Model versions
Once a carrier or a lender prices against a version of this model, that version is a commitment. A referenced score whose model can change silently cannot be relied on in a contract — so changes ship as new versions with an effective date and a changelog, and every issued score states which version produced it.
Every version, newest first
OCML 6GL + G4 fabric on the Implicit H3 Field. Occupancy, SDS OSM/CDM, ISO 8601 binding, PMD, radar fusion.
- Does not replace TIM v1.0 as TIA Score issuer.
- OCML is a sixth-generation language: INTENT and constraint blocks compile to G4 policy (file id $OML).
- OCML compiles Gap 1 (5-year PMD) and Gap 3 (derelict/crew/fuel right-of-way) onto IHF face clusters.
- G4 curvature K is projected mass-energy; leapfrog samples paths off the depression.
- CCSDS CDM KVN ingest and 3-sigma radar fusion lock.
- Space Data Standards OSM ($OSM) and CDM ($CDM) JSON/KVN/FlatBuffers file-id binding.
- ISO 8601 audits on TCA, CREATION_DATE, screening windows, and OSM PASS_START.
- Surfaces SDS $CCT capability classes as live vs declared orbital verticals.
Coupling analysis index — physical × financial × aggregate behavioural entanglement. Never issues TIA Scores and never self-executes.
- Does not replace TIM v1.0 as TIA Score issuer, and does not replace HL-IRRI as the place index.
- Gaussian entanglement 1 − det(R) on the 3×3 Pearson correlation of the three layers.
- Pairwise mutual information from published 5-bin histograms; decoupling→sync detector.
- Stylised core–periphery cascade; metric 2 is ∂shortfall/∂behavioural.
- Twenty named analogue trajectories by RMSE; GNN matching stays in the lab.
- Frozen out-of-sample backtests for 2008, 2011 Thailand, 2020, and 2022 energy.
Place-risk analysis index — H3 cells, threat × (1 − resilience), vertical sub-indices.
- Does not replace TIM v1.0 as TIA Score issuer.
- Eight threat dimensions and four resilience capacities, DAW local+frequency weights.
- Deterministic spatial diffusion and physics constraints (flood, grid).
- 12-month causal forecast with published coefficients.
- Merkle tick commitments on the Bureau transparency log.
Published analysis layer — adversarial path, controllability, provenance, and uncertainty.
- Promoted from shadow after calibration evidence met promotion threshold.
- Adds adversary timeline analysis (EASI weakest-path scoring).
- Controllability decomposition: inherited vs. earned risk split.
- Evidence provenance and manipulation sensitivity index.
- Score uncertainty with decision-safe score (10th percentile) and evidence acquisition plan.
- Does not replace TIM v1.0 composite — v1 remains the published TIA Score issuer.
First published version of the Threat Impact Model — TIA Score issuer.
- Six weighted layers: crime intelligence, environmental (CPTED), infrastructure, incident history, operational, community intelligence.
- Standard weighting 30/15/20/15/12/8; oil and gas sector profile re-weights to 18/17/23/20/17/5.
- Thirteen grade bands from A+ (97–100) to F (below 60).
- Data-confidence assessment that withholds credit the evidence does not support, rather than scoring thin data optimistically.
- Distance-squared decay on crime events to a 1,600m radius, with severity and age weighting.
A score is a record of what was issued
A score issued under an earlier version stays the record of what was issued, and says which version produced it. We do not retroactively restate scores when the model changes — that would rewrite a number somebody may already have acted on.
Assessments issued before the registry existed carry no stamp and are labelled unversioned. They are not back-dated into a version they never ran under, because a stamp we invented after the fact would be worth less than the honest gap.
Read the model itself
Layer weights, sector profiles, grade bands, and how the model handles data it does not have — published in full.
The methodology