# Changelog

## 3.0 — 13 November 2026

- Added the inference and interpretation contract, versioned competing accounts, bridge assumptions, quantitative prediction matrix, and six-outcome adjudication.
- Added conditional design assurance with null/target/rival calibration, 95% Wilson uncertainty gates, scientific-basis digests, pinned evidence, and an optional synthetic-calibration summarizer. No simulator outputs or scientific decisions are fabricated.
- Added critical-perspective review, explicit proponent/proxy/internal-critic labels, conflicts/overlap disclosure, and design/interpretation acceptance tied to the current review digest.
- Added analysis masking/unblinding plans, proportionate replication requirements, and outcome-independent reporting/amendment commitments.
- Added the methodological critique, detailed contract reference, review worksheet, calibration input template, and prediction/assurance CSV exports.
- Migration now copies v1 or v2 to v3 with unresolved new records and cleared active approvals; existing decisions and historical reviews are preserved. V1/v2 remain draft-readable but cannot create new freezes with this compiler. Existing archives are unchanged.
- Updated website, prompt, example, and download. The proposed occlusion study remains unapproved and is scoped as a candidate mechanism benchmark.

These are original Mind Guild workflow choices informed by public work, not a review or endorsement by Lucia Melloni or COGITATE. The interval-based engine is a restricted two-account scalar design; scientific adequacy and real review remain human responsibilities.

## 2.0 — 12 November 2026

- Added structured construct/claim-to-measure maps, competing explanations, expected result patterns, directly encoded-effect interpretation, and conclusion limits.
- Separated implementation status, demonstration/verification status, and study inclusion for each component.
- Added conditional records for offline calibration, persistent runtime state, and online parameter learning, including the teaching/error/reward signal and evaluation update policy.
- Added per-field source locators, agreement status, unresolved speaker attribution, named decision records, explicit analyst-proposal adoption, and field-value digests.
- Added artifact handoff and a prospective-results-access declaration, including prior results, who saw them, design influence, reserved runs, and analysis-only timing when appropriate.
- Added CSV decision/artifact exports, a ten-minute protocol close, an updated meeting worksheet and extraction prompt, and a detailed record reference.
- Added explicit v1 migration and undecided provenance scaffolding. Compilation does not change input values or infer agreement. Legacy drafts remain readable; v2 completion and renewed reviews are required for freeze.
- Updated the proposed occlusion example, generated packet, downloadable kit, and website guide. The example remains unapproved and incomplete.

Scientific adequacy, genuine agreement, and remote artifact integrity remain human responsibilities. A local freeze is not an OSF registration.

## 1.0 — 11 November 2026

Initial meeting worksheet, transcript extraction prompt, protocol template, proposed occlusion example, packet compiler, local freeze convention, and OSF handoff guide.
