ESG Oracle

Proprietary technology. The ESG Oracle scoring methodology, weighting model, redistribution logic, and source code are proprietary intellectual property of Untitled_ LuxPerpetua Technologies, Inc. API access is available to approved beta partners. Self-hosting requires a separate license agreement — apply via the beta access page.

The DPX ESG Oracle produces live E, S, G scores from 8 real-world institutional data sources. Those scores flow directly into settlement fees: a higher score means a lower fee. 100% of ESG fee revenue is redistributed to verified on-chain impact programs — making every DPX settlement a direct contributor to measurable environmental and social outcomes.

The Social (S) dimension includes human rights metrics (child labour, labour rights enforcement, safety from violence) and a gender equity sub-score (women in management, earnings parity, gender-responsive institutions) — making DPX the first stablecoin settlement rail with human rights and gender equity embedded directly in its fee structure.

Data Sources

The oracle pulls from 8 institutional data sources across the three ESG dimensions:

Environmental (E)

Source	API	Metrics
World Bank	`api.worldbank.org`	CO2 emissions (metric tons per capita) — `EN.ATM.CO2E.PC`
Climate Monitor	`climatemonitor.info/api`	Live CO2 levels (ppm), CH4 levels (ppb)

Source	API	Metrics
UN SDG API	`unstats.un.org/SDGAPI`	SDG 4.1.1 — education; SDG 3 — health
UN SDG API — Human Rights	`unstats.un.org/SDGAPI`	SDG 8.7.1 — child labour; SDG 16.1.3 — safety from violence; SDG 16.b.1 — non-discrimination; SDG 8.8.2 — labour rights compliance
UN SDG API — Gender Equity	`unstats.un.org/SDGAPI`	SDG 5.5.2 — women in management; SDG 5.c.1 — gender-responsive budgeting; SDG 8.5.1 — earnings parity

Governance (G)

Source	API	Metrics
IMF	`dataservices.imf.org`	Consumer Price Index (CPI) — economic stability proxy
OECD	`sdmx.oecd.org`	GDP data — economic governance indicator
SEC EDGAR	`data.sec.gov`	Corporate governance disclosures (XBRL company facts)

The Social (S) score is a proprietary weighted composite across four sub-dimensions, drawing on OHCHR Universal Periodic Review frameworks, WEF Gender Gap Index methodology, and UN SDG data standards:

Sub-dimension	Source	What it captures
Human Rights	UN SDG 8.7, 16.1, 16.b	Child labour, safety from violence, non-discrimination, labour rights compliance
Education	UN SDG 4.1.1	Long-run human capital development
Gender Equity	UN SDG 5.5, 5.c, 8.5	Women in management, earnings parity, gender-responsive institutional frameworks
Health	UN SDG 3	Population wellbeing baseline

The weighting model is proprietary intellectual property. Full methodology documentation is available to approved institutional partners under NDA — apply via the beta access page.

Fee Component

The ESG Oracle score directly determines the ESG fee component of each settlement. A higher score means a lower fee. The component ranges from 0.00% (score 100) to 0.50% (score 0). The on-chain contract, oracle, and API always apply the same calculation consistently.

The scoring methodology, sub-score weighting model, and fee calculation logic are proprietary intellectual property of Untitled_ LuxPerpetua Technologies, Inc. Full methodology documentation is available to approved institutional partners under NDA — apply via the beta access page.

ESG Redistribution Mechanism

ESG fee revenue from bad actors is redistributed to a verified on-chain impact wallet. This is enforced by the ESGRedistribution contract — it is not discretionary.

A company qualifies as a bad actor when its weighted aggregate ESG score (across WorldBank, UN, ClimateMonitor, IMF, OECD, and SEC) falls at or below the redistribution threshold (40/100). At that point, fees generated by that company are redirected from the general fee pool to the impact wallet.

Industry Buckets

The contract maps each company to an industry. When redistribution is triggered, fees route to the impact wallet tagged by industry for reporting purposes.

Industry	Coverage
FossilFuels	Oil, gas, coal extraction
Energy	Renewable energy, utilities
Manufacturing	Industrials, chemicals, materials
Agriculture	Food, farming, fishing
Technology	Software, hardware, data centres, telecom
Finance	Banks, insurance, investment
Consumer	Retail, fashion, apparel
Healthcare	Pharma, medical devices, biotech
RealEstate	Property, construction, infrastructure
Transportation	Shipping, aviation, logistics

How it works on-chain

Company’s ESG score is set per-provider by the oracle updater on ESGOracle
ESGRedistribution.aggregatedScore(company) computes the weighted average across all active providers
If score ≤ 40, isBelowThreshold(company) returns true
redistributeERC20(token, company, amount) routes fees to the impact wallet
A RedistributionExecuted event is emitted — auditable by any counterparty, regulator, or auditor

Impact wallet

All redistributed fees consolidate to a single on-chain impact wallet (0x5b26325d3f6f3A2B5FAdEbe54c780A8C3dAFd6Ba) — a multi-signature Safe wallet on Base mainnet. Destination addresses can only be updated via setFundingAreas — callable by the contract owner only.

Impact at scale

Annual Volume	Bad Actor Fee Pool (score 20)	Redirected
$100M	~$200K	100% to impact wallet
$1B	~$2M	100% to impact wallet
$5B	~$10M	100% to impact wallet

Every redistribution is verifiable on-chain. Any agent, auditor, or regulator can independently confirm funds reached the impact wallet via Base Blockscout.

AI Intelligence Layer

Proprietary technology. The AI synthesis methodology, qualitative signal sources, and inference infrastructure are proprietary intellectual property of Untitled_ LuxPerpetua Technologies, Inc.

The ESG Oracle includes an embedded AI intelligence layer that runs after scores are computed. It synthesises the quantitative E, S, G scores alongside live qualitative signals into a structured institutional briefing returned as an intelligence object on every oracle response.

What it produces:

Output	Description
`reasoning`	Plain-language explanation of the primary ESG signal drivers and material risk areas
`confidence`	0.0–1.0 reflecting data freshness, source coverage, and signal clarity
`alerts`	Up to 3 ESG risk items material to institutional counterparties or regulators
`outlook`	`IMPROVING` / `STABLE` / `DETERIORATING` / `UNCERTAIN`

The AI layer draws on qualitative signals beyond the quantitative data sources — including real-time news events from global sources — to identify emerging ESG risks before they appear in institutional data feeds. The specific sources and synthesis methodology are proprietary.

If synthesis is unavailable, the oracle returns the full quantitative result and omits the intelligence field. The quantitative scores are always the authoritative input for fee calculations.

Adaptive Layer

Proprietary technology. The adaptive learning architecture, weight regression model, entity bootstrapping methodology, and policy execution logic are proprietary intellectual property of Untitled_ LuxPerpetua Technologies, Inc.

The ESG Oracle includes a fully autonomous adaptive layer that continuously improves E/S/G score weighting, calibrates confidence, and bootstraps ESG profiles for new counterparties — running entirely on Cloudflare native infrastructure.

Architecture overview:

Component	Technology	Role
Prediction ledger	Cloudflare D1	Logs every oracle run; resolves E/S/G predictions against actuals; scores per-dimension accuracy
Weight regression	Cloudflare Workflows (correlation)	Adjusts the E/S/G weighting using 30-day accuracy correlation — weekly
Confidence calibration	Cloudflare Workflows (Platt scaling)	Sigmoid calibration of raw confidence scores weekly
Semantic memory	Cloudflare Vectorize (3-dim signals)	Fingerprints each run as normalised [E, S, G] vector; recalls top-3 similar historical scenarios for AI prompt injection
Entity profiles	Cloudflare Vectorize (5-dim profiles)	Stores per-counterparty [E, S, G, settlement_count, avg_fee] vectors; nearest-neighbour bootstrapping for new entities
Async event pipeline	Cloudflare Queues	Non-blocking — oracle response latency is never affected by adaptive writes
Policy execution	Cloudflare Workflows	5-gate safety check before any on-chain call to `ESGCompliance.setESGFee()`

Entity bootstrapping:

When a new counterparty address has no ESG history, the ESG Oracle queries Vectorize for the nearest known entity by profile similarity (cosine score > 0.90). If found, the new entity inherits that neighbour’s E/S/G priors as a starting point — rather than defaulting to an arbitrary equal-weight score. This means first-settlement pricing is informed by comparable actors rather than a cold start.

Adaptive weight bounds:

The minimum weight floor per E/S/G component is 10% (higher than Stability Oracle’s 5% — justified by the smaller number of components). Max shift per week: 2%. All bounds enforced by immutable TypeScript SAFETY_BOUNDS.

Cron schedule:

Cron	Job
`0 * * * *` (every hour)	Oracle update
`0 5 * * 0` (Sunday 05:00 UTC)	Weight regression workflow
`0 6 * * 0` (Sunday 06:00 UTC)	Calibration workflow

Adaptive status endpoint:

GET /api/adaptive/status

Returns current E/S/G adaptive weights, prediction count, and circuit breaker state.

Live Endpoint

GET https://esg.untitledfinancial.com/esg-score

Returns the current E, S, G scores, composite average, active fee rate, tier label, and AI intelligence synthesis. Use scores.average as the esgScore parameter in settlement quotes on the Stability Oracle.