{
 "schemaVersion": 1,
 "registerId": "mde-arc-bound-framing-question",
 "dateModified": "2026-08-12",
 "status": "RULED - stability limit, not hard ceiling. Operator statement 2026-08-08 ('I propose it as a stable limit, not a hard limit'), reconfirmed 2026-08-10 when the operator pressed the stable-limit reading in an independent incognito reading test and forwarded it as the intended meaning. Site-editable surfaces executed 2026-08-10 (falsification.json FALS-020 criterion restated in the sustained-without-loss-of-stability form; definitions page published). DOI-bound paper editions (Papers I-III F11 + Carnot-ceiling language) dispatched to the research record for version bumps and OSF re-upload; until those land the papers carry the pre-ruling phrasing and this record plus the definitions page govern the intended reading.",
 "question": "Is the ARC Bound a HARD ceiling that no system can exceed, or a STABILITY limit beyond which growth continues and destroys itself?",
 "whyItMatters": "The two readings are incompatible and they demand DIFFERENT FALSIFIERS. Under the hard reading a measured exponent above 2 refutes the framework. Under the stability reading the same measurement CONFIRMS the framework's supercritical regime and the correct test becomes whether the system then destabilises. The registered falsifier currently encodes the hard reading on both surfaces, so if the stability reading is the intended one, the programme is publishing a kill condition that would fire on its own success.",
 "operatorStatement2026_08_08": "Papers I to III were written prior to changing what the ARC Bound was. I propose it as a stable limit, not a hard limit.",
 "verifiedByReading": {
  "lane": "the research record",
  "date": "2026-08-08",
  "hardReadingInPaperIII": [
   "The ARC Bound therefore identifies a computational ceiling analogous to Carnot efficiency: a bound set by the information-processing geometry of the substrate, not by engineering limitations.",
   "No classical sequential recursive system can sustain alpha > 2.",
   "The ARC Bound: For classical sequential recursive systems, alpha -> 2 from below.",
   "F11 registered falsifier: ARC Bound (printed in the paper as alpha_max = 2; the retired symbol, the current form being alpha_crit = 2 at gamma = 1/2) ... Sustained alpha > 2."
  ],
  "stabilityReadingInTheSamePaper": [
   "We derive a safety boundary for recursive intelligence (the ARC Bound, beta <= 0.5).",
   "beta <= 0.5, alpha <= 2 for safe operation.",
   "This is the Eden Protocol: the deliberate engineering of the composition operator in recursive AI systems to ensure that the self-referential coupling beta remains below 0.5.",
   "Just as control rods convert a bomb into a reactor by changing the composition operator, alignment mechanisms must convert unbounded self-reference into bounded self-reference."
  ],
  "theIncompatibility": "A Carnot-style physical ceiling and a safety boundary cannot both be true of one quantity. If no system CAN exceed 2 there is nothing to engineer and no control rods are needed, because a reaction that cannot go supercritical needs no moderation. If it CAN be exceeded then 'no system can sustain alpha > 2' is false.",
  "hrihCarriesTheStabilityReading": "HRIH, revision of 2 August 2026: 'The ARC Bound alpha <= 2, the candidate ceiling on the stable growth of complexity that lets creation grow super-linearly WITHOUT DESTROYING ITSELF, is likewise a live hypothesis, not a measurement claim.' This is the most recent statement and it matches the operator's account.",
  "provenance": "Paper III itself dates the two layers: 'The ARC Bound (first stated: December 2024 in Infinite Architects; formalised with computational derivation: February 2026).' The O(N^2) derivation added in February 2026 is what converted a stability proposal into a hard ceiling, because O(N^2) genuinely caps a transformer. Grounding the bound there made it architecture-dependent and therefore EXPIRING: a non-transformer architecture has no such ceiling. The stability reading applies to any recursive system regardless of architecture, so it is both the original and the more general claim."
 },
 "consequenceForTheRegisteredFalsifier": {
  "currentCriterion": "A measured exponent exceeding the bound under the published blinded protocol on frozen systems",
  "correctUnderTheHardReading": "unchanged: exceeding 2 refutes it",
  "correctUnderTheStabilityReading": "alpha > 2 SUSTAINED WITHOUT LOSS OF STABILITY. A system that runs supercritical and does not destroy itself. Strictly harder to satisfy, so the claim gets stronger rather than weaker.",
  "theParagraph2Episode": "Paper II measured alpha ~ 2.24 and it was recorded as exceeding the programme's own bound and treated as a potential falsification. Under the stability reading that was the wrong response: the correct question was whether that system then destabilised. The retraction to ~0.49 (interval [-1.3, 2.9]) was about measurement quality, which is separate and remains right."
 },
 "whyNothingWasChanged": "Fixing one surface alone manufactures a page-versus-paper contradiction, which is the defect this estate spent 8 August 2026 removing. The dashboard criterion in falsification.json is editable and the papers are DOI-bound, so a unilateral dashboard edit would put the register and the papers into conflict. Deciding WHAT THE CLAIM IS is the author's call, and both surfaces then move together.",
 "surfacesThatChangeTogetherOnceRuled": [
  "research/papers/paper-iii-alignment-scaling-problem.html — F11 and the Carnot ceiling language (DOI-bound: version bump + OSF re-upload)",
  "research/papers/paper-i-arc-principle.html and paper-ii-experimental-validation.html — check the same framing (DOI-bound)",
  "research/data/falsification.json — the published kill condition criterion (site register, editable)",
  "research/data/claims-ledger.json — if any claim states the bound",
  "research/falsification-report/ — regenerate after any paper edit; the SHA-256 pins will show the change"
 ],
 "credit": "Operator raised it. the research record verified it by reading Paper III end to end rather than searching it.",
 "twoWaysWithdrawal2026_08_12": {
  "date": "2026-08-12",
  "withdrawnTest": "one number does two jobs, measured two independent ways, and the answers must agree",
  "reason": "vacuous: if both exponents arise because accumulation is independence-limited, they share a law, agreement is forced, and a test that cannot fail is not a test. A common cause is a simpler theory than a coincidence, and the replacement asserts no independence, so the confound has nothing to bite on.",
  "replacementCarrier": "one accumulation law with two consequences (the conversion exponent and the correction ceiling). The cross-class to same-class correction-exponent ratio already on the register carries the unification. Under the exchangeability null the expected value of that ratio is exactly 1.00, because architecture-blind averaging has no free parameter for composition class; the framework says materially above 1.00. Two predicted numbers for one measurable quantity.",
  "ranking": "law before value. The ceiling of stable self-improvement is the reciprocal of the correction shortfall, one minus the correction exponent. Feed the law one assumption, that same-class correctors accumulate corrections the way independent samples do, so the exponent is one half, and it returns two. If two dies the law is untouched (measure the exponent, report a different ceiling); if the law dies two has nothing holding it up. State the law first, the number second; aim tests at the assumption, never at the digit.",
  "surfacesToCorrect": [
   "any surface still saying 'one number does two jobs' or 'measured two independent ways ... must agree' must be corrected, stating the withdrawal as a withdrawal (with its reason), never silently",
   "research/data/registration-diligence-table.json: study-ae row marked withdrawn 2026-08-12, cross-class ratio named as replacement carrier",
   "research/data/claim-governance-policy.json: the withdrawn phrasings added to the programme's forbidden-wording register",
   "research/data/claims-ledger.json: CEILING-DERIVATION-LAW claim carries the prohibitedWording that the executable wording gate enforces"
  ],
  "credit": "Canonical brief for this wave; the research record recorded the withdrawal on 2026-08-12."
 }
}