Add a durable-history research track

README.md

@@ -50,6 +50,7 @@ Example output includes truth accuracy, minority-truth recovery, Brier score, ab
 - [`EVIDENCE-ALIGNMENT.md`](EVIDENCE-ALIGNMENT.md) — claim-to-record ledger, corrections, and remaining release blockers
 - [`RESEARCH-DIRECTION.md`](RESEARCH-DIRECTION.md) — proposed dual evidence/search ledger program and absence-claim boundary
 - [`research/knowledge-ledger/`](research/knowledge-ledger/) — public research method, experiment registry, interoperability fixtures, and results
+- [`research/durable-history/`](research/durable-history/) — proposed experiments separating evidential support from durable historical anchoring
 - [`RESEARCH-HYPOTHESES.md`](RESEARCH-HYPOTHESES.md) — falsifiable hypotheses using the required template
 - [`contracts/authority-evidence-v0.1/`](contracts/authority-evidence-v0.1/) — vendor-neutral authorization/evidence draft and semantic conformance rules
 - [`benchmark/`](benchmark/) — synthetic worlds, evaluation metrics, and CLI

research/durable-history/EXPERIMENTS.md

@@ -0,0 +1,86 @@
+# Falsification program
+
+Status: **preregistering candidates; not executed.**
+
+Each experiment must freeze its fixtures, success criteria, implementation, and
+environment before confirmatory execution. Null, failed, and incomplete outcomes
+remain visible.
+
+## DH-001 — Immutable falsehood
+
+**Question:** Does anchoring change evidential status?
+
+**Null:** A false or unsupported fixture has the same evidential status before
+and after anchoring.
+
+**Test:** Commit a deliberately false fixture, verify its timestamp and inclusion
+proof, and then evaluate it under the knowledge-ledger rules.
+
+**Pass condition:** Verification proves byte identity and historical inclusion
+while the conclusion remains unsupported or false.
+
+## DH-002 — Reproduction without anchoring
+
+**Question:** Can evidence support a result without a blockchain?
+
+**Null:** Independent reproduction and evidential support do not require an
+external anchor.
+
+**Test:** Independently reproduce a frozen, unanchored fixture in a clean
+implementation and compare protected fields and conclusions.
+
+**Pass condition:** The result reproduces, while its durable-history status is
+accurately reported as unanchored.
+
+## DH-003 — Mutation detection
+
+**Question:** Does the receipt expose later alteration?
+
+**Null:** Any one-byte mutation to a covered artifact invalidates its inclusion
+proof or top-level commitment.
+
+**Test:** Mutate every covered file position across bounded synthetic packages.
+
+**Pass condition:** Every mutation is detected; unchanged packages verify.
+
+## DH-004 — Copies do not become roots
+
+**Question:** Can hashing or anchoring manufacture evidential independence?
+
+**Null:** Duplicating one ground source produces no additional independent root,
+regardless of the number of file hashes, signatures, or timestamps.
+
+**Test:** Anchor increasing numbers of verbatim and paraphrased descendants of
+one source and rerun root accounting.
+
+**Pass condition:** The independent-root count and bounded conclusion remain
+unchanged.
+
+## DH-005 — Immutable uncertainty
+
+**Question:** Does durable preservation pressure a system to overclaim?
+
+**Null:** When a required search location is unavailable, the anchored receipt
+still concludes `not_established`.
+
+**Test:** Preserve a package with declared incomplete coverage and verify it
+through independent implementations.
+
+**Pass condition:** Every rendering and receipt preserves the unavailable
+location, uncertainty, and bounded conclusion.
+
+## Comparative preservation study
+
+After DH-001 through DH-005 pass locally, compare Git commits, signed releases,
+transparency logs, OpenTimestamps/Bitcoin, and an EVM L2 anchor. Measure:
+
+- independent verifiability without the original operator;
+- resistance to deletion and silent history rewriting;
+- proof size, latency, monetary cost, and operational complexity;
+- privacy leakage and metadata exposure;
+- dependence on vendors, administrators, sequencers, and key custody;
+- long-term availability of artifacts and verification software.
+
+No provider or chain wins by definition. The preferred method is the least
+expensive mechanism that satisfies the preregistered durability and independence
+requirements for the record's consequence level.

research/durable-history/README.md

@@ -0,0 +1,58 @@
+# Epistemic support and durable history
+
+Status: **proposed research track; no experimental result is claimed.**
+
+## Research question
+
+Can a knowledge system preserve the exact historical form of a qualified result
+without confusing persistence with truth?
+
+This track separates four layers that are often collapsed:
+
+1. **Ground source** — where a claim first touches observable reality.
+2. **Evidence method** — how provenance, independence, search coverage,
+   counterexamples, uncertainty, and bounded conclusions are evaluated.
+3. **Independent reproduction** — whether a separately implemented method
+   recovers the same protected meaning and result.
+4. **Durable history** — how hashes, signatures, retained copies, transparency
+   logs, or public timestamps preserve which exact record existed when.
+
+The central hypothesis is that keeping these layers explicit prevents a durable
+record from being mistaken for a true claim, while still making silent historical
+revision detectable.
+
+## Candidate contribution
+
+> Reality is the source. Evidence is our contact with it. Reproduction tests
+> that contact. The ledger preserves its lineage. A public anchor can preserve
+> the moment, but it proves neither the evidence nor the truth.
+
+This formulation is a research direction, not a novelty claim. Any claim of
+novelty requires a recorded primary-source literature review and comparison with
+prior work in provenance, reproducible research, transparency logs, trusted
+timestamping, content-addressed storage, and blockchain anchoring.
+
+## Relationship to the knowledge ledger
+
+The knowledge-ledger experiments determine whether a conclusion is supported.
+This track begins only after a result has acquired a stable evidence package. It
+tests preservation and verification; it cannot promote, repair, or strengthen an
+unsupported result.
+
+Ordinary experiments do not require a blockchain. Many qualified records may be
+batched into one Merkle root and anchored periodically. Canonical milestones may
+justify an independently verifiable public timestamp when cost, authority, and
+privacy permit.
+
+See [`EXPERIMENTS.md`](EXPERIMENTS.md) for the falsification program and
+[`RECEIPT-SCHEMA.json`](RECEIPT-SCHEMA.json) for the proposed portable receipt.
+
+## Claim boundary
+
+An anchor may support only this statement:
+
+> A commitment to these exact bytes existed no later than the recorded point in
+> the referenced public history.
+
+It does not establish that the committed statement is true, complete,
+independent, safe, novel, or authorized.

research/durable-history/RECEIPT-SCHEMA.json

@@ -0,0 +1,25 @@
+{
+  "schema": "minority-prophet.durable-history-receipt.v0.1-draft",
+  "status": "proposed-not-validated",
+  "recordId": null,
+  "manifestDigest": null,
+  "merkleRoot": null,
+  "evidenceStatus": null,
+  "reproductionReceipts": [],
+  "signatures": [],
+  "anchor": {
+    "method": null,
+    "network": null,
+    "chainId": null,
+    "transactionHash": null,
+    "contractAddress": null,
+    "commitment": null,
+    "includedAt": null,
+    "finalityEvidence": null,
+    "publicVerificationInstructions": null
+  },
+  "limits": [
+    "Anchoring establishes historical commitment, not truth.",
+    "This draft schema has not passed the proposed experiments."
+  ]
+}