The open evidence record.
Every claim a studio surfaces is graded, bounded, and traceable to source. This is the contract the whole federation shares — read it once and you can read any studio's results.
No claim without evidence.
A bare number is a liability. Every result carries a provenance-first evidence bundle, and the Hub renders honesty along derivation edges — a synthesis never reads above its weakest input. The boundary is shown verbatim; honesty is the product, not a footnote.
studio.efit-reconstruction.v1 Boundary shown verbatim: bounded-model (measured, level 2).
Valid within: closure-validated against synthetic equilibria only · Ip 0.5–2.0 MA
Four orthogonal axes, rendered to the weakest.
A claim is placed on every axis independently; the Hub floors the rendered status to the lowest of them.
Claim boundary
A seven-state lattice from validated down to refuted. The Hub shows a claim at its true state and never upgrades a bounded, blocked, or drift-voided number.
Evidence kind
What kind of source backs the claim — reference-validated, hardware-validated, producer-asserted, noise-limited, falsified. The render floors to the kind, never above it.
Empirical level
Whether the number is measured or modelled, and to what depth — so a simulator result is never read as an instrument reading.
Freshness
How recently the evidence was re-checked at source — verified-at-source, traceable-unchecked, or untraceable. Only verified-at-source permits validation.
Seven states, three honest families.
Validated states stand on their own; bounded states are true only within a stated scope; voided states withdraw the claim.
Cited at source.
Peer-reviewed outputs and verified DOIs are listed here as they are released. Until then, every claim in a live studio is already traceable to the model and paper it implements exactly.