Research roadmap

Near-term research work should tighten the connection between observed data, discovery assumptions, and simulation validation: benchmark published systems with versioned generators; compare derivative estimators under controlled noise; add out-of-sample rollout evaluation to standard reports; and test unit-aware feature filtering.

The architecture describes broader targets such as regime-aware discovery, uncertainty-aware selection, richer data interchange, and causal workflows. Those are not current guarantees. Any addition must enter with an explicit data model, failure semantics, reference fixture, and independently checkable validation criterion—not merely a route or configuration name.