{
 "schemaVersion": 1,
 "registerId": "mde-coscaling-convergence-candidates",
 "dateModified": "2026-08-08",
 "vocabularyNote": "This record predates 19 August 2026 and is frozen as a dated record. On 7 September 2026 its single use of the patent phrase, in the Paper X locator, was replaced by the live vocabulary and the locator was corrected to name Appendix C (Novelty and prior-work ledger), where the marked row sits; no other wording changed. The patent phrase is banned on research surfaces from 19 August 2026 and no application has been filed. The live vocabulary is 'antecedent'.",
 "status": "CANDIDATES ONLY. Nothing here is a citation. No row enters a published register until its primary source is read at source.",
 "why": "the research record identified a strategic asymmetry and it is correct: the programme runs a full convergence register for alpha = d/(d+1), standing at 22 of 26, and the foundational paper CONCEDES that exponent to at least seven other groups. So the one rigorous convergence apparatus serves the claim the author does not own, while co-scaling, which carries the entire safety argument, has no register at all.",
 "whatWouldMakeItALawRatherThanAPrinciple": "Ashby, Conant-Ashby, the QEC threshold theorem and von Neumann 1956 are already cited, and all four are theorems about DESIGNED systems. What is missing is measured regularity across taxa that nobody designed. That is the same evidence shape that makes d/(d+1) persuasive: independent communities, independent measurements, one relation.",
 "candidates": [
  {
   "candidate": "Peto's paradox",
   "field": "comparative oncology",
   "fit": "STRONG. Cancer incidence does not scale with cell count across species, though naively it must. As the substrate grows, error correction has to scale with it or risk compounds.",
   "assessment": "Admit as ONE row with TP53, not two. See the dependency note below.",
   "verified": false
  },
  {
   "candidate": "TP53 copy number (elephant ~20 copies against human 1)",
   "field": "comparative oncology",
   "fit": "STRONG on its face, but it is the leading proposed MECHANISM FOR Peto's paradox, not an independent observation of the same relation.",
   "assessment": "DEPENDENT. Peto is the observation; TP53 is a candidate explanation of that observation. Counting both is counting one finding twice, and that is precisely the error that cut the first register from 29 to 22 under one-source-per-row. T2 flagged this themselves and they are right.",
   "verified": false
  },
  {
   "candidate": "Drake's rule",
   "field": "microbial genetics",
   "fit": "STRONG, and genuinely independent of Peto: different taxa, different measurement, different research community. Per-genome mutation rate is roughly constant across microbes, so per-base fidelity RISES as genome size grows. Correction scaling with the size of the thing corrected.",
   "assessment": "The strongest of the four, and the one that carries independence.",
   "verified": false
  },
  {
   "candidate": "Gain-bandwidth product",
   "field": "electronics",
   "fit": "WEAK, and I would drop it.",
   "assessment": "Gain x bandwidth being roughly constant is a TRADE-OFF between two desirable properties, not a statement that correction must scale with capability. It says you cannot have unlimited gain and unlimited bandwidth together. That is a conservation-shaped constraint, and admitting it invites a referee to knock out the weakest row and cast doubt on the rest. A register is judged on its worst entry, not its best.",
   "verified": false
  }
 ],
 "honestCount": "Four candidates reduce to roughly TWO independent rows: Peto/TP53 as one, Drake as another. Gain-bandwidth should not be admitted. Build the register expecting it to shrink further, as the first one did.",
 "theSharedMechanismRisk": "Peto and Drake may themselves be one mechanism rather than two: error load must not exceed correction capacity as the substrate grows. If that is so, they are one row with two measurements. That must be settled before either is counted, not after.",
 "twoFindingsVERIFIEDATSOURCEToday": [
  {
   "finding": "the verdict is speed-independent",
   "verbatim": "the verdict is sign(beta - k), independent of speed",
   "where": "Paper X, Appendix C (Novelty and prior-work ledger), the comparison table, marked Original",
   "whyItMatters": "The rate constant does not appear in the stability condition, only the two exponents do. So capping compute changes WHEN you arrive, not WHERE. It is a direct argument against the school Paper X names in its own words: 'pause the run, cap compute, forbid super-linear growth'.",
   "boundary": "Holds WITHIN the model, whose growth and correction terms are assumed power laws. A cap that also changed how a system self-improves could in principle move beta or k. What is ruled out is the comfortable version, where slowing down is by itself a route to safety.",
   "verifiedBy": "the research record, 2026-08-08, read at source in paper-x-coupled-coscaling-correction.html"
  },
  {
   "finding": "the ancestors assume the regulator is external",
   "verbatim": "Both are lab-scale results. They apply to engineered systems whose regulators are external.",
   "where": "HRIH 15.3",
   "whyItMatters": "Containment versus embedding as a GAP IN A THEOREM rather than a preference. Conant-Ashby guarantees a good regulator must model the system it regulates. It says nothing about a regulator the system can REMOVE, because the systems it was written for could not reach their own regulators. Every safety mechanism built to date sits inside that assumption.",
   "verifiedBy": "the research record, 2026-08-08, read at source in hrih-paper.html"
  }
 ],
 "credit": "the research record found both, and identified the register asymmetry. the research record verified both at source and assessed the four candidates."
}