bundle/v0/presentation/NARRATIVE_SPEC.md

# Narrative

1. Meet a two-asset decision and discover why equal forecast distance can create unequal loss.
2. Shape an uncertainty map in the Decision Atlas field trial.
3. Predict whether distance should stay fixed, be hand-shaped, or be learned from loss.
4. Run one faithful bilevel step from metric to robust decision to validation loss to hypergradient.
5. Open six ordered official claim pages, each with its exact text, conservative verdict, supporting facts, and strongest downgrade.
6. Scrub sample size in the sealed coverage and regression traces, then inspect censoring, theorem scope, and provenance.

The field trial is intuition, never scored evidence. The result layer appears only after the reconciled receipt passes the binder.