A reproducible path from candidate set to scientific shortlist

Freeze the comparison contract, run each candidate once, persist every success or failure, and expose both the available signal and the evidence still missing.

Rank boundary

Catalog size is not benchmark size. Formal ranks remain empty until context, model, coverage and hard gates all pass.

Known

—

Visible research inventory

Structure-ready

—

Computable, not automatically comparable

Assessable

—

Bounded fixed-context path

Formal rank

—

Gates and coverage not yet satisfied

Candidate universe

0 matching

Every safe candidate remains visible; readiness controls what can happen next.

Select	Candidate	Readiness	Structure	Source coverage	Next

Loading the deployment-safe universe…