft, stage, execute, emergency block.
Permitted working, episodic, semantic, procedural, prospective, founder, and outcome planes.
Quantized evidence or decision confidence, 0–255.
Quantized escalation or hazard score, 0–255.
Permitted artifacts, messages, APIs, physical commands, or World emission state.
E*_t = E_proposed(t) ∩ E_allowed(t) ∩ E_available(t)
Codex V2 · The modulated agentic system
A conventional multi-agent system assigns fixed roles and fixed workflows. Neuruh modulates all twelve of these variables per opportunity, while identity, contracts, memory, evidence requirements, governance, and operating doctrine stay fixed.
Where the run executes and what infrastructure is actually available.
The class of value being pursued: transition, land, business, asset, IP.
Which registered agents compose the temporary topology.
Which model serves which node, routed by task and cost.
Which Vault planes a node may read, and nothing beyond them.
Which tools and adapters are bound to the run.
How nodes communicate: the N-Cell graph for this run.
Observe → recommend → draft → stage → execute → emergency block.
How much hazard a run may carry before it must escalate.
Budget, cost envelope, and expected value of the opportunity.
Which artifacts the run may emit, and to whom.
The harness and thresholds the result is scored against.
Codex V3 · Full Notion audit
The audit exposed the backbone the earlier codex under-described: a revenue spine that turns signals into outcomes, a knowledge spine that turns canon and evidence into repeatable intelligence, the formal t0–t6 lanes, and the Galaxy-to-Product deployment model.
Signals become outcomes, and outcomes reweight the system.
Canon becomes retrievable intelligence, and receipts flow back into canon.
| Lane | Role | Status at the July 15, 2026 build session |
|---|---|---|
| t0 Operator | Human fire: approvals and irreversible actions | OPERATOR-DEPENDENT |
| t1 Engine | LandOS engine + ingest | DOWN LOCALLY demo fixtures carried the visual |
| t2 Vault | Vault Galaxy substrate | PARTIAL git vault present; Galaxy Live UI down |
| t3 AEGIS | IAR decision and model-routing plane + governance | LIVE :8090 |
| t4 Emitter | World Emitter | LIVE :4174 |
| t5 CRM | APG Land CRM cockpit surface | COCKPIT |
| t6 Cockpit | Neuruh Cockpit + AXON execution plane | DEGRADED AXON daemon healthy; gateway degraded with Bonsai offline |
AEGIS/IAR is the decision plane. AXON is the execution plane. MOTHER is health and clock, built but not wired as a continuous supervisor. The Governance Core is an allow / escalate / block evaluator behind a gated boundary.
Section 10 · Immune system
The immune system protects the organism from unsupported conclusions, legal risk, financial overreach, and operational chaos. Policy-driven and auditable: every policy area defines what is blocked automatically, what escalates to a human, and what proceeds with a visible warning.
| Policy area | Automatic block | Escalation | Allowed with warning |
|---|---|---|---|
| Evidence sufficiency | BlockARV report with fewer than three acceptable comps and no documented fallback. | EscalateSparse data but a transparent confidence score. | WarnThree comps present, confidence below target. |
| Seller-facing offers | BlockOwnership unverified or seller authority unclear. | EscalateConflicting seller notes or title flags. | WarnVerified seller, approved offer, sufficient confidence. |
| Creative finance | BlockNo loan details, arrears, or seller-consent path. | EscalateDue-on-sale or legal structure unclear. | WarnTerms modeled with an attorney-review flag. |
| Zoning and development | BlockBuildability claim without zoning or source evidence. | EscalateAmbiguous ordinance or conflicting GIS layers. | WarnConceptual HBU marked non-binding. |
| Tax and legal | BlockDirect legal or tax conclusion without professional review. | EscalateEntity, lien, probate, or tax exposure involved. | WarnEducational estimate with disclaimer and review path. |
| Data licensing | BlockRestricted provider data used in a public report. | EscalateReuse rights unclear. | WarnInternal-only source properly marked. |
Risk categories under watch: title, lien, probate, ownership, authority, tax delinquency, foreclosure, and redemption; zoning, land-use, floodplain, wetland, environmental, utility, access, and easement; structural, roof, foundation, mechanical, occupancy, tenant, code, and permit; market liquidity, rental demand, investor appetite, buyer financeability, and disposition; data quality, conflicting facts, stale evidence, missing comps, and low-confidence model output.
Section 9 · Memory, learning, and the Vault
The Vault is not a flat chat log. It holds typed memory with permissions, decay rules, and an outcome calibration loop, so the organism develops continuity instead of amnesia.
Specific runs and deal histories.
“The 935 S 28th St underwriting run and its final disposition.”
Durable knowledge and market facts.
“River Park 2/1 homes lease within a supported band, given the data.”
How to perform a workflow.
“The residential ARV comp-selection relaxation order.”
Future follow-ups and pending actions.
“Re-check the inspection report when the seller uploads it.”
Current active objectives.
“Build Real Estate Organ Pack V1 integration.”
Stable preferences and templates.
“APG visual language, report formatting, acquisition thresholds.”
Store the original prediction with its formulas, assumptions, evidence, and model version.
Record the actual outcome: purchase price, repair cost, resale or rent, DOM, profit, timeline.
Calculate the variance between predicted and actual values.
Tag the variance source: bad evidence, market shift, repair miss, comp mistake, or execution issue.
Calibrate only through governed rule updates. Never silent prompt drift.
N-Cell paper · Prior art and novelty boundaries
The paper's own discipline, applied to itself: the architecture is real enough to formalize and test, not yet proven enough to market as AGI. Microsoft, OpenAI, Anthropic, LangChain, Palantir, ServiceNow, Salesforce, and the dynamic-topology research line already cover many of the primitives. The defensible ground is the integration.
| Claim | Classification | Defensibility |
|---|---|---|
| Eight-neighbor lattice and bitmask routing | Established prior art | Do not claim as invention. |
| Dynamic agent-team and communication topology | Established, rapidly advancing | Use as foundation; differentiate implementation and objective function. |
| Agent operating system | Established research category | Do not claim category ownership. |
| Governance intersection: proposed ∩ allowed ∩ available | Synthesis | Implementation claim only if contracts, receipts, and runtime behavior are specific and novel. |
| Intent-to-topology compiler for opportunity objects | Potentially distinctive synthesis | Requires a precise algorithm, baselines, and experiments. |
| Topology receipts tied to evidence, authority, cost, and outcome | Potentially strong claim area | Requires schema, runtime implementation, and demonstrated learning value. |
| Galaxy-to-World emission from one organism | Potential category claim | Must be defined in machine-verifiable contracts, not narrative alone. |
| Cross-domain opportunity discovery + execution + outcome compounding | Strongest commercial synthesis | The moat depends on proprietary recipes, data rights, outcome history, and repeatability. |
| N-Cell/64 header | Potential protocol claim | Novelty depends on exact field semantics, interoperability, and measured advantage. |
Task-adaptive topology improves success versus the best fixed topology, using identical models and tools.
N-Cell state signatures reduce debugging and incident-resolution time without losing semantic traceability.
Governance graph reduction lowers the unauthorized-action rate while preserving useful completion.
Outcome-linked memory changes future topology selection in a direction that improves calibration.
Sparse route activation reduces token and latency cost at equal or better quality than dense communication.
One canonical substrate transfers across at least five Worlds with limited domain-specific orchestration code.
Topology receipts improve reproducibility of stochastic agent runs.
Opportunity-centered composition produces higher measurable economic value than task-centered automation alone.
Sections 14–16 · Build plan and commercial shape
The build is fixture-first with hard stop conditions per phase, and every asset carries a status on a strict ladder. Built, deployed, live, and revenue-active are different states, and conflating them is a named failure mode: registry inflation.
| Phase | Deliverables | Stop condition |
|---|---|---|
| 0 · Contract Reconciliation | Inventory of existing modules, reuse / refactor / retire map, core and domain schemas. | All schemas validate; no duplicate organism services introduced. |
| 1 · Universal Intake + Evidence | Opportunity assembler, evidence ledger, conflict detection, source policy. | A deal object builds from sparse and conflicting inputs. |
| 2 · Core Real Estate Organs | Comparable Intelligence, Market Velocity, Photo Repair, Residential ARV, Land Zoning. | Fixtures pass deterministic formula and schema tests. |
| 3 · Strategy Organs | Rental / BRRRR, multifamily, development, creative finance, offer architecture. | Scenarios compare profit, risk, capital, and time. |
| 4 · Governance + Risk | Kill-switch policies, approval gates, data-sufficiency checks. | Blocked and escalated actions behave predictably. |
| 5 · APG Cockpit + Emitters | Deal views, review screens, report templates, CRM notes. | An operator can review and approve a deal from intake to report. |
| 6 · Outcome Intelligence | Outcome capture, variance scoring, calibration workflow. | Predicted vs. actual feeds the governed learning loop. |
Six fixture types gate every phase: baseline clean, sparse data, conflict, unique asset, risk, and outcome. Acceptance requires schema validation, unit-tested formulas with known outputs, evidence or an explicit assumption label on every claimed fact, and a run receipt with versioned prompts, formulas, sources, inputs, warnings, and output IDs.
APG CRM + organ pack finds and structures deals faster. Customer: the Neuruh / APG operating team.
Paid underwriting packets and deal opinions. Customer: wholesalers, buyers, agents, lenders.
APG Land CRM powered by Neuruh organs. Customer: land buyers, wholesalers, small brokerages.
Underwriting, comp scoring, market heat, repair scope endpoints. Customer: proptech, CRMs, agencies, investors.
Market heat, parcel unlocks, owner signals, distressed lists. Customer: investors and acquisition teams.
Lead analysis, offer packages, seller scripts, investor packets. Customer: operators who need output, not software.
Resale, TaxShield, StudioGPT, Bookeit, defense, optimization. Customer: industry-specific buyers.
“Neuruh finds hidden value, proves it, and builds the system that captures it.”