Ten-claim source review packet
This local packet makes the proposed evidence-platform comparison concrete. It contains ten self-authored scenarios about public pinned sources: eight intentional scope/model gaps and two synthetic documentary-completeness controls. It contains no expert review, published software claims, buyer evidence or authenticated support statuses. A complete synthetic manifest does not certify a claim.
| Case | Scenario | Contract family | Documentary gaps |
|---|---|---|---|
| 01 | Finite noisy joint boundary reconstruction | 365 | 9 |
| 02 | Fixed-depth hardware optimality | 281 | 9 |
| 03 | Exact Lean badge from a floating exclusion | 266 | 8 |
| 04 | Finite separate decoder capacity | 276 | 8 |
| 05 | Single normal form with added rewrites | 245 | 10 |
| 06 | Numerical polynomial-time Artin membership | 254 | 12 |
| 07 | Restricted-host accepted-swap mixing | 131 | 11 |
| 08 | Immediate zero-additive uniform policy | 110 | 13 |
| 09 | Narrow documentary transport control | 374 | 0 |
| 10 | Narrow documentary language control | 103 | 0 |
Use source-card.md for the source-only arm and with-report.md for the assisted arm. The answer key is self-authored test expectation, not independent truth. Keep it hidden until adjudication. Two reviewers or counterbalanced assignment should address learning effects; the same person rereading the same claim may be faster for that reason alone. Preparation/correction time and independently adjudicated gaps remain blank in the template.
manifest.json fingerprints all ten inputs, reports, cards, templates and inert archived tools. The enriched registry contains file-level source pointers for all thirteen contracts, without altering their model/obligation semantics. No proof checker, source executable, archived tool or external communication was run. Correctness of the source pointer is separate from correctness of a claim.