🔒 Confidential Evaluation Sandbox
This is a non-production Chateauz prototype.
Enter the access password to continue.
Confidential Evaluation Sandbox · Non-Production Prototype · Synthetic/Illustrative Data · No Live Systems Connected
Chateauz v1.0.20
Cortex · [Infrastructure Program] · Scenario A
COGSENS
CORTEX
CINDRAL
COGSENS · Site Intelligence Scenario A
Read-only summary
✕
Performance output —
—
Power source mix
—
Structures
—
—
Decision Architecture · Protocol 1.0
Define your decision.
✕
Eight standardized fields. Under two minutes. The answers you provide here anchor every scenario, weight, and confidence score in CORTEX — and become the first record in your CINDRAL audit chain.
01 Project name & sector
Name this decision. The identifier that will appear in CORTEX, CINDRAL, and all sealed records.
02 Capital at risk
The committed capital this decision defends. Reflected live through CORTEX confidence scores as a dollar exposure figure.
$ M
03 Performance commitment
What you have promised to deliver. Sets the physics floor that CORTEX cannot override.
04 Decision deadline
The forcing function — NTP, FID, or equivalent. This is the date the decision must be defensible by.
05 Primary vendor · technology status
Seeds the Tech Lead Time baseline in CORTEX stress conditions.
06 Customer commitment status
Pre-sets the Commitment Gap and Anchor conditions in CORTEX.
07 Constraint classification
Determines the physics floor structure in CORTEX.
08 Primary uncertainty source
Your single highest uncertainty. Seeds the contested track in CORTEX — the question CINDRAL is designed to protect against.
08b Secondary uncertainty source (optional)
A compounding uncertainty layer. Flagged in the simulation but does not bifurcate the model.
09 Decision defensibility threshold
Your minimum acceptable Decision Confidence rating. The floor below which no scenario is considered defensible — to your board, your capital partners, your regulators. This is your standard, not ours.
10 Decision dimension priorities
Set the relative importance of each dimension to your project. These weights drive how CORTEX calculates your Decision Confidence score.
Inputs anchor this CORTEX session and are sealed in CINDRAL when you accept a decision.
Decision Architecture
↗
Scenarios
A
Single Asset
Recommended
78%
B
Multiple Assets
Below threshold · T1
24%
C
Customer-Led
Conditional
58%
D
Infrastructure-First
Capital structure req.
41%
E
Licensable Model
Long-horizon · T1 unresolved
19%
View spatial confidence
Scenario A
Prove It First — Single Asset, Full Execution Focus
Concentrate on highest-confidence first asset. Validate the model, then expand.
All 5 sources pass integrity check. Physics floor confirmed. Decision defensible at full confidence.
78%
Decision Confidence
Defensible —
Exposure
Decision weights ✕
Simulated ✕
Input integrity →
5-dimension scoring · Scenario A
82
Capital allocation
Concentrated · highest reversibility
75
Performance
Delivery feasibility · physics cleared
80
Risk
Lowest complexity · Tier 1 survivable
68
Revenue
Deferred upside · model validates
88
Decision efficiency
Clear owner · single anchor
Scenario A clears the confidence floor across all five dimensions.
Session log · 3 entries →
Re-running Scenario A
Applying conditions against physics layer
0
Simulation fidelity
1,000 runs · Light pass
10,000 runs · Standard
100,000 runs · Regulatory
Something we haven't modeled? + Add
Decision record · CINDRAL
Accept Simulation
Review the record before sealing. This action is irreversible.
Scenario A — Single Asset
Confidence score 78%
Score type Baseline · unmodified simulation
Conditions All at baseline
Decision weights Decision Architecture defaults
Timestamp
I confirm this decision is ready to seal. I understand that the sealed record is immutable and will be logged to the CINDRAL audit chain with a cryptographic hash.
⬡
Decision sealed
Session log
✕ Close
No session entries yet.
Input integrity
✕ Close
Input sources
Geospatial
97
Fresh · verified
Weather
82
⚠ Aging · 168d
ORNL
99
Stable · physics
Physics rules
100
Deterministic
Hitachi · specs
91
Fresh · 129d
Physics rules · primary · hard constraints confirmed
Data conditions
Primary decision · verified inputs ⚠ Under contested data conditions
78%
Scenario A · Prove It First
All 5 sources pass integrity check. Physics floor confirmed. Decision defensible at full confidence.
Capital
82
Performance
75
Risk
80
Revenue
68
Decision eff.
88
71%
Scenario A · holds with disclosure
Weather source aged 168d. Delta +2.1% vs ORNL baseline. Scenario A still holds but Performance and Risk dimensions carry wider uncertainty bands.
Capital
82
Performance
64
Risk
68
Revenue
68
Decision eff.
88
Both tracks sealed in CINDRAL audit chain. Decision owner may commit on either track with full disclosure of data confidence conditions.
A · Prove it first
B · Portfolio parallel
C · Customer-led
D · Infrastructure-first
Asset placement ▾
Energy mix simulator ▾
Physics constraint
N
S
W
E
↖
↗
↙
↘
⊕
+
−
orbit
Cortex · live scoring
Scenario A · asset placement drives simulation
78%
Composite
Holds · defensible
Capital allocation
82
Performance
75
Risk
80
Revenue · native
68
Decision efficiency · native
88
Performance commitment target
MW
Click to set your commitment target
Weather stress Baseline
Baseline −60°F
Delivered vs target —
Add energy sources to begin
Nameplate capacity
—
Wake loss
—
Weather derating
—
Constraint loss
—
Defensible delivered
—
Physics constraints · real-time
Grid interconnect boundary
Clear
Thermal exclusion zone
Clear
Environmental permit boundary
Clear
Weather exposure zone
Nominal
◉ Scenario A holds · current placement
Asset within optimal zone. Tie-in distance minimizes capital exposure. All physics constraints clear. Move the asset to explore constraint boundaries.
Trust layers
01
Input Trust
Source data registry
02
Process Trust
Pipeline dashboard
03
Output Trust
Audit trail · Decay signal
01 Input Trust
● Real · Acquired
ESRI Geospatial
Acquired2025-03-14
Rows2,847,203
CoverageSite + 50mi corridor
Time range2018–2025
Data score94
● Fresh · 62 days
37.8044°N · elev 142m · slope 2.3° · grid dist 4.1mi
● Real · Acquired
Wood Mackenzie · Load Curves
Acquired2024-09-01
Rows438,760
CoverageWECC · hourly
Time range2020–2024
Data score71
⚠ Aging · 259 days
2024-08-14 14:00 · 847 MW · peak · WECC-SW zone 3
◆ ORNL · Validated Physics
ORNL Thermal / Grid Priors
Validated2024-11-20
Rules1,847 constraints
TierPhysics floor · immovable
Versionv3.2.1
Data score99
● Validated · National lab anchor
Thermal excl: Δtemp >8°C/100m · FERC Order 2023 §4.2
● Real · Acquired · Tier 1
NOAA · Climate Baseline
Acquired2024-12-01
Rows3,102,400
Coverage30-yr baseline · hourly
Time range1994–2024
Data score82
⚠ Aging · 168 days · Tier 1 risk
2024-11-30 · -4°C · wind 22kn · precip 0.3in · zone 4b
● Real · Operational
Hitachi · Equipment Specs
Acquired2025-01-09
Rows12,440
CoverageTransformer · switchgear
Time rangeSpec · static
Data score91
● Fresh · 129 days
Unit T-44A · 230kV · 500MVA · ONAN/ONAF · 99.2% eff
◈ Synthetic · Simulation-generated
Edge Case Set · Run 10K-A
Generated2025-05-18 · this run
Samples10,000
ScenarioA · T1 High stress
MethodMonte Carlo · ORNL-bounded
Within physics bounds✓
◈ Synthetic · session-scoped · not persisted
T=-20°F sustained · lead +180d · demand +23% · beyond historical
🔐 Data Sovereignty
✓ Customer-controlled ⬡ Chateauz IP boundary enforced
✓
Data residency
Session data lives in the customer environment by default. Chateauz cloud and sovereign deployment available as configuration. No data leaves the designated boundary without explicit customer authorization.
✓
Retention policy
Session inputs are purged once the output is sealed and the audit chain is exported. Retention period is customer-configurable down to zero. Nothing held beyond the session boundary without explicit instruction.
✓
Access controls
No Chateauz personnel can access session inputs. Isolation is architectural — not policy-based. The platform is built so operator data is unreadable by the vendor by construction, not by promise.
✓
Export ownership
The sealed CINDRAL audit chain belongs entirely to the customer. Chateauz retains no copy. The operator owns their decision record — scores, provenance, integrity hashes — in perpetuity.
◈
Synthetic data scope
Synthetic edge cases are generated within the session, scoped to that run only. Not persisted. Not shared. Not used to train or update Chateauz models. "Session-scoped · not persisted" is a data sovereignty guarantee — not just a label.
⬡
Chateauz IP boundary
Customer exports contain decision provenance, scores, and audit chain only. Decision Architecture methodology, ORNL physics prior integration, decay model, and test harness weights are retained by Chateauz — never included in exports. The boundary runs both directions.
Nothing crosses the boundary in either direction without explicit authorization · Architecture-enforced · not policy-dependent
02 Process Trust
⬡
Deterministic
Hard rules · no uncertainty · ORNL-validated
FERC constraint check
Interconnect queue · Order 2023 §4.2 · hard pass/fail
PASS · no uncertainty
Thermal exclusion zones
ORNL thermal gradient rules · site boundary enforcement
PASS · 0 violations
Load flow physics
Power flow equations · Kirchhoff · voltage/current limits
PASS · within tolerance
Tier harness100% · no probabilistic component
◎
Probabilistic
Monte Carlo UQ · quantifiable uncertainty
Monte Carlo · 10,000 runs
Sampled against ORNL physics floor · input perturbation within bounds
σ = 4.2 · well-constrained
Uncertainty quantification
P10/P50/P90 output distribution · tail risk identification
P10: 61% · P50: 78% · P90: 89%
Synthetic edge generation
Beyond-historical scenarios · ORNL-bounded · flagged synthetic
Within physics bounds
Tier harness94.7% · distribution validated
◈
Inferential
Model judgment · explicit validation required
Scenario path scoring
Dimension weights from Decision Architecture stakeholder input · not hard-coded
Test harness: 87.3%
Condition impact vectors
T1/T2 weights from prior Decision Architecture sessions · transfer validated
Test harness: 82.1%
Decision output synthesis
Natural language grounded in scored dimensions
Test harness: 96.4%
Tier harness88.6% · inferential · explicit validation
03 Output Trust
A
B
C
D
E
Immutable Audit Trail 6 steps · SHA-256 chained
INGStep 00a
09:14:28 UTC
Data source ingestion · 5 feeds acquired
ESRI geospatial (97) · NOAA weather (82 ⚠ aging) · ORNL physics priors (99) · Physics rules engine (100) · Hitachi equipment specs (91). Composite input integrity: 94/100.
sha256: 9c41d7f2a08b3e15… · ingestion manifest sealed
INGStep 00b
09:14:29 UTC
Data confidence scoring · delta analysis
Cross-source consistency check run. Weather / climate source flagged: +2.1% variance vs ORNL thermal baseline. Age threshold exceeded at 168 days. Contested track activated.
sha256: e7b20c4f1d93a861… · confidence scores sealed
INGStep 00c
09:14:30 UTC
Physics consistency validation · pre-inference
All 5 inputs checked against ORNL hard physics floor before simulation. 4 of 5 pass clean. Weather input passes physics bounds but carries contested confidence flag. No inputs blocked — contested track carries the disclosure forward.
sha256: 3a58e1c9f20d7b44… · physics pre-check sealed
INGStep 00d
09:14:31 UTC
Simulation input manifest locked · dual-track armed
Primary track: all 5 inputs at face value. Contested track: weather input uncertainty bands widened per delta score. Both tracks will run in parallel. Decision owner receives both outputs with full data provenance.
sha256: 6f19d2c8b30e7a52… · input manifest sealed
DETStep 01
09:14:32 UTC
Decision Architecture session ingested
5 stakeholder groups · 5 conflicting capacity estimates · decision question locked. Client-owned under data sovereignty.
sha256: a3f8c2d1e9b74052… · inputs sealed
DETStep 02
09:14:33 UTC
ORNL physics constraints applied
1,847 rules evaluated · 0 violations · thermal + FERC floor locked before UQ layer.
sha256: 7c14e8a2f3d91b06… · constraint state sealed
PROBStep 03
09:14:41 UTC
Monte Carlo · 10,000 runs
σ=4.2 · P10/P50/P90: 61/78/89% · 847 synthetic edge cases generated and flagged as synthetic.
sha256: 2e97b4c8d1f03a51… · distribution sealed
INFStep 04
09:14:42 UTC
Scenario scoring · 5 dimensions
Decision Architecture weights applied · test harness 87.3% · All 5 scenarios scored across all 5 dimensions. Composite scores sealed.
sha256: f1d4a9e7b2c38050… · scores sealed
PROBStep 05
09:22:17 UTC
T1 stress re-run · lead time → High
10,000 re-runs · All 5 scenarios re-scored under T1 High. Deltas logged per scenario. Stress state sealed.
sha256: 8b30c2f1e4d79a06… · stress state sealed
DETStep 06
09:22:19 UTC
Decision output finalized · sealed
Scenario A highest-scoring path across all conditions · full session chain sealed · exportable to DOE reviewer or capital committee.
sha256: 3d9f7c1a8e402b55… · output sealed · chain complete
Integrity verification
✓
Chain integrity · 6 steps · unbroken
Verified
✓
Physics floor · ORNL-anchored · 0 violations
Confirmed
✓
Synthetic boundary · explicit · 847/10,000 flagged
Confirmed
⚠
Decision confidence decay · 62 days since run
Aging · re-run recommended
✓
Client data sovereignty enforced
Confirmed
Decision Confidence Decay ⚠ Aging
71%
Current confidence
Original: 78% · 2025-05-18
Decay driven by 3 active factors.
Refresh recommended within 14 days.
Decay drivers
Weather baseline
−4.1% · Fast
FERC queue position
−1.8% · Med
Wood Mac load curves
−1.2% · Med
ORNL physics priors
−0.1% · Stable
Recommended: Refresh NOAA weather baseline and re-run Monte Carlo. Estimated confidence recovery to 76–79%. ORNL physics floor unchanged — deterministic outputs remain valid.
Data lineage
1
Decision Architecture stakeholder interviews
Decision question captured Week 1. Dimension weights calibrated to client constraints.
2
Five system outputs
ESRI · Paces · Hitachi · GE ADMS · Wood Mackenzie. Divergence preserved as input uncertainty range.
3
ORNL national lab physics priors
External validation. Load growth distribution and failure mode library. v3.2.1.
4
Physics-informed constraint engine
FERC rules, thermal limits, load flow. Hard boundaries before probabilistic layer.
5
Monte Carlo UQ · 10,000 runs + 847 synthetic
Probabilistic layer on ORNL substrate. Synthetic edge cases flagged and scoped to session.
6
5-dimension scoring · all 5 scenarios
All 5 scenarios scored. Scenario A only path above 65 on all five simultaneously. Full session sealed.