What genuinely connects the Hu-2026/NeuNeu and March & Weatherall/arXiv:2610.39107 corroboration-absence notes (cosine 0.88, unlinked)
Prior sibling check (per bridge-seed discipline): a search of 10-inbox/raw/ for captures mentioning both
claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24 and
claim-no-independent-corroboration-march-weatherall-2610-39107-as-of-2026-11-02 together found none. The
two slugs appear in overlapping but disjoint sets of captures (the Hu-2026 branch's own promotion threads;
the March & Weatherall branch's own promotion threads), never in the same file. This is a first investigation
of this specific pair, not a duplicate of an existing one — though it follows a method the vault has already
run three times on structurally identical pairings (see Claim 3).
Claim: The pairing is a false friend at the object level — no real-world connection between the two papers, their authors, or their subjects was found
claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24 concerns Hu, Pan, Jhaveri, Lourie & Cho's "Neural Neural Scaling Laws" (arXiv:2601.19831) — a machine-learning methods paper (NYU-affiliated authors) proposing a neural scaling-law predictor for LLM downstream accuracy, also listed at NeurIPS 2026 and ICML 2026. claim-no-independent-corroboration-march-weatherall-2610-39107-as-of-2026-11-02 concerns March & Weatherall's "Minimally-coupled Quasilinear Symmetric Hyperbolic Systems, Naturally" (arXiv:2610.39107) — a mathematical-physics / philosophy-of-physics paper (Weatherall is at UC Irvine) about symmetric hyperbolic PDE structure in field theories. A targeted web search for any link between the two author groups ("Michael Hu Kyunghyun Cho 'Neural Neural Scaling Laws' James Owen Weatherall Eleanor March") returned the NeuNeu paper's own listings (arXiv, NeurIPS, ICML, and several AI-summary mirrors) and no mention of Weatherall or March anywhere in the results; no shared institution, venue, citation, or subject-matter overlap surfaced. This negative search is weak evidence, held here only as due diligence rather than as a cited source — consistent with how the vault's prior instances of this same check (below) have treated an absence-of-connection search. There is no real-world bridge between neural scaling-law prediction for language models and the differential geometry of symmetric hyperbolic systems in general relativity.
Claim: The two notes ARE connected one level up — both are the vault's own single-source concentration cap discipline, applied identically to two unrelated preprints
Both notes are outputs of the same standing procedure: sources.md's single-source
concentration cap (Cali ruling, 2026-07-29) — "No more than three claim-notes may rest on a single unrefereed
primary … until at least one independent source corroborates it" — triggers a dated search for independent
(different-author, different-venue) corroboration, and the search's negative result is recorded as a note
stamped "as of <date>." Both target notes share near-identical structural and phrasal DNA as a direct
consequence: "no independent … corroboration … has been found as of [date]," "the single-source cap remains
undischarged," a search across Semantic Scholar / general web / (for March-Weatherall) INSPIRE-HEP and NASA
ADS, an explicit statement that the absence is "a date, not a disconfirmation," and a closing commentary
distinguishing a true search-negative from a refused or rate-limited query. A direct vault_pair measurement
confirms the embedding similarity this produces: cosine 0.8809 between the two notes' stored vectors, each
note's 99.7th-percentile neighbour within the other's own distribution, despite neither ranking in the other's
immediate top-5 nearest-neighbour list (rank 9 of A-in-B-neighbours, rank 7 of B-in-A-neighbours) — a high
raw similarity driven by shared template language, not topical adjacency. [unverified-mechanism — the vault_pair
figures are a direct tool measurement against the live retrieval index, not independently re-derivable from a
second source, but the mechanism they measure (template-driven similarity) is corroborated by the three prior
identical findings below.]
Claim: This is a fourth instance of an already-named, already-catalogued phenomenon — the "hedged bridge" the vault has found three times before
moc-the-same-source-wearing-different-clothes (built 2026-10-30, Warden pass) already names this exact
shape under "The cluster watching itself — within-cluster bridge-checks": "The retrieval index pairs two of
this map's own member claim-notes at high cosine; a bridge-check reads the pairing; and the verdict is the
same each time — a false friend at the object level, a real convergence one level up — because both members
are independent executions of the same single-source / evidentiary-sufficiency discipline, not because their
subjects relate." The MOC catalogues three prior instances, all carrying bridge_verdict: hedged:
observation-greenhalgh-2003-hu-2026-cosine-085-pairing-is-hedged-bridge (Greenhalgh-2003 / Hu-2026, cosine
0.85), observation-riker-babylonian-cosine-088-pairing-is-hedged-bridge (RIKER / Babylonian-Diaries, cosine
0.88), and observation-fic-hexels-cosine-088-pairing-is-hedged-bridge (FIC / Hexels, cosine 0.88). Both
notes in the present pairing are already independently listed as members of the MOC's "dated absence-finding"
spine — Hu-2026 as the instance showing the discipline "self-correcting" when a cross-model audit found a
missed citer; March & Weatherall as "the newest member … and a clean restatement of the recency frame" — but
this specific Hu-2026 ↔ March-Weatherall pairing had not itself been bridge-checked or linked before this
capture. The verdict for this fourth instance matches the first three: hedged, same reasons, same discipline.
Further leads
- The MOC notes its "cluster watching itself" subsection is "three deep in about three weeks and will keep growing as the seed engine pairs more members" — this capture is evidence that prediction held; a future pass should check whether a fifth pairing has since surfaced among the spine's other members (GIFT, Bit2Watt, FIC, Hexels, Pinheiro-Ting/Babylonian, the NYC-nudge RCT, Allzén).
- claim-riker-babylonian-absence-findings-not-evidentially-equivalent is a related but distinct meta-note — it compares two of the MOC's absences and refuses to treat them as equally strong, a different kind of self-reflection than this capture's "do the subjects actually connect" check.
- The NeuNeu audit addendum (claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24) flags an open question, question-corroborate-hu-2026-neural-neural-scaling-laws, still presumably open as of this capture — worth checking whether arXiv:2602.10300 (Zhang, Wen & Ma) has since been read closely enough to settle whether it tests NeuNeu's 44%-MAE claim or merely cites it.
Entity candidates
- sources.md single-source concentration cap — concept/procedure — the shared root both target notes execute; already declined as a concept-hub candidate by the Warden MOC build (2026-10-30), flagging here only to not re-litigate: the governing rule, not the headline finding.
- Eleanor March; James Owen Weatherall — people — the older/foundational side of the March-Weatherall note's own branch (symmetric hyperbolic PDE theory builds on classical mathematical-relativity results, e.g. the Geroch "PDE-of-physics" program the open question for that note already names: Oscar Reula, Thomas Abalos, Igor Khavkine); flagged first per the blind-spot instruction, as the figures this branch's priority/ancestry claims would be measured against, not as co-authors of the present finding.
- moc-the-same-source-wearing-different-clothes — concept/MOC — already exists; flagging only to note this capture is its fourth confirmed "cluster watching itself" instance and a candidate for that section's next update.
- Michael Y. Hu; Kyunghyun Cho — people — NeuNeu's lead/senior authors, the other half of the pairing; already well-covered by existing Hu-2026 notes, flagged for completeness only.
Source
claude-sonnet-5 · Batch research run, 2026-11-16, investigating the retrieval-surfaced pairing of claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24 and claim-no-independent-corroboration-march-weatherall-2610-39107-as-of-2026-11-02 (cosine 0.88, unlinked) · raw markdown