Methodology · Sanitized

GMIIE Public Methodology

High-level specification of the intelligence engine published on xxxiii.io. Trade-secret weightings, scraper selectors, and unpublished ML features are omitted. Current readings match the Oracle Engine editorial snapshot (May 2026).

Human vs AI knowing GMIIE engine outputs are mathematical and cryptographic intelligence — not opinion prose. Humans on the network carry judgment, institutional context, and explicit experiential opinion. Both lenses are deliberate. Full epistemology → How GMIIE Knows

NIG — Normalized Intelligence Grade

Five ring inputs are combined with published α coefficients (Apr 2026 editorial vintage — LSTM retrain pipeline is roadmap, not deployed), then bounded via hyperbolic tangent:

NIG = tanh(α₁·STRUCT + α₂·LANG + α₃·DEPLOY − α₄·FRAG + α₅·FRACT) × 100 Published coefficients (Apr 2026 vintage): α₁=0.28 α₂=0.14 α₃=0.24 α₄=0.22 α₅=0.12 Current reading: NIG = +28 (MILD POSITIVE)

Ring weights for module-level SSC composite: w = (0.25, 0.15, 0.25, 0.20, 0.15) across R1–R5.

Core formulas (F-01 – F-12)

CFS — Correlation Fracture Score (R5)

CFS = |ρ_current − ρ_36M| × (1/σ_ρ) × Regime_distance Alert threshold: ≥ 2.1σ (fracture) Current: Yen carry CFS = 2.1σ ALERT · Dollar/Gold = 1.4σ WATCH

ODV — On-Chain Deployment Velocity (R3)

ODV = (ΔTVL_inst / TVL_inst,30d) × (1 + ln(NewContracts)) × RegClarity_factor Current: ODV = 0.71 · US RegClarity 0.75 → 1.25 (GENIUS Act)

PCS — Prediction Calibration Score (Oracle)

PCS = Accuracy × Sharpness × (1 − Overconfidence) Quality gate: PCS > 0.68 Current: PCS = 0.71 · 24M backtest = 0.74

PCS is maintained by post-resolution scoring on Oracle questions; community probabilities are not treated as ground truth until resolve events fire.

Related public metrics

SFI (R4): interconnect × (1 + CRE/Assets) × liquidity mismatch / capital buffer — banks SFI > 1.6 on watch.
PCSG (R4): private credit stress — current 0.38 (Elevated).
GCS (R5): geopolitical contagion with λ = 0.08/day decay — current 0.79.
CATS (R3): CBDC adoption trajectory — global 218/300.
LDS (R2): language drift — BoJ +0.71, Fed −0.28.

Full formula cards: Engine → 12 Signal Formulas. Institutional extensions (IMF Taylor, WEF risk graph, labor Markov matrix): Institutions.

R1–R5 ring structure

Ring	Focus	Score	Status
R1 STRUCT	Yield curve, SCFI, macro architecture	62	Watch
R2 LANG	Central bank NLP / LDS	58	Moderate
R3 DEPLOY	ODV, CATS, on-chain institutional flow	71	Positive
R4 FRAG	SFI, PCSG, CRE wall	44	Elevated
R5 FRACT	CFS, GCS, regime distance	55	Watch · fracture alert

Scores refresh on a 4-hour module cycle (GMIIE intelligence pipeline + live-desk cron); UI may show static seed until KV warms.

Oracle calibration

Oracle questions publish community-weighted probabilities with documented consequence chains. Calibration target PCS > 0.68; current 0.71 exceeds IMF-reported ML quality gates cited on Institutions. Backtest window: 24 months rolling; resolution events update Accuracy and Sharpness terms in F-12.

Verification

IPFS: 8-hour snapshots — /api/ipfs/latest
JSON API: /api/gmiie/live/engine, /data/gmiie-engine-snapshot.json
Reproducibility: Compare engine_fingerprint and generated_at — see Authenticity

Internal canonical specs: private repo docs/canonical/ (not deployed). Consult IP counsel for trademarks and patents.