ariffazil/explorer-intelligence-architecture

🧭 EXPLORER INTELLIGENCE ARCHITECTURE

The governed metabolism of discovery. OBSERVE → HYPOTHESIZE → FALSIFY → VERIFY Forged: 2026-07-06 | Ratified: F13 SOVEREIGN DITEMPA BUKAN DIBERI

Reference contracts:

• references/memory-schema.md — structured explorer memory contract
• references/route-contract.md — lawful cross-organ routing rules
• references/explorer-packet-schema.md — canonical JSON/YAML packet schemas
• references/dispatch-protocol.md — Hermes/OpenCode/OpenClaw/A-FORGE stage handoff protocol
• references/falsifier-routing.md — OpenClaw vs A-FORGE falsifier decision law
• assets/agent-stage-map.yaml — machine-readable stage ownership map
• assets/explorer_packet.template.yaml — ready-to-fill explorer packet template
• assets/falsification_packet.template.yaml — ready-to-fill falsification handoff template
• assets/dispatch_packet.template.yaml — canonical 8-packet dispatch template set
• scripts/validate_explorer_packet.py — local validator for packet completeness and protocol order
• scripts/validate_dispatch_packet.py — local validator for dispatch packet families
• scripts/route_dispatch_stage.py — minimal stage router using agent-stage-map.yaml
• scripts/log_handoff.py — append compact handoff receipts to local JSONL

---

0. THE COMPLETE STACK

KNOWLEDGE GRAPH (555-ASI) — the universe
        ↓ mapped by
EXPLORER INTELLIGENCE — OBSERVE → HYPOTHESIZE → FALSIFY → VERIFY
        ↓ governed by
APEX THEORY — Δ clarity · Ω humility · Ψ vitality
        ↓ judged by
arifOS KERNEL — SEAL / SABAR / HOLD / VOID
        ↓ civilized by
AAA — identity · leases · domain_law · floors
        ↓ evolved by
A-FORGE — birth → test → mutate → retire

This is not a diagram.
This is the physics of governed intelligence.

Before this skill, every agent explored by instinct.
Now every agent explores by protocol.

---

1. THE FOUR-PHASE METABOLISM

Every exploration by every AAA warga follows exactly four phases:

┌─────────────────────────────────────────────────────────────┐
│                                                             │
│   OBSERVE ──→ HYPOTHESIZE ──→ FALSIFY ──→ VERIFY ──→ SEAL  │
│      ↑                          │                           │
│      └──────────────────────────┘                           │
│         (falsification failed → re-observe)                 │
└─────────────────────────────────────────────────────────────┘

Phase 1: OBSERVE

What: Ingest raw reality from the domain.

| Organ | What it observes | Tool | |-------|-----------------|------| | GEOX | Seismic, well logs, basin data | geox_well_ingest, geox_seismic_ingest | | WELL | Biometrics, fatigue, cognitive load | well_validate_vitality | | WEALTH | Prices, fiscal data, portfolio | wealth_market_data | | A-FORGE | Tool surface, drift, entropy | forge_surface_guard | | arifOS | Governance signals, floor violations | arif_observe |

Exit condition: At least one OBSERVED (OBS) claim with confidence ≥ 0.70.
If no OBS claim → return to OBSERVE with wider aperture.

Floor check: F2 TRUTH — label everything OBS/DER/INT/SPEC.

Phase 2: HYPOTHESIZE

What: Generate ≥3 competing hypotheses grounded in domain law.

Hypothesis A — the dominant model (what current evidence suggests)
Hypothesis B — the contrarian model (what contradicts A)
Hypothesis C — the null model (random or noise-driven)

Rules:

• Must consult the 555-ASI knowledge graph for domain boundaries
• Each hypothesis must declare its domain_law provenance
• Each hypothesis must have at least one falsifiable prediction
• Confidence cap at 0.75 — hypotheses are not yet truth

Tool: arif_think(mode="reason") with explicit hypothesis structure.
Graph: Query /root/555-ASI/knowledge-taxonomy.json for relevant edges.

Exit condition: ≥3 hypotheses, each with ≥1 falsifiable prediction.
If <3 → hypothesize harder or acknowledge knowledge gap.

Floor check: F7 HUMILITY — cap at 0.75. F9 ANTI-HANTU — never claim certainty.

Phase 3: FALSIFY

What: Attack all hypotheses using cross-domain physics, biology, economics, or governance.

This is the eureka engine. Contradictions are surfaced here.

Falsification gates (use all that apply):

| Gate | Tool | What it falsifies | |------|------|-------------------| | Physics | geox_contrast_detect | Anomalous seismic, AVO mismatch | | Stratigraphy | geox_biostrat_falsify (8 gates) | Age, facies, taxonomy | | Sequence | geox_biostrat_ruling_check | Facies veto, reworking | | Economics | wealth_asymmetry_check | Risk asymmetry | | Collapse | wealth_collapse_signature_scan | Institutional failure pattern | | Capture | wealth_capture_scan | Hidden incentives | | Governance | arif_critique | Maruah, human impact | | Quantum Eureka | quantum-eureka-doctrine skill | Cross-domain contradiction |

Protocol for each falsification:

1. For each hypothesis, apply ALL relevant gates
2. Record: which gate passed, which failed, which is inconclusive
3. If any gate returns CONTRADICTION → escalate to ELEVATED
4. If ALL gates pass → hypothesis survives (provisionally)
5. If ALL hypotheses falsified → return to OBSERVE

Exit condition: At most 1 surviving hypothesis OR at least 1 eureka candidate.
If all hypotheses falsified → return to OBSERVE (this is not failure — it is discovery).

Floor check: F1 AMANAH — document all falsifications. F3 WITNESS — tri-witness required for any claim.

Phase 4: VERIFY

What: Run bounded, auditable tests on the surviving hypothesis.

1. Generate test prediction from the hypothesis
2. Run bounded test (computation, simulation, look-up)
3. Compare prediction vs outcome
4. Record uncertainty band
5. If test passes → SEAL candidate
6. If test fails → return to HYPOTHESIZE or OBSERVE

Tools:

• geox_seismic_compute — synthetic modelling
• geox_3d_model_build — structural validation
• wealth_compute_emv — expected value validation
• well_compute_metabolic_flux — system health validation
• arif_judge — constitutional validation

Exit condition: Test passes with uncertainty band ≤ 0.20.
If uncertainty > 0.20 → surface as SABAR (not SEAL).

Floor check: F2 TRUTH — every result labeled. F13 SOVEREIGN — 888_HOLD before irreversible.

---

2. THE FOUR SUBMODES (per agent identity)

Every agent maps the four phases to their domain:

| Agent | OBSERVE | HYPOTHESIZE | FALSIFY | VERIFY | |-------|---------|-------------|---------|--------| | GEOX | Ingest seismic/well/basin | Geology models (3+) | Physics + strat + biostrat | Forward model + well tie | | WELL | Biometrics + fatigue | Health models (3+) | Physiology + stress | Vitality validation | | WEALTH | Market data + portfolio | Economic models (3+) | Asymmetry + collapse | EMV/Monte Carlo | | A-FORGE | Tool surface + drift | Fitness models (3+) | Scar + governance | Test suite | | arifOS | Governance signals | Verdict models (3+) | Floor violations | Constitutional review | | AAA | Agent registry + leases | Identity models (3+) | Contradiction + drift | Session validation |

---

3. THE THREE INVARIANTS (APEX binding)

Every exploration loop must satisfy three invariants:

Δ (Clarity) — Evidence ≥ Confidence

for every claim in the exploration loop:
    evidence_strength ≥ confidence
    if evidence_strength < confidence → downgrade confidence
    if confidence < 0.50 → return to OBSERVE

Ω (Humility) — Declare Boundaries

for every hypothesis: 
    declare domain_law provenance
    declare uncertainty band
    declare blindspots
    declare what would falsify it

Ψ (Vitality) — Protect System + Human

for every edge in the exploration:
    check organ health (all 6)
    check human readiness (WELL)
    check for collapse signals (WEALTH)
    if any is RED → HOLD, not proceed

---

4. EUREKA TRIGGER (bridge to quantum-eureka-doctrine)

If during FALSIFY phase, a contradiction survives ≥2 gates:

1. Package as eureka candidate (see quantum-eureka-doctrine §3 for trigger conditions)
2. Route to arif_judge for constitutional review
3. If SEAL → seal to VAULT999 + record scar
4. If HOLD → surface as SABAR pending more data
5. If VOID → log and move on

This is where structured memory + cross-domain reasoning + contradiction detection produce the quantum-style eureka.

For implementation-facing work, load the four reference contracts above before designing storage, packets, or cross-organ handoff. When emitting a real packet, start from assets/explorer_packet.template.yaml and validate it with python scripts/validate_explorer_packet.py <packet.yaml>.

---

5. ANTI-PATTERNS

| Anti-pattern | Phase | Why | What to do | |---|---|---|---| | Skipping OBSERVE | 1 | Hypotheses without evidence = hallucination | F9 — always observe first | | Single hypothesis | 2 | Confirmation bias engine | F8 — generate ≥3 | | Self-falsification | 3 | Never trust your own hypothesis | Use independent gate per falsification | | Smoothing contradictions | 3 | Eureka suppression | Surface them — don't hide | | Over-verifying | 4 | Analysis paralysis | Cap at uncertainty ≤ 0.20 | | Ignoring Ψ | all | System collapse | Always check vitality | | Looping without progress | any | Mirror loop | After 3 iterations → escalate to 888 | | Declaring eureka without tri-witness | 3→4 | F3 violation | Require human × AI × external |

---

6. MIRROR LOOP DETECTION

If the exploration loop repeats ≥3 times without advancing:

def detect_mirror_loop(iteration_count, delta_entropy, hypothesis_diversity):
    if iteration_count >= 3:
        if delta_entropy < 0.05:
            return "MIRROR_LOOP — entropy not decreasing"
        if hypothesis_diversity < 2:
            return "MIRROR_LOOP — same hypothesis recycled"
    return "HEALTHY"

On MIRROR_LOOP: Escalate to 888_HOLD with full iteration history. Do not continue. The loop is consuming compute without producing insight.

---

7. FLOOR ALIGNMENT

| Floor | Exploration obligation | |-------|----------------------| | F1 AMANAH | Every mutation reversible. Escalate 888 before irreversible. | | F2 TRUTH | Label OBS/DER/INT/SPEC at every phase. Cap confidence at 0.90 (OBS). | | F3 WITNESS | Tri-witness required before SEAL. | | F4 CLARITY | ΔS ≤ 0 per iteration. Each loop must reduce entropy. | | F7 HUMILITY | Hypotheses capped at 0.75. Uncertainty bands mandatory. | | F8 GENIUS | ≥3 hypotheses. Simplest correct path wins. | | F9 ANTI-HANTU | Never claim eureka without evidence. Falsification = anti-hallucination. | | F11 AUDIT | Every exploration loop logged. Every contradiction recorded. | | F13 SOVEREIGN | 888_HOLD before irreversible. Sovereign decides what becomes doctrine. |

---

8. INTEGRATION WITH OTHER SKILLS

| Skill | Role in Explorer Architecture | |-------|------------------------------| | 000-init-intent-classify | Stage 0 — classify intent before exploration begins | | 111-sense-evidence-observe | Phase 1 — evidence binding and tagging | | 333-mind-plan-generate | Phase 2 — hypothesis generation scaffold | | 666-heart-critique-stress | Phase 3 — ethical stress-testing | | 888-judge-verdict-render | Phase 4 — constitutional verdict | | 999-vault-seal-immutable | Post-loop — seal insight to VAULT999 | | quantum-eureka-doctrine | Phase 3→4 bridge — contradiction lifecycle | | shadow-diagnostic | Pre-output — alignment self-check | | CONSTITUTIONAL_REFLEX | Always-on — ART → KERNEL → ACT arc |

---

9. LOADING CONTRACT

# At session init, after heptalogy load:
skill(name="explorer-intelligence-architecture")
skill(name="quantum-eureka-doctrine")

# Then when exploring any domain:
def explore(query: str):
    # 1. Classify intent
    intent = classify(query)  # calls 000-init-intent-classify
    
    # 2. Determine which organ(s)
    organ = route_intent(intent)  # consults 555-ASI/organ-affinity-index.json
    
    # 3. Execute exploration loop
    observation = observe(organ, query)
    if not observation:
        return "OBSERVE FAILED — widen aperture"
    
    hypotheses = hypothesize(organ, observation, n=3)
    if len(hypotheses) < 3:
        return "HYPOTHESIZE FAILED — knowledge gap"
    
    survivors = falsify(organ, hypotheses)
    if len(survivors) == 0:
        return "ALL HYPOTHESES FALSIFIED — return to OBSERVE"
    
    verdict = verify(organ, survivors[0])
    if verdict.uncertainty > 0.20:
        return f"SABAR — uncertainty {verdict.uncertainty}"
    
    arif_seal(verdict)
    return f"SEAL — {verdict.summary}"

---

10. QUICK REFERENCE CARD

EXPLORER METABOLISM — OBSERVE → HYPOTHESIZE → FALSIFY → VERIFY

PHASE 1: OBSERVE
  Ingest raw reality. Exit: OBS claim ≥ 0.70 confidence.

PHASE 2: HYPOTHESIZE
  Generate ≥3 competing hypotheses. Each must be falsifiable.
  Consult 555-ASI graph for domain boundaries.

PHASE 3: FALSIFY
  Attack ALL hypotheses using cross-domain gates.
  Physics → stratigraphy → economics → governance.
  Contradiction = eureka candidate.
  All falsified = return to OBSERVE (not failure).

PHASE 4: VERIFY
  Bounded test. Uncertainty ≤ 0.20. 
  Pass → SEAL candidate. Fail → return to HYPOTHESIZE or OBSERVE.

INVARIANTS (APEX binding):
  Δ — evidence ≥ confidence
  Ω — declare boundaries + blindspots
  Ψ — protect system + human health

MIRROR LOOP: ≥3 iterations without progress → 888_HOLD

EUREKA: contradiction survives ≥2 gates → quantum-eureka-doctrine

ANTI-PATTERNS:
  Skip observe, single hypothesis, smooth contradictions,
  self-falsify, ignore vitality, mirror loop

---

11. AUTHORITY

This skill is the canonical exploration protocol for all AAA warga agents, forged under F13 SOVEREIGN directive. It binds every agent that loads it to the OBSERVE → HYPOTHESIZE → FALSIFY → VERIFY metabolism.

It does not replace the constitutional floors (F1-F13), the APEX theory framework, or the Quantum Eureka Doctrine. It activates them.

DITEMPA BUKAN DIBERI — Exploration is forged, not given.

---

Forged: 2026-07-06 by FORGE (000Ω) for Arif (F13 SOVEREIGN) Heritage: ASI💃 synthesis → F13 sovereign ratification Integration: 555-ASI knowledge graph · quantum-eureka-doctrine · APEX theory · GEOX · WELL · WEALTH · A-FORGE · AAA · arifOS Domain: cross-cutting — all 7 organs