AUROS
Connexion
AUROS Ledger
AUROSAUROS

Public methodology

How the notation pipeline actually runs

Not a single LLM pass. Not a mega-score. Five public sources stay unfused; stats produce an envelope; template NLI reads claim vs evidence; a lab desk argues both sides from the same numbers. Indicative. Contestable. Not a licensed CRA.

Open the notation desk

ScoreActivityGapVarianceIoT

The diagram below is the product — coded from the same rules as the scores, not a slide pasted on the page. Lab and research sit on dashed / dotted lanes so they cannot be mistaken for live coverage.

AUROS rating pipeline — from public file to contestable note

livelabresearch

Sources

live

Five inputs, never fused. The lab desk reads this bus — no sixth feed.

  • Compare
  • Activity
  • Freshness
  • Transparency
  • Friction

Named scores

live

Separate layers. Each 0–100 + badge. Never a mega-rating.

S = { Activity, Gap, Freshness, Depth, Friction }

Envelope

live

Leave-one-dimension-out → disagreement band. Stats object, not a vibe.

method = lodo_disagreement · band = [min(point, LOO) − σ̂, max(point, LOO) + σ̂]

Conformal 95% (research)

research

Not live — labeled holdout n=0 (split / CV+). Adjacent live: retrospective LODO coverage on Gap history.

coverage target 95% — holdout empty · diagnostic = next∈prior LODO band

Claim vs evidence (NLI)

live

Template labels on Promise Gap densified with claim sub-signals. Not a legal verdict. Not a chat pass.

label ∈ {entailment, contradiction, neutral} (+ yield/guarantee/press rules)

Adversarial desk

lab

Bull, bear, arbiter — same inputs, reproducible.

3 reads(S) → {bull, bear, arbiter} — not a live 3-LLM panel

Cite and contest

live

Public files. A note you cannot recompute is marketing.

cite getauros.com/technology/methodology

Exploratory research, not live: zkTLS · entity graph · wash clustering · Bayesian changepoint · satellite / IoT

Solid = live. Dashed = lab. Dotted = research. Not a licensed CRA. Not a price oracle. Not a single LLM pass.

Why this is not “just AI”

  • Envelope = leave-one-dimension-out + residual spread. Deterministic. Same inputs, same band.
  • NLI live = labelled rules on structured fields (entail / contradict / neutral), not a chat completion you cannot replay.
  • Desk = three briefs from those metrics. A live multi-agent LLM would be a later overlay, never a substitute for the arithmetic.

Retrospective LODO coverage (T1)

auros.envelope_coverage.tip.v1 · method retrospective_lodo · calibrated_conformal=false · holdout labeled rows = 0

Did the next Gap edition land inside the prior LODO envelope? Cite-ready on Entity + Gap brief + hub edition tip. Template NLI densifies with claim sub-signals (yield / guarantee / press). Not split/CV+ conformal 95% until the holdout is non-empty.

auros.research_honesty.tip.v1 · holdout_n=0 · calibrated_conformal=false · zktls_live=false · adjacent=retrospective_lodo

Research map honesty (P4 cite) — never market 95% coverage or zkTLS settlement. Live adjacent = retrospective LODO only. See docs/AUROS-RESEARCH-HONESTY.md · GET /api/v1/intelligence → research_honesty.

Variance physical window (T3)

auros.variance_window.tip.v1 · silence live · history thin · HITL/mock · mainnet_custody=false · shared_lab

AUROS shared lab physical window as_of=2026-10-31 · fill=0.94 · health=94 · silence=live · collateral_wad=940000000000000000 · history=thin · settlement=hitl chain_write=mock mainnet_custody=false · shared_lab=true (not a market slug meter) · production_settlement=false · fused=false

Fact Plane agreement (T4)

auros.fact_agreement.tip.v1 · independence tiers/hosts/authority · posture thin/sourced/cross_checked/contested · not_a_verified_asset

Field×agreement matrix on free Entity. Conflicts stay listed — never averaged into a mega-score. D-tier press ingest stays blocked until dated independent sources exist.

Conformal holdout

Calibrated 95% coverage needs a labeled split (or CV+) set. The registry currently has 0 labeled rows. Until that number is non-zero, envelopes stay leave-one-dimension-out disagreement bands.

Settlement Reality

HITL path (4 steps) — not a mint oracle. See /technology#settlement-reality.

Formal mint-cap (Halmos)

AUROS formal mint-cap · Halmos ResourceOracle 3/3 · PorMintCore 5/5 · property totalSupply<=provenSum · Echidna CI · HITL/maxProof fail-closed · fused=false · production_settlement=false · not Certora · not production settlement claim

Symbolic Halmos proofs on the real UUPS+ECDSA mint path (P7 densify). Echidna runs in protocol CI. SMTChecker is a sweep only. On-chain fail-closed: provenSum · HITL · maxProof · frozen sole oracle. Not a custody or production settlement warranty · certora=false · mainnet_custody=false.

First-meter IoT path

AUROS first-meter path · profile=auros-industrial-json-v0 · MQTT EIP-191 · outbox · 6h window · HITL · real_device=false · invented_wh_volume=false · production_settlement=false · mTLS preferred · resource_layer tip · fused=false · not SCADA · not custody

Industrial JSON gateway profile maps Modbus/IEC/DNP3 registers → Wh, then the same signed MQTT production envelope as the lab simulator. Mosquitto lab only (loopback · auth/ACL · TLS · mTLS). Not a certified SCADA stack. real_device=false · production_settlement=false until a founder meter is connected. Outreach: docs/LINKEDIN-FIRST-METER.md · mailto resources@. Pairs with auros.resource_layer.tip.v1 (lab=true · production_settlement=false). Variance lab stays shared HITL/mock. No invented Wh. /builders/first-meter

Resource Layer (lab · HITL)

AUROS Resource Layer · lab=true · production_settlement=false · hitl=true · real_device=false · multi_resource_live=false · settlement_sku=false · mTLS preferred · first-meter → variance HITL/mock → formal mint cite · not SCADA · not custody · no invented Wh · fused=false

AUROS Resource Layer densify (P3 · Étape 22) — Energy stack stays LAB/HITL. lab=true · production_settlement=false · hitl=true · real_device=false · mainnet_custody=false · invented_wh_volume=false · scada_certified=false · multi_resource_live=false · settlement_sku=false. Path honesty: first-meter industrial JSON → MQTT EIP-191 → outbox → window → HITL (mTLS preferred · Mosquitto loopback lab only · not SCADA). Variance window tip = shared_lab HITL/mock. Formal mint-cap = Halmos/Echidna cite — not Certora · not settlement warranty. Lab demos (/lab · /producer · /trade · /watt) ≠ production custody. Gap ≠ Activity ≠ Variance · F1–F8 intact · F8 lab≠prod · no invented Wh. /resource-layer · /lab · /protocol/variance-oracle.

Technology hub · Resource Layer · Lab · Freshness · Activity · Promise Gap · Score envelope

+What is the AUROS intelligence stack?
The AUROS Compliance Score (/score) is the core product. Activity, Promise Gap, Freshness, Transparency and Friction are calculation annexes — each kept separate so desks can inspect drivers one by one.
+Do AUROS scores come with confidence intervals?
Live scores publish a leave-one-dimension-out envelope (disagreement band). Gap history also publishes a retrospective LODO coverage diagnostic (did the next edition land inside the prior band?). Calibrated conformal 95% stays research until a labeled holdout exists — holdout n=0 today. Cite auros.research_honesty.tip.v1 (calibrated_conformal=false · zktls_live=false).
+How does Promise Gap check claims?
Live contradiction checks use template NLI: entail / contradict / neutral between a yield claim and Activity, densified with claim sub-signals (yield mention / guarantee language / press amplify). Fine-tuned legal NLI stays on the research map.
+What is the bull/bear/arbiter desk?
The lab desk generates three briefs from the same published metrics so debate stays reproducible. Scores remain usable without a multi-model Copilot panel.
+What about zkTLS attestations?
zkTLS / zk-attestations sit on the research map for RWA disclosure — auros.research_honesty.tip.v1 locks zktls_live=false · zktls_settlement=false. The live desk today runs on published metrics, Activity overlays, and HITL settlement.
+Is Freshness a Bayesian changepoint detector?
Not yet. Freshness ships a pre-changepoint tip: cross-section cache-age regime (stable / aging / stale_heavy) and age vs class median. calibrated_changepoint stays false until a real issuer-clock series exists. Not a trade clock.
+What is the first-meter IoT path?
Lab path: industrial JSON profile (auros-industrial-json-v0) maps Modbus/IEC/DNP3 registers to Wh, then the same EIP-191 MQTT production envelope, durable outbox, window aggregate, HITL risk oracle. Prefer Mosquitto mTLS (loopback lab · not SCADA). real_device=false until a founder meter is connected — no invented volume. Track /builders/first-meter · tip auros.first_meter.path.tip.v1.
+Is the mint-cap formally verified?
Yes on the lab contracts: Halmos proves totalSupply ≤ provenSum on the real ResourceOracle UUPS+ECDSA path (3/3) and PorMintCore twin (5/5). Echidna runs in protocol CI. SMTChecker is a sweep only. See FORMAL_VERIFICATION.md — not a Certora badge, not a production settlement warranty. Tip: auros.formal.mint_cap.tip.v1.
+Is the Resource Layer production settlement?
No. auros.resource_layer.tip.v1 locks lab=true · production_settlement=false · hitl=true · real_device=false. Energy demos (/lab · /producer · /trade) and Variance HITL/mock stay labeled — F8 lab≠prod. No invented Wh volume. Cite /resource-layer · GET /api/v1/arl/resource-layer.
+What do AUROS scores measure?
Activity, friction, disclosure depth, and claim-alignment — preparation and post-issue life for RWA desks.
+How does Settlement Reality work?
Settlement Reality is a HITL path: signed physical packet, variance desk, human confirmation. Keeper updates run when an RPC is configured. First-meter industrial JSON feeds the same lab pipe when a real device exists.
+Where is the rating pipeline documented?
getauros.com/technology/methodology — five separated sources, leave-one-dimension-out envelope, template NLI, lab desk on the same inputs. Settlement Reality is HITL.

AUROS

Intelligence de tokenisation pour actifs réels.

Jurisdiction posture: EU-facing operations · imprint & entity pack on written request via /legal · /press
resources@getauros.com · legal@auros.app

AUROS LedgerEnergy Lab

Desk

Signaux unfused · journal

  • AUROS Ledger
  • Promise Gap
  • Activity
  • Variance
  • Freshness
  • Brief 1 page
  • Custom desk pack

Lab

Énergie · compteur · HITL

  • Energy Lab
  • First meter
  • Producer
  • IoT & contrats
  • Resource Layer

Issuer

Score · juridictions · prep

  • Compliance Score
  • Starter Kit
  • Juridictions
  • Tarifs
  • Pilotes 30 jours
  • AUROS Shield

Plus

Entreprise · formation · présence

  • À propos
  • Investisseurs
  • Presse
  • Academy
  • Partenaires
  • Status
  • Blog
  • Mentions légales
  • CGU
  • Confidentialité
  • Confiance
  • Sécurité
  • À propos

© 2026 AUROS · Tous droits réservés.