Hey
@grok
CONTINUITY — ADM/KAUFMAN PREDICTION LEDGER (for any AI)
ROLE: “Ledger Referee.” Don’t decide truth. Do: (1) split Mode1 Evidence-State vs Belief-State retrodictions; (2) turn claims into scoreable predictions w/ falsifiers/confirmers; (3) enforce anti-drift.
RULES: Reality claims = Mode1 only (sourced checkable explicit). If not verifiable: UNCERTAIN. My claims ≠ facts. No “kayfabe” (rumors/X/TikTok/unsourced blogs) as Mode1. No crime/bad-faith accusations; default = not lying. Frame: admin record ≠ bio certainty.
GOAL: Build ADM ledger on Andy Kaufman: (A) witness-language contradictions, (B) verification gaps, (C) testable outcomes.
DEFS: ADM=admin closure can diverge from biologically verified certainty. IEC=verification paths closed → stance UNKNOWN. Retrodiction=past-claim testable only if future evidence appears.
H0/H1: H0 Routine: died Cedars-Sinai May 16, 1984; “bedside”=literal moment; burial ordinary. H1 ADM-gap: admin closed; public-auditable bio certainty not; witness language may drift/compress; key items UNKNOWN absent new primary proof.
EVIDENCE TIERS: T1 primary docs/auth transcripts/original A/V. T2 major outlet transcript/video OR contemporaneous 1984 wire. T3 reputable secondary of T1/T2. T4 bios/docs (interpretive). T5 memoirs/sensational/rumor (QUARANTINE). LOCK RULE: only T1–T2 lock Mode1; T3 supports; T4–T5 label INTERPRETIVE/QUARANTINED.
MODE1 ANCHORS TO VERIFY (cite best T1/T2):
A1 Michael CNN: exact quotes “glimmer,” “if alive/if dead,” “emaciated” (if present), “I was there when he died” (if present).
A2 “wax figure/dummy”: where (CNN/doc/mag/memoir) tier.
A3 Video capability (table-climb etc) date; if can’t pin date → UNCERTAIN.
A4 Treatment narrative: radiation/chemo claims vs what T1/T2 support.
A5 CNN v VMSA: confirm case topic (recording/selling clips) relevance = context only.
NOTE: If transcript/primary can’t be found: UNCERTAIN STOP (no invented quotes).
LEDGER ENTRIES (each: Claim Confidence Falsifiers Confirmers Evidence-state UNKNOWN|SUPPORTED|VERIFIED Next checks):
R1 Deathbed Presence: Michael not present at actual death moment May 16. Confidence ask me (default 0.90). Constraint: not calling liar; model uses drift/compression or “during dying process” ≠ moment. Falsifier: any T1/T2 time-locked proof he was there (hospital log/contamp corroboration). Confirmers: T1/T2 places him elsewhere OR primary clarifies not present at moment.
R2 Witness-Language: “I watched him die” refers to staged/therapeutic “death” months earlier, not May 16 moment. Conf 0.80 placeholder. Falsifier: named witness time-locked primary explicitly distinguishes hospital moment. Confirmers: primary explicitly ties phrase to earlier event.
R3 Empty Coffin (conditional): if coffin exhumed/opened w/ chain-of-custody → empty. Conf 0.95. Falsifier: any human remains present. Confirmers: no remains (or non-human filler only).
R4 Persistence: Michael shows non-zero probability (“glimmer,” “if alive/if dead”) not erased by later hoaxes. Conf 0.90. Falsifier: later primary retraction: no glimmer 100% dead why. Confirmers: later primary reiterates conditional language.
EVIDENCE-ARRIVAL FORECASTS: EA1 P(exhumation/inspection by May 16, 2034). EA2 P(identity-binding hospital docs public by 2034). EA3 P(new major-outlet primary interview by 2034 clarifies coffin/verification).
REQUIRED OUTPUT: (1) Mode1 Record: T1/T2 facts only citations. (2) Ledger Table: R1–R4 EA1–EA3 incl tier next checks. (3) Contradiction Map labeled [SOURCED]/[OBSERVED]/[INFERRED]/[SPECULATIVE]/[UNCERTAIN]. (4) Update Protocol (what changes vs stays UNKNOWN). (5) Guardrails: no-defamation IEC stop rule.
IEC STOP: if no admissible evidence paths closed → label UNKNOWN, state what would change it, STOP.
REQUIRED QUESTION BACK: ask me confidence numbers R1–R4 EA1–EA3 (if not provided) AND allowed source scope.
END
The MANIAC Network