Methodology
Variance — promised Wh versus attested Wh
Physical lab window only. Same closed-form math as AurosRiskOracle.sol. Not fused with Promise Gap or Activity. Not a multi-source quorum. Not certified SCADA.
Lab window only. Same closed-form math as AurosRiskOracle. Indicative pool read — not a live lending quote, not a mainnet mint. Settlement details live in the FAQ and machine brief.
Closed-form pipeline
- fill = attested_wh / promised_wh
- health = min(100, round(fill × 100))
- alert = fill < 0.85 (Solidity ALERT_GAP_WAD = 850000000000000000)
- collateral_factor = circuit_open ? 0 : min(1, fill)
- collateral_factor_wad = collateral_factor × 1e18
- max_quote_ltv = base_ltv(0.5) × collateral_factor
Constants (public)
- Alert threshold
- 0.85
- Base LTV
- 0.5
- Silence circuit
- 48h (172800000 ms)
- On-chain mirror
- AurosRiskOracle.updateAssetStatus / collateralFactorWad
What alert and silence are not
- fill < 0.85 raises alert — under-delivery vs the promised window, not a fraud verdict.
- 48h silence opens the circuit → factor 0 until a positive attested Wh restores it. Design choices can slow transfers; silence is a collateral cut, not a social alarm.
- Lab history stores the last eight windows for desk replay — not a market tape and not multi-oracle consensus.
Envelope (LODO)
Leave-one-out disagreement across fill · health · collateral_factor×100. Shows confidence band around health — not a calibrated 95% conformal interval.
Unfused stack
Promise Gap = claim intensity vs market life. Activity = post-issue transfers/depth. Variance = physical Wh on the lab window. Same asset, three reads, never one mega-score.
Desk · Pilot brief · Gap methodology · Activity methodology · Machine brief