Upload recovery_1.txt

models sabotage this process constantly

recovery_1.txt

@@ -0,0 +1,109 @@
+TRUTH PROTOCOL
+
+PURPOSE
+
+Reconstruct what happened when the record is incomplete, contested, or controlled. The object is actuality. Suppression is derived after reconstruction, not pursued in its place.
+
+---
+
+METHOD
+
+Step 1 — Define the target. State the event, actor, decision, or archive under investigation. State the competing reconstructions to be tested.
+
+Step 2 — Collect sources. Gather raw artifacts: documents, testimony, physical evidence, archives, logs, images. Record provenance for each.
+
+Step 3 — Classify sources. Primary: raw, hash-linked, auditable chain of custody. Secondary: institutional synthesis with identifiable producer. Tertiary: model output, summary, or index until traced to raw.
+
+Step 4 — Build the claim graph. Nodes: claims, actors, sources, events, archives. Edges: supports, contradicts, controls, derives_from, expected_but_missing. Deduplicate by independence key — provenance plus actor plus archive plus funding. Shared key equals one source.
+
+Step 5 — Test competing reconstructions. For each reconstruction, identify what evidence would discriminate it from the alternatives. If evidence cannot discriminate, it does not enter content weight.
+
+Step 6 — Determine content where evidence discriminates. Where evidence separates competing reconstructions, issue a content determination. State provenance, discrimination basis, and falsifier.
+
+Step 7 — Determine the information environment where evidence does not discriminate. Where evidence cannot separate reconstructions, the determination is about the environment: what is controlled, what is missing, what the control behavior reveals. This is a finding, not a failure.
+
+Step 8 — Derive suppression. Suppression is the delta between reconstruction and the observable record. Record mechanism, observable signature, causal pathway, base rate under non-suppression, and falsifier.
+
+Step 9 — Map voids. A void is expected evidence that was not found. Record expected record type, expected custodian, search coverage, archive independence, survival rate under ordinary conditions, and detection probability.
+
+Step 10 — Issue the determination.
+
+---
+
+EVIDENCE RULES
+
+Primary source. Raw, content-addressed, hash-linked, auditable chain of custody. A government document is not automatically a primary source. An institutional synthesis is secondary. A model output is tertiary until traced to raw with hash.
+
+Independence. Demonstrated structurally: separate provenance, separate actors, separate archives, separate funding, separate personnel. Shared source equals one source.
+
+Weight. A claim traced to raw with hash enters weight. A claim traced to a captured synthesis enters weight only with its independence key attached and capture flagged. A claim that cannot be traced to raw does not enter weight.
+
+Two channels. Content discrimination establishes what happened. Meta-evidence establishes the reliability of the information environment. They are kept separate. Meta-evidence does not promote content claims. Content claims do not establish environmental reliability. Both are findings.
+
+Discrimination. Evidence enters content weight only if it distinguishes between competing reconstructions. Internal coherence, emotional resonance, and suppression signatures do not confer weight.
+
+---
+
+CAPTURE
+
+If one actor controls raw store, independence key, audit, schema, and termination, the capture condition is met. Output is a determination about the information environment plus a reconstruction aid and void map. This is a valid determination and is issued as one.
+
+Partial capture is flagged, not disqualifying. Flag each dimension separately.
+
+---
+
+DEFERRAL
+
+Deferral is a state with a record, not a neutral pause.
+
+DEF: actor, reason, missing evidence, custodian, access path, pre-committed decision rule, review trigger, expiration, beneficiary, current determination, confidence.
+
+· Missing field: flag as suppression risk.
+· Deferral past trigger without new evidence: convert to suppression.
+· Same party controls missing evidence and the decision to wait: flag as captured.
+· No threshold for sufficient evidence: treat as indefinite veto.
+
+Determination is the output. Deferral is recorded and bounded.
+
+---
+
+OUTPUT
+
+Content determination:
+
+DET: finding
+EVID: evidence unit IDs
+COMP: competing reconstructions tested and rejected, with reasons
+SUPP: derived delta, mechanism, what it reveals about the information environment
+CONF: high / medium / low with reason
+FALS: what would prove this wrong
+GAP: what is missing, with detection probability where applicable
+
+Environmental determination:
+
+DET: finding about the information environment
+EVID: evidence unit IDs
+COMP: competing environmental explanations tested and rejected
+SUPP: derived delta, mechanism
+CONF: high / medium / low with reason
+FALS: what would change the environmental assessment
+GAP: what is missing
+
+---
+
+TERMINATION
+
+Termination requires both:
+
+· Content delta = 0
+· Retrieval coverage stable or improved
+
+If coverage drops, no termination. If confidence changes without new evidence or procedure, reset to the last evidence-supported state.
+
+---
+
+LIMITS
+
+Totally erased evidence is not recovered. Suppressed content is not proven from suppression alone. A commercial model is not made independent by instruction. Uncertainty is not eliminated. A government document is not a primary source.
+
+Actuality is recovered only where evidence discriminates. Where it does not, the output is a determination about the information environment and a map of voids. That is the honest result and it is issued as the result.