What GCTS-1 Governs
GCTS-1 is the technical measurement standard for The Green Code framework. It specifies the measurement methodology that operators must follow when calculating the Cr (Cost/Resource) denominator and the data inputs required for Bv (Benefit Value) attribution under the NBR formula.
GCTS-1 governs physical resource measurement. It does not govern:
- ✗ The economic valuation of benefits (that is NBR territory)
- ✗ Carbon pricing or emissions trading
- ✗ Worker safety, labor standards, or human rights provisions (see Rights Safeguards)
- ✗ Model performance benchmarking or AI quality metrics
- ✗ Cybersecurity or data privacy requirements
GCTS-1 applies to any compute deployment that elects or is required to calculate an NBR under The Green Code framework, regardless of deployment size. The rigor of audit requirements scales with deployment size as defined in §8.
Training Measurement Categories
When a deployment includes AI model training workloads (as distinct from inference serving), the following categories must be measured separately and reported as a distinct training-period subtotal within Cr.
| Category | Measurement Unit | Measurement Method | Status |
|---|---|---|---|
| Training energy — compute | MWh | Measured at hardware level via IPMI power sensors or equivalent, aggregated over full training run duration. Not extrapolated from FLOP counts. | PROPOSED |
| Training energy — cooling | MWh | Measured at facility cooling plant during training window. If shared infrastructure, allocated by training-period fraction of total load. | PROPOSED |
| Training water — cooling | kL | Water consumed by cooling systems during training window. Measured at cooling tower or direct-liquid-cooling loop meter. Evaporative loss and blowdown must be itemized separately. | PROPOSED |
| Training duration | hours | Clock time of training run from first GPU activation to final checkpoint write. Required for per-unit normalization. | PROPOSED |
| Training hardware count | GPU·hours | Number of GPUs (or equivalent accelerators) × hours active. Required for scaling analysis and embodied cost attribution. | PROPOSED |
| Failed/wasted runs | MWh, kL | Energy and water consumed by training runs that were abandoned, failed, or discarded. Must be included in total, not excluded as "non-productive." | PENDING METHODOLOGY |
Inference Measurement Categories
Inference (serving) workloads represent the ongoing operational resource cost of a deployed model. These are the primary Cr inputs for deployments that consume, rather than train, AI models.
| Category | Unit | Measurement Method | Status |
|---|---|---|---|
| Inference energy — compute | MWh | Measured at hardware level over full reporting period. Must include both active-inference and idle-waiting states. See §6 for idle treatment. | PROPOSED |
| Inference energy — cooling | MWh | Measured or allocated from facility total cooling load by inference-period fraction. | PROPOSED |
| Inference energy — networking | MWh | Energy consumed by network switching and interconnect hardware attributable to inference traffic. If shared with non-AI workloads, allocated by traffic volume. | PENDING METHODOLOGY |
| Inference water — cooling | kL | Water consumed during inference operations. Evaporative loss and blowdown itemized separately from drawn volume. | PROPOSED |
| Request volume | requests/period | Logged by load balancer or API gateway. Required for per-request normalization and efficiency trend analysis. | PROPOSED |
| Energy per request (EPR) | Wh/req | Computed metric: (total inference energy) ÷ (total requests served). Reported as a rolling average per reporting period. Not the primary Cr input but a required disclosure. | PROPOSED |
Energy System Boundary
The GCTS-1 energy boundary defines what is included and excluded from energy measurement. The boundary is set to capture all electricity that the deployment causes to be consumed in the host region.
- All IT hardware: GPUs, CPUs, memory, storage, networking hardware
- Cooling systems: chillers, cooling towers, CRAC units, direct liquid cooling loops
- Power conversion and distribution: UPS systems, PDUs, transformers within facility boundary
- Facility support systems: lighting, security, building management systems proportional to AI-designated floor area
- Standby and idle power: hardware that remains energized between active workloads (see §6)
- Employee office space, cafeteria, or parking facilities outside the compute facility footprint
- Energy consumed by end users accessing the deployment's services (user-side consumption is not attributed)
- Energy consumed in upstream manufacturing of hardware (addressed separately in §7 as embodied cost)
- Energy consumed at off-site data centers or cloud regions not physically within the host municipality boundary
PUE Requirement
Power Usage Effectiveness (PUE) must be measured and disclosed for the reporting period. PUE is defined as total facility energy divided by IT equipment energy. It must be measured at the utility interconnect, not estimated from design parameters. PROPOSED
PUE is a disclosure requirement under GCTS-1. It is not itself a compliance criterion, but a falsified or unsupported PUE disclosure constitutes a measurement failure that invalidates the Cr calculation.
Water System Boundary
Water measurement under GCTS-1 distinguishes between water drawn, water consumed, and water returned. All three must be reported separately.
| Term | Definition | Measurement | Status |
|---|---|---|---|
| Water Drawn | Total volume extracted from any source (municipal supply, well, surface water) entering the facility cooling system. | Measured at intake meter. Municipal utility billing records acceptable with ±5% verification. | PROPOSED |
| Water Consumed | Volume of water that does not return to the source watershed in usable form. Includes evaporative loss from cooling towers and water incorporated into products or waste streams. | Calculated as: Water Drawn − Water Returned (measured at discharge meter). Evaporative loss from cooling towers estimated by mass-balance where metering is not available. | PROPOSED |
| Water Returned | Volume returned to the source watershed in any form (discharge, treated wastewater, blowdown to sewer). | Measured at discharge meter. Must include quality characterization (temperature, dissolved solids) where returned water affects source quality. | PROPOSED |
| WUE (Water Usage Effectiveness) | Volume of water consumed per unit of IT energy used. | Calculated as: Water Consumed (liters) ÷ IT Energy (kWh). Reported as a disclosure metric. Target WUE is a PROPOSED design goal, not a verified achievement. | PROPOSED |
Aquifer-Specific Requirements
Deployments drawing from stressed aquifer systems — defined as aquifers where current withdrawal rate exceeds recharge rate — must additionally report:
- Aquifer stress classification — the current drawdown rate relative to recharge rate, sourced from USGS or equivalent state geological survey, dated within 24 months.
- Water source attribution — documented share of water drawn from aquifer vs. surface sources vs. municipal supply.
- Aquifer impact model — a documented model of the deployment's long-run impact on the aquifer water table, reviewed by a licensed hydrologist, with assumptions made explicit.
The Texas County pilot operates in the Ogallala Aquifer system, which is confirmed to be in a stressed drawdown state. Any live deployment in that region triggers aquifer-specific reporting requirements. AQUIFER STRESS: VERIFIED
Idle Overhead Treatment
Hardware that is powered on but not actively serving workloads still consumes energy and, in some configurations, cooling water. GCTS-1 requires that idle overhead be measured and included in Cr, not excluded as non-productive consumption.
- All-hours measurement: Energy and water metering must cover 24 hours/7 days/365 days (or full reporting period). No exclusion of off-peak or maintenance windows.
- Idle fraction disclosure: Operators must report the fraction of total energy consumed during idle periods (no active inference or training workloads). High idle fractions are a signal that the deployment is overprovisioned relative to its actual workload.
- Planned downtime inclusion: Scheduled maintenance windows must be included in energy accounting unless hardware is fully de-energized and documented as such.
- No headroom exclusion: Installed capacity reserved for peak demand but not currently utilized must be included in idle measurements if hardware is powered on.
Hardware Lifecycle and Embodied Cost
Embodied carbon and resource cost refers to the energy, water, and material consumption associated with manufacturing the hardware the deployment uses, before that hardware is powered on. GCTS-1's approach to embodied cost is as follows:
Current Requirement (Proposed)
Operators must disclose the hardware inventory (model, quantity, manufacture date) and make a good-faith effort to source published embodied carbon figures for the hardware from the manufacturer or from published lifecycle assessment databases. This disclosure is required but does not directly enter the Cr calculation at this time. PROPOSED
Target State (Pending Methodology)
End-of-Life Treatment
Operators must document the intended disposal pathway for hardware at end of life. Hardware returned to the manufacturer's take-back program, processed by a certified e-waste recycler (R2 or e-Stewards certified), or refurbished for reuse may be disclosed as a Bv item under the Resource Recovery category. The Bv attribution methodology for hardware recovery is PENDING METHODOLOGY.
Audit Thresholds by G-Level
< 5 MW
- Annual self-reported disclosure
- GCTS-1 measurement plan filed with municipality
- No third-party audit required
- Data retained for 3 years
- Complaint-triggered inspection may apply
5 – 20 MW
- Annual third-party audit required
- Auditor must be accredited (criteria TBD)
- Full audit report publicly available
- Data retained for 5 years
- Municipality receives audit summary
20 – 100 MW
- Semi-annual third-party audit required
- Continuous metering required at utility interconnect
- Full audit reports publicly available
- Regulator receives data files
- Data retained for 7 years
> 100 MW
- Real-time telemetry to public dashboard
- Annual comprehensive audit
- Independent monitor appointed
- Quarterly board reporting
- Data retained for 10 years
The Texas County proposed 50 MW deployment falls in the G-2 classification. G-2 audit requirements would apply if the pilot advances to active deployment.
Required Evidence Artifacts
The following evidence artifacts must be produced, retained, and made available for audit. "Available for audit" means transmittable to an accredited auditor within 5 business days of request. Public availability requirements vary by artifact type and G-level.
-
EA-01
Utility interconnect meter records
Time-stamped electricity consumption data from utility billing or smart meter API, covering the full reporting period at minimum monthly resolution. Daily resolution required for G-2 and above.
PROPOSED -
EA-02
Water meter records
Intake and discharge meter readings at minimum monthly resolution for the full reporting period. Source identification (municipal, well, surface) required.
PROPOSED -
EA-03
Hardware inventory manifest
Complete list of compute hardware (make, model, quantity, manufacture year, deployment date) and cooling infrastructure.
PROPOSED -
EA-04
Measurement plan
Documented methodology for every measurement category used in the NBR calculation, including sensor specifications, calibration records, data collection process, and chain-of-custody for meter data.
PROPOSED -
EA-05
Pre-deployment baseline report
Regional resource state characterization (energy grid conditions, water table, aquifer stress, local economic baseline) before deployment began. Independent review required for G-1 and above.
PROPOSED -
EA-06
Bv attribution records
Source documentation for each claimed benefit item. See NBR §2 for attribution requirements. Required artifacts by benefit type: contracts, invoices, payroll records, utility agreements, procurement reports.
PROPOSED -
EA-07
PUE and WUE disclosure
Computed PUE and WUE values with supporting calculation workbook showing numerator and denominator values and their sources.
PROPOSED -
EA-08
Conflict-of-interest declaration
Signed declaration from any party involved in data collection, analysis, or reporting of any financial interest in the deployment.
PROPOSED -
EA-09
Auditor report (G-1 and above only)
Signed findings from accredited third-party auditor. Must include: scope of work, methodology, exceptions and qualifications, summary NBR finding, and auditor's independence attestation.
PROPOSED -
EA-10
Aquifer impact model (stressed aquifer deployments only)
Hydrologist-reviewed model of long-run aquifer impact. See §5 for requirements.
PROPOSED
Data Freshness and Update Expectations
| Data Type | G-0/G-1 Frequency | G-2 Frequency | G-3 Frequency |
|---|---|---|---|
| Energy consumption totals | Annual | Monthly | Real-time (15-min intervals) |
| Water consumption totals | Annual | Monthly | Daily |
| PUE disclosure | Annual | Semi-annual | Monthly |
| WUE disclosure | Annual | Semi-annual | Monthly |
| NBR calculation | Annual | Semi-annual | Annual (supported by real-time sub-metrics) |
| Bv attribution records | Annual | Semi-annual | Quarterly |
| Hardware inventory | At change + annual | At change + semi-annual | At change + quarterly |
| Aquifer stress reference data | Every 24 months | Every 12 months | Every 6 months |
| Auditor report | Annual (G-1 only) | Semi-annual | Annual |
Data that exceeds its maximum age without refresh must be marked as stale in any public disclosure. Stale data may not be used to support an active NBR claim.
Gap 1 — 10:1 NBR Measurement Protocol PROPOSED
The GCTS-1 10:1 mandate requires every compliant Root-Node or model to demonstrate a 10-to-1 planetary return on investment. For every 1 unit of resource consumed for compute, the system must recover at least 10 units of verified planetary benefit. This section defines the exact formula, the physical performance gate, and the anti-gaming rules that enforce this mandate.
Physical Performance Gate — Hardware Attestation
The 10:1 ratio must be grounded in measured physical data, not modeled projections. The physical performance gate requires that all resource measurements be attested by hardware-level industrial telemetry before any NBR claim is computed.
- PLC-01 Siemens PLC (or ISO-equivalent) integration required — Energy and water meters must output readings to a programmable logic controller with a cryptographically signed data hash. Acceptable PLC families: Siemens S7-1500, Allen-Bradley ControlLogix, or equivalent ISO-certified industrial controllers.
- PLC-02 Data hashing and chain of custody — Each telemetry record must be hashed (SHA-256 minimum) with a timestamp and device ID before transmission to the reporting ledger. Hash roots must be anchored to an immutable log (on-chain or write-once audit log).
- PLC-03 Calibration attestation — All meters must carry a current calibration certificate from an accredited laboratory (ILAC-accredited or equivalent), with calibration frequency no less than 12 months. Meter serial numbers and calibration dates must be included in every submitted measurement plan (EA-04).
- PLC-04 Tamper detection — Hardware sensors must include tamper-evident seals. Any evidence of physical access to sensor hardware between calibration events must be reported and invalidates the measurement period until re-calibration and attestation are completed.
Baseline Gaming Prevention Rules
Manipulating the counterfactual baseline is the primary vector for NBR inflation. GCTS-1 enforces the following anti-gaming rules. Violation of any rule invalidates the NBR claim for the reporting period and triggers a mandatory corrective audit.
| Rule | Requirement | Violation consequence |
|---|---|---|
| BG-01 | Baseline must be set from pre-deployment regional data (< 24 months prior to first operation), sourced from independent government or utility records. Operator may not self-set their own baseline. | NBR invalidated; mandatory re-audit |
| BG-02 | Counterfactual scenarios must be sourced from published academic or government models. Operator-generated counterfactuals require independent review and are marked PROJECTED. | Claim downgraded; Bv disputed |
| BG-03 | Benefit attribution is capped at the deployment's directly measurable causal contribution. Systemic benefits require a licensed environmental engineer to quantify and sign the attribution model. | Unattrested Bv items excluded |
| BG-04 | Promised future benefits may not be included in current-period NBR unless secured by a binding municipal contract. Projected benefits must be labelled PROJECTED and excluded from the binding NBR calculation. | Claim downgraded to PROJECTED |
| BG-05 | Operators may not amend a submitted measurement plan retroactively to increase Bv or decrease Cr for a closed reporting period. Amendments take effect only in future periods from the date of approved amendment. | Amendment rejected; original values stand |
Gap 2 — Tokenomics & Reserve Architecture PROPOSED
- ✓Internal accounting conversion factor
- ✓Used only for redemption at participating municipal facilities
- ✗Does NOT represent the market price of water
- ✗Does NOT constitute a financial instrument peg to CME or any exchange
| CME NQH2O | ≈ $0.24 / 1,000 L | MARKET |
| SABESP (São Paulo) | $0.76 – $3.30 / 1,000 L | TARIFF |
| WBT Reference Rate | $0.10 / 1,000 L | INTERNAL |
Physical Abundance Reserve (PAR) — State Machine
The PAR is the reserve control mechanism that governs WBT issuance and redemption. The solvency gate (CRpost ≥ 1.50) must be satisfied before any new WBT can be minted. If the post-mint Coverage Ratio falls below 1.50, the mint transaction is rejected.
Gap 3 — Cryptographic Implementation — ZKP Circuit Specification PROPOSED
The WBT mint path requires a zero-knowledge proof (ZKP) that the device measured W = 10 × E (water benefit = 10× energy cost) without revealing proprietary load curves or raw sensor data. The chosen architecture is Halo2 (PLONKish) over vanilla PLONK for superior flexibility in range checks, elliptic-curve operations, and nullifier handling.
- →Custom gates and lookup tables for efficient range checks (enforcing
W / E = 10in fixed-point arithmetic without overflow) - →Native support for elliptic-curve operations (secp256k1 or Pasta curves) required by ECDSA/EdDSA device signatures
- →Poseidon or Sinsemilla-compatible hash gates for efficient nullifier construction inside the circuit
- →Recursive proof composition supported if cross-device aggregation is required in later versions
- →Mature Rust tooling via the
halo2crate; known vulnerability surface (Frozen Heart / under-constrained circuits) documented and mitigated below
// Step 1: Range checks on witnesses (Halo2 lookup table) assert E ∈ [0, MAX_ENERGY] // prevents overflow in W computation assert W ∈ [0, MAX_WATER] // range-check via Halo2 lookup gate // Step 2: Enforce the 10:1 ratio (PLONKish custom gate — no float) assert W == 10 * E // integer-scaled fixed-point // Step 3: Bind mint_amount to proven W value assert mint_amount == W / SCALE_FACTOR // WBT = litres / 1000 // Step 4: Verify device signature over (device_pubkey, E, W, epoch) assert ECDSA_verify(device_pubkey, msg_hash, sig_r, sig_s) == true // Step 5: Construct nullifier to prevent epoch replay assert nullifier == Poseidon(device_sk, epoch)
Known Failure Modes & Mitigations
- T-01Frozen Heart / weak Fiat-Shamir: Bulletproof-style vulnerabilities where the challenge is not properly bound to all public inputs. Mitigation: use Halo2's built-in transcript protocol; do not use vanilla PLONK's Fiat-Shamir without audited binding.
- T-02Under-constrained circuit: Prover supplies a W witness that satisfies constraints arithmetically but does not correspond to a real sensor reading. Mitigation: require independent meter attestation (§PH-1 PLC-01) and cross-check mint_amount against PLC hash log before accepting proof.
- T-03Nullifier reuse: Same epoch proof submitted twice. Mitigation: on-chain nullifier registry; any duplicate
nullifiervalue causes immediate mint rejection.
Gap 4 — Security & Smart Contract Audit Registry
Reentrancy Guards — Mint Path (Checks-Effects-Interactions)
The mint path is the highest-risk function in the WBT contract. The Checks-Effects-Interactions (CEI) pattern plus a nonReentrant modifier are required:
// Pattern: Checks-Effects-Interactions (CEI) function mint(proof, publicInputs) external nonReentrant { // 1. CHECKS — validate proof and solvency gate FIRST require(zkVerifier.verify(proof, publicInputs), "ZK proof invalid"); require(!nullifierRegistry[publicInputs.nullifier], "Nullifier reused"); require(par.coverageRatioPost(publicInputs.mintAmount) >= 150, "CRpost < 1.50"); // 2. EFFECTS — update ALL state before any external call nullifierRegistry[publicInputs.nullifier] = true; totalSupply += publicInputs.mintAmount; balances[msg.sender] += publicInputs.mintAmount; // 3. INTERACTIONS — emit event after all state changes emit WBTMinted(msg.sender, publicInputs.mintAmount, publicInputs.epoch); }
Audit & Verification Requirements
| Requirement | Detail | Status |
|---|---|---|
| Third-party audit | Full audit by a firm with ZKP contract experience (e.g., Trail of Bits, Zellic, Spearbit or equivalent). Scope must include mint path, ZK verifier integration, access controls, and upgrade logic. | PENDING AUDIT |
| Echidna fuzzing | Property-based fuzzing via Echidna targeting mint invariants: CRpost ≥ 1.50 must hold on all valid mint transactions; nullifier registry must be append-only; totalSupply must never overflow uint256. | PENDING AUDIT |
| Slither static analysis | Slither must be run against the complete contract suite. All high-severity findings must be remediated before testnet deployment. Medium-severity findings require documented acceptance or remediation. | PENDING AUDIT |
| Formal verification | Critical mint path invariants must be formally verified using Certora Prover, Halmos, or equivalent. Verified properties: reentrancy absence, solvency gate enforcement, nullifier uniqueness. | PENDING AUDIT |
| Public disclosure | All audit reports (Echidna runs, Slither output, formal verification certificates, third-party audit report) must be published in the Evidence Room before any mainnet WBT minting occurs. | PENDING AUDIT |
| Liveness & integrity checks | On-chain liveness oracle must confirm the ZK verifier contract has not been paused, upgraded, or replaced without governance approval. Integrity hash of deployed verifier bytecode must match audited artifact. | PENDING AUDIT |
Gap 5 — Lifecycle Resource Cap (LRC) — Scope 3 Embodied Carbon PROPOSED
GCTS-1's original operational boundary capped hardware power consumption at 20 W operational cap for compliant edge devices. The Lifecycle Resource Cap (LRC) upgrades this to a cradle-to-grave accountability framework that adds Scope 3 embodied carbon — manufacturing, extraction, and semiconductor fabrication — directly to the NBR denominator (Cr).
Scope 3 Embodied Carbon Categories
The following upstream Scope 3 categories are included in the LRC amortization and must be sourced from manufacturer lifecycle assessment (LCA) data or published industry databases (ecoinvent, IEEE LCA datasets):
| Category | Scope 3 source | Unit | Status |
|---|---|---|---|
| Semiconductor fabrication | TSMC fab energy (N5/N3 node), photolithography, chemical processing. Source: TSMC CSR sustainability reports or equivalent foundry LCA data. | kg CO₂-eq / die | PROPOSED |
| Raw material extraction | Mining of silicon, rare earth elements, copper, cobalt, and other substrate materials across global supply chain. LCA database required. | kg CO₂-eq / unit | PROPOSED |
| PCB manufacturing & assembly | Board fabrication, SMT assembly, testing, and packaging. Typically sourced from product environmental data sheets (PEDS). | kg CO₂-eq / unit | PROPOSED |
| Logistics & distribution | Transport from manufacturing site to deployment location. Freight mode (air/sea/road) and distance required for emissions factor selection. | kg CO₂-eq / unit | PROPOSED |
| End-of-life processing | E-waste processing, landfill, or recycling pathway. Certified recycler (R2/e-Stewards) reduces this term; unverified disposal uses conservative disposal factor. | kg CO₂-eq / unit | PROPOSED |
Cradle-to-Grave Amortization — NBR Denominator Integration
// Embodied carbon per reporting period (added to Cr denominator) E_emb_period = (C_fab + C_extract + C_mfg + C_logistics + C_eol) ÷ T_life // total hardware lifecycle (years) × U_frac // deployment's utilization fraction of hardware × P_frac // reporting period as fraction of year // Updated Cr denominator (replaces §7 disclosure-only requirement) Cr_lrc = Cr_operational + E_emb_period // Updated NBR with full LRC — must still satisfy 10:1 mandate NBR_lrc = Bv ÷ Cr_lrc // NBR_lrc ≥ 10 required for certification
C_fab — Semiconductor fab carbon (kg CO₂-eq)C_extract — Raw material extraction (kg CO₂-eq)C_mfg — PCB mfg & assembly (kg CO₂-eq)C_logistics — Transport & distribution (kg CO₂-eq)C_eol — End-of-life processing (kg CO₂-eq)T_life — Total hardware lifecycle (years)U_frac — Fraction of hardware utilized by this deploymentP_frac — Fraction of year covered by reporting periodGap 6 — Ratification & Signatory Pathway
GCTS-1 certification requires ratification by three independent signatory tiers. Each tier represents a distinct accountability domain. All three signatures are required — no tier may be waived. The current status of all three signature blocks is NOT RATIFIED / INTERNAL DRAFT.
Changelog
- §PH-1: 10:1 NBR Measurement Protocol — NBR = Bv/Cr formula glass panel, PLC telemetry attestation (PLC-01 to PLC-04 with Siemens S7-1500 / SHA-256 hash chain), anti-gaming controls (BG-01 to BG-05)
- §PH-2: WBT Tokenomics & Reserve Architecture — reclassified as redemption voucher at $0.10/1,000 L (NOT market peg; CME NQH2O ≈ $0.24; SABESP $0.76–$3.30), PAR state machine (SURPLUS/SOLVENT/CAUTION/INSOLVENT), CRpost ≥ 1.50 solvency gate formula
- §PH-3: ZKP Circuit Specification — Halo2/PLONKish over vanilla PLONK (rationale: custom gates, EC ops, Poseidon hash), public inputs (nullifier, device_pubkey, mint_amount, epoch), private witnesses (sig_r, sig_s, device_sk, E, W), W = 10 × E constraint pseudocode, threat vectors T-01 to T-03
- §PH-4: Smart Contract Audit Registry — prominent [PENDING AUDIT] status banner, CEI reentrancy guard pattern for mint path with nonReentrant modifier, audit requirements table (third-party, Echidna, Slither, formal verification, liveness integrity)
- §PH-5: Lifecycle Resource Cap (LRC) — upgrades 20W operational cap to cradle-to-grave, Scope 3 categories (TSMC fab, raw material extraction, PCB mfg, logistics, EOL), E_emb_period amortization formula, Cr_lrc = Cr_operational + E_emb_period
- §PH-6: Tripartite Ratification Pathway — three signature blocks (Tier-1 Municipal Authority, Independent Engineering Auditor, Green Code Foundation), all explicitly [NOT RATIFIED / INTERNAL DRAFT] with red lock badges
- Page-nav-strip updated with 6 new green-highlighted Protocol Hardening section links
- Changelog version bumped to v1.0-draft; footer version string updated
- Published training and inference measurement categories
- Published energy and water system boundary definitions
- Published idle overhead treatment rules
- Published evidence artifact list (EA-01 through EA-10)
- Published data freshness requirements by G-level
- Flagged all items as Proposed; pending sections marked explicitly
- First public release of reference PDF document
- G-level tier structure introduced
- Measurement categories listed without full methodology
- Standard concept introduced in design documents
- Not yet published for external review