The Global Risks Index
NXS-GRX · Methods · model registry 1.0.0

Every risk method, standardized

670 methods across hazard science, exposure, vulnerability, loss, catastrophe, financial, sovereign, systemic, climate, health, cyber and assurance, each with a stable ID, an honest implementation state and the standards that govern it.

Page GRX-MTHEdition Baseline Edition 2026Evidence close 2026-09-30Values not yet publishedSnapshot 31a753f76
670
methods
23
fabrics
169
kernel-bound
48
connected or reference
01Fabrics

23 method fabrics

Every quantitative method the index and its platform may use is registered once, in one fabric, with a stable capability ID; a title is display text only.

FabricCapabilitiesKernel-boundPlanesStandards
Universal mathematics and statistics3826julia, typescript
Hazard science377julia, python
Exposure and asset intelligence184typescript, python, sql
Vulnerability, fragility and resilience198julia, python
Consequence and loss186julia, python, r
Catastrophe risk2313julia, python, r
Networks, dependency and systemic risk2110julia, python
Financial risk5713julia, python, r
Insurance and reinsurance3814r, julia
Markets, trade and supply chains307julia, python
Sovereign, macroeconomic and development risk285julia, python, r
Policy, regulation and governance171python, julia
Water risk267julia, python
Food, agriculture and fisheries252python, julia
Energy risk267julia, python
Health and epidemiology259julia, python, r
Climate, nature and biodiversity243python, julia
Critical infrastructure and industrial risk276julia, python
Cyber, AI and frontier technology566python, rust, julia
Geopolitics, conflict and human security215python, julia
Projects, programmes and infrastructure investment272julia, python
Risk data240typescript, python, rust, sql
Scientific assurance4516julia, python, r, rust
02State

What is implemented

169 capabilities are bound to hand-written scientific kernels with golden-value tests in Python, Julia or R. Every capability is release-held, with a claim ceiling of integrated.

model registry 1.0.0 · 670 capabilities
Connected9
Reference implementation39
Adapter required327
Specialist model required159
Data required136
Kernel bindings by plane
python161
julia4
r4
Connected
An existing governed engine path is available; specialist physics still needs validation for the context of use.
Reference implementation
Transparent reference mathematics for diagnostics and parity, not institutional-grade calibration.
Adapter required
Registered; an execution adapter is still to be built. Requests fail closed.
Specialist model required
Needs a specialist domain model. Requests fail closed rather than fabricate an output.
Data required
Needs governed data that is not yet available. Requests fail closed.
03Contract

What every method must carry

The same contract applies to an NXS-GRX (canonical data, provenance and risk) index, a custom index in Index Studio and a model run in Model Lab.

Evidence
sealed RiskDataCut for data-grounded execution; source release and exact-record lineage; rights/purpose decision; chronotope; model/method version; uncertainty and applicability evidence
Uncertainty
aleatory where applicable, epistemic, parameter, model-form, data, scenario
Verification and validation
verification, calibration or justification, validation or benchmark, applicability, reproducibility
Outputs
uncertainty-packet, evidence-envelope, vvuq-receipt
Refusals
NO_DATA_IS_NOT_ZERO NO_UNGROUNDED_MONETARY_TRUTH NO_FAKE_SPATIAL_MATERIALIZATION NO_CAUSAL_CLAIM_WITHOUT_IDENTIFICATION NO_CALIBRATION_CLAIM_WITHOUT_EVIDENCE NO_AUTHORITY_ESCALATION
04Honesty

What registration does not mean

  • Capability registration is not model execution, calibration, observation, validation or assurance.
  • CONNECTED means an existing governed engine path is available; specialist domain physics still requires context-of-use validation.
  • IMPLEMENTED_REFERENCE means transparent reference mathematics exists for diagnostics/parity, not institutional-grade calibration.
  • SPECIALIST_MODEL_REQUIRED and DATA_REQUIRED fail closed rather than fabricate outputs.
  • Cryptographic integrity proves content/provenance relationships, not scientific truth.

Propose a method

New methods are registered with a stable ID, a hand-derived golden test and their evidence obligations before any page uses them.