Skip to content
DPX DPX Docs

Predictive Intelligence Layer

Proprietary technology. The predictive intelligence architecture, signal weighting model, regime estimation methodology, probability recalibration approach, scenario memory encoding, and causal graph methodology are proprietary intellectual property of Untitled_ LuxPerpetua Technologies, Inc. This page describes what the system does and what it returns — not how it is implemented internally.

The DPX Predictive Intelligence Layer is a set of four interconnected systems that give the Stability Oracle a forward-looking dimension: it does not just describe current conditions, it estimates where conditions are going, tracks whether those estimates were right, learns from every call, and maps the causal structure driving regime changes.

Overview

Section titled “Overview”
SystemEndpointWhat it does
Regime transition probabilities/v1/intelligence/transition-risk30/60/90-day forward probability distributions over 5 regimes
Prediction ledger/v1/intelligence/ledgerStores every forecast at call time, scores it when the horizon arrives
Scenario memory(internal — returned in transition-risk response)Matches current conditions against historical precedents
Causal graph/v1/intelligence/causal-graphWeighted graph of signal influence; calibrates against live observation history

All four systems are live and share a single data plane: each transition-risk call seeds the ledger and scenario memory; a daily job resolves predictions and updates causal edge weights.

Regime Transition Probabilities

Section titled “Regime Transition Probabilities”

Endpoint: GET /v1/intelligence/transition-risk

Synthesises multiple lead signals — including climate, tectonic, water-risk, gender-risk, earth-systems, resonance, and mycelium indicators — into a current composite stress index and forward probability distributions over five regimes at 30, 60, and 90-day horizons.

Five regimes:

RegimeComposite score band
STABLE0–20
ELEVATED20–40
STRESSED40–60
CRITICAL60–80
CRISIS80–100

Longer-lead signals are weighted more heavily at longer horizons, while shorter-lead signals dominate the near-term call — the 90-day forecast skews toward structural/climate drivers, the 30-day forecast toward financial-system signals. The exact weighting and recalibration model is proprietary.

Response shape (abridged):

{
  "generatedAt": "2026-06-13T14:00:00Z",
  "currentScore": 55,
  "currentRegime": "STRESSED",
  "transitionSignal": "DETERIORATING",
  "horizons": {
    "h30": { "mostLikely": "STRESSED", "confidence": 0.72, "probabilities": { "STABLE": 0.03, "ELEVATED": 0.11, "STRESSED": 0.49, "CRITICAL": 0.28, "CRISIS": 0.09 } },
    "h60": { "mostLikely": "STRESSED", "confidence": 0.61, "probabilities": { "...": "..." } },
    "h90": { "mostLikely": "CRITICAL", "confidence": 0.54, "probabilities": { "...": "..." } }
  },
  "scenarioMemory": {
    "similarScenarios": [
      { "date": "2026-06-12", "similarity": 0.99, "regime": "STRESSED", "daysAgo": 1 }
    ],
    "patternNote": "2 similar configurations found. Most common precedent regime: STRESSED."
  }
}

Cache: 1 hour.

Prediction Ledger

Section titled “Prediction Ledger”

Endpoint: GET /v1/intelligence/ledger

Every fresh (non-cache-hit) transition-risk call stores a prediction for each horizon (30, 60, 90 days) — the predicted regime, its probability, and the full distribution at call time. Once a horizon arrives, a scheduled job resolves the prediction against the actual outcome and scores it. Accuracy scores accumulate per horizon and feed a periodic recalibration pass that corrects systematic over- or under-confidence in the probability outputs.

Ledger response (abridged):

{
  "summary": {
    "totalPredictions": 24,
    "resolved": 6,
    "pending": 18,
    "byHorizon": {
      "h30": { "predictions": 8, "resolved": 2, "regimeAccuracy": 1.0 },
      "h60": { "predictions": 8, "resolved": 2, "regimeAccuracy": 0.5 },
      "h90": { "predictions": 8, "resolved": 2, "regimeAccuracy": 0.5 }
    }
  }
}

The ledger is the mechanism that makes DPX self-improving: as predictions resolve, accuracy accumulates, calibration drift is detected, and probability outputs are recalibrated over time. The longer the system runs, the more precisely it understands which signal configurations matter.

Scenario Memory

Section titled “Scenario Memory”

Scenario memory runs inside every fresh transition-risk call. It is not a separate endpoint — it surfaces as the scenarioMemory field in the transition-risk response.

Each call encodes the current signal configuration and compares it against previously stored configurations, surfacing the closest historical precedents along with their similarity, stored regime, and (once resolved) what regime they actually transitioned to. The specific encoding and matching methodology are proprietary.

As the history grows, this gives the oracle a genuine pattern library — the ability to say “this configuration appeared before; conditions typically deteriorated to CRITICAL within 60 days.”

Causal Graph

Section titled “Causal Graph”

Endpoint: GET /v1/intelligence/causal-graph

A weighted graph of signal influence spanning lead, amplifier, coincident, and lagging signal tiers. Edge weights start from theoretical priors grounded in established macro/financial theory, then calibrate toward empirically observed signal relationships as daily observations accumulate.

Calibration phases:

PhaseBehaviour
PRIOREdge weights reflect theory only; too few observations for empirical calibration
BLENDINGEmpirical data begins to take a share of edge weights
EMPIRICALEdge weights are predominantly data-driven

The full node set, edge weights, and calibration methodology are proprietary. The response surfaces a topology summary — the dominant causal path through the network and which signal is currently most influential — without exposing the underlying model.

Graph response (abridged):

{
  "topology": {
    "phase": "PRIOR",
    "dominantCausalPath": ["climate", "water-risk", "instability", "macro-stress", "currency-stress"],
    "mostInfluential": "climate",
    "description": "DISTRIBUTED — risk is distributed; no single node dominates"
  }
}

How the four systems work together

Section titled “How the four systems work together”

Every fresh call to /transition-risk computes the current probability distributions, records predictions in the ledger, and checks scenario memory for historical precedents. A daily scheduled job resolves matured predictions, updates calibration, and refreshes the causal graph’s edge weights from newly observed data.

The result is a system that improves with time in three distinct ways: probabilistic calibration improves as forecast accuracy accumulates; pattern matching improves as the scenario memory history grows; and causal understanding improves as empirically observed signal relationships gradually displace theoretical priors.

Auth and pricing

Section titled “Auth and pricing”
EndpointPriceCache
GET /v1/intelligence/transition-risk$0.751h
GET /v1/intelligence/ledger$0.25no-cache
GET /v1/intelligence/causal-graph$0.505min

Auth: x402 USDC on Base, or X-API-Key header. Base URL: https://intelligence.untitledfinancial.com

Terminal window
curl https://intelligence.untitledfinancial.com/v1/intelligence/transition-risk \
  -H "X-API-Key: your-key"