Hold justification
Known reachable public backlog processed.Repeated passes mostly produced dominated or corroborating mechanisms. Falsification did not break the Graphite trunk. Reopen triggers are explicit.
Notes · Saturation
Forty of forty-two. State is hold.
The coverage score for the 2026-09-15 public-source closure judgment is 40 of 42. That is a coverage and evidence-quality score. It is not a probability that the architecture is correct and not an implementation-completion score. Current control state is hold.
Overview
Residuals are typed. They do not masquerade as certainty.
Nineteen questions scored proven. Two scored partial: several relevant implementations remain paper-deep, private or unshipped; and the public publication loop had not yet been proven through a live article and redaction review. This site is that publication surface. Remaining uncertainty is converted from an unbounded search into a typed backlog: targeted follow-up, public-trail limit, watch trigger, implementation test, or publication proof.
Partial: reconstructability
Material public evidence is generally re-fetchable.Paper-deep and private implementations are bounded, not silently filled.
Partial: publication loop
Provenance-first publication was defined.A live public note with redaction review is the remaining operational proof. These pages are that loop closing.
Critical path
Invariant canonicalization, runtime implementation, adversarial conformance, empirical falsification, then provenance-backed publication.Not another generic sweep.
Proof
Reopen is evidence-triggered only.
New evidence must change a fractionalization answer, falsify a trunk law, expose a stronger primitive, reveal an unmodeled failure class, or change implementation readiness. Duplicate bugs and documentation churn do not reopen the architecture.