Coverage and cost

All 39 native attempts have returned: four whole-paper attempts, 29 bounded local reviews and six arbitrations. The accepted mathematical reports are one whole-paper Sol report and 25 local Sol reports. Four local attempts were invalid. All six native arbiter deliveries were rejected for metadata/schema errors; separately logged metadata-only copies passed the strict source validator. They are not six originally accepted model deliveries.

The total is usage-based accounting against the authorized $25 ceiling, not an invoice reconciliation. No additional paid inference was used for the source comparison, exact arithmetic, local repair or publication. Fable ran at high, not max. Gemini, Opus and Fable did not yield complete usable whole-paper reports at their caps; this is not a mathematical rejection vote.

The local wave covered 19 named claims and ten narrative segments. Assignment of the mathematical body does not establish correctness or complete coverage. External inputs, quantitative unsmoothing and the full qualitative sieve chain remain unresolved in this review. The 37 bibliography nodes identify references, not 37 independently verified papers.

The neutral map preserves source references and adds a limited set of explicitly attributed proof-input edges. It is incomplete. A terminal node means outdegree zero in this recorded graph; it does not classify all headline results. Unadjudicated nodes remain yellow. Red statements and unresolved consumers are not conflated.

The attempt ledger retains native outcomes, model, effort, token usage and accounted cost. Selected structured assessments and derived arbitration provenance are in the curated panel download. The source manifest and edition notes are in the separate reading download.