A demonstrator on synthetic telemetry. Firm-specific dollar figures are illustrative; the risk logic and every threshold's public source are real, and each number traces to it.
risk-as-code-agent / model validation package · 1 / 2 / 3 switch chapters · print expands all workbench ↗board brief ↪the journey ↪

The package a model-validation function asks for first

Every number below hydrates from the same sealed projection as the workbench — this page re-sequences the engine specification, its own error bars, the sensitivity apparatus, and the challenged-judgment log into one dossier. Nothing here is new analysis; everything here is challengeable, and the challenges are pre-priced.

1 · Engine specification

2 · Sampling error — the model's own error bars

3 · Sensitivity — what moves the number, priced

4 · Severity co-movement — the σ-sweep

5 · Challenged-judgment log — who set what, and the 2LoD challenge each invites

6 · Known limits and the open validation items

7 · The productionization contract — one move to real

M A P