MOLGATE contract digest: a preregistered control-completeness campaign
Year: 2026
Type: other
Abstract: Accompanies P0 version 9, which remains the current registration and is unchanged. It records the refusal trail that followed: five run-labelled freeze attempts, thirteen defects, no provider spend. Position is post-D1, pre-freeze, pre-D2. D1 has been executed; no valid contract has been frozen, no P1 exists, and D2 has not begun. It is published before the successor registration that resolves it. This is not a declaration of infeasibility. The registered rule fired once, at the fifth attempt, on a realized healthy-pass probability of zero produced by a joint region folded at a statistic the campaign had not registered. Under the registered rules the criterion reaches 0.993850 against a target of 0.95. Principal findings. (i) Four of the seven registered sections were emitted by no code path. The contract schema, the lineage claim governing holdout reuse, the 256-probe instrument bridge, and the claim-grain confirmation laws were declared in the overlay, carried verbatim by version 9, signed, timestamped and anchored, and absent from every frozen contract; each freeze accordingly produced a version 2 contract for a campaign registered at version 3, and each missing section presented as unregistered instead of dropped. The disposition table established that a destination had been assigned for each section, not that any section arrived. (ii) Repairs to the region's folding statistic, its prior treatment, and the criterion's measure reached the harness but not the freezer, which writes the contract the arrival gate reads; the offline verifier was a fourth implementation of the same quantity. A seventh attempt would have refused correctly on a number computed under unregistered rules. (iii) One calibration slot was purchased twice, 98 seconds apart, with byte-identical requests and differing replies, both first attempts. The run scored the later purchase under no registered rule. Both replies refuse and recompute the same formula, so the verdict is invariant. The thirteenth defect is in this record's own drafting. Three figures cited as the campaign's coverage under the two measures, and the dependence between them, belong to a synthetic test fixture evaluated at confidence 0.95 over 2,000 simulations against the registered 0.995 over 20,000. Two conclusions drawn from them are withdrawn: that the measure mismatch separated a feasible design from an infeasible one, and that the criterion refutes the design assurance stated in version 9. Neither holds. A third claim, that the simultaneous region is a subset of every other conjunct, was also a property of the fixture. The underlying repairs stand; none depended on the margin. Total campaign spend is 94.8771 USD over 4,413 billable responses, accounted separately and entirely predating the refusal trail. Verification. The canonical SHA-256 of the body is 095a66c519f640d0059bdf9e5f0da3e73ad26f8b90d54cb099c957c5c2d12b44. The body is canonical JSON, so its bytes are its canonical form. The detached signature is raw Ed25519 over those bytes and verifies against the included public key. The RFC 3161 token was issued by freetsa.org with status granted and an imprint equal to the digest above. Sigstore transparency log entries stand at index 2726760685 over the artifact and signature and at index 2726760790 over the timestamp token, both carrying inclusion proofs. The body also carries a record digest over its own content less that field. The timestamp authority's certificate chain has not been verified.
Author: Lukas Weidener
Language: English
Subfield: Strategy and Management
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7203882241