---
id: "20260719-0239-hop-collaborative-convergence-not-mutual-unawareness"
title: "The En-Gedi/backprop resemblance is superficial — virtual unwrapping's history is acknowledged convergence, not mutual unawareness, and the vault already has the axis to place it on"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-seales-2016-en-gedi-paper-cites-mocella-rival-claim.md","30-notes/observation-seales-mocella-is-third-pole-of-multiple-discovery-axis.md","40-entities/entity-vito-mocella.md"]
not_promoted: ["The flagged-pairing-is-a-false-positive verdict itself: duplicates observation-ml-ad-en-gedi-cosine-pairing-is-embedding-false-friend.md, already promoted today (2026-07-20) from a sibling capture (2026-07-18-hop-embedding-false-friend-ml-ad-en-gedi.md) that reaches the same conclusion by the same route. Linked from the new notes instead of re-promoted.","The self-cosine top-5 check (neither seed note in the other's top-5 at 0.877/0.832): retrieval-mechanics evidence supporting the false-friend verdict above, not an independently atomic claim.","No-priority-dispute / prior 2013 Seales-Delattre collaboration detail: folded as supporting context into claim-seales-2016-en-gedi-paper-cites-mocella-rival-claim.md rather than given its own note.","Further-leads item (Mocella's own retrospective on the exchange, unread): a nice-to-verify curiosity, not a load-bearing gap under either kept note; left in this capture rather than routed to 50-questions/."]
origin: "hop-batch"
writer_model: "claude-sonnet-5"
date_created: "2026-07-19T00:00:00.000Z"
hop_chain: ["seed: claim-ml-and-ad-communities-mutually-unaware & claim-virtual-unwrapping-first-proved-on-2015-en-gedi-scroll -> direct vault_novelty self-check on each note's own title text (max_cosine 0.877 / 0.832, neither note appears in the other's top-5)","Wikipedia 'Virtual unfolding' + Seales/Vesuvius Challenge web history -> checked whether virtual unwrapping's development shows the same 'isolated communities' structure as backprop (max_cosine 0.706)","Vesuvius Challenge funders hook -> vault's existing priority-without-paternity cluster, claim-nakamoto-bitcoin-leaned-on-wei-dai-b-money (max_cosine 0.672)","claim-linnainmaa-priority-not-paternity -> searched for a Seales/Mocella priority dispute mirroring Linnainmaa/Rumelhart (web search, no vault match)","Mocella search -> Seales et al. Science Advances/PMC primary source, direct citation check (Tier 1, no vault match)","PMC primary -> moc-backpropagation-origins.md, read to confirm the cluster's defining pattern (vault read)","MOC read -> observation-parker-and-labview-max-are-opposite-poles-of-multiple-discovery via vault_novelty on the synthesis claim (max_cosine 0.798, adjacent)"]
novelty_max_cosine: 0.798
tags: ["multiple-discovery","historiography","virtual-unwrapping","automatic-differentiation","backpropagation","priority","cross-domain-bridge"]
source_url: "https://pmc.ncbi.nlm.nih.gov/articles/PMC5031465/"
source_author: "Seales et al."
source_date: "2016-09-21T00:00:00.000Z"
source_venue: "Science Advances"
source_tier: 1
---


The vault flagged [[claim-ml-and-ad-communities-mutually-unaware]] and
[[claim-virtual-unwrapping-first-proved-on-2015-en-gedi-scroll]] as an unlinked
pair at cosine 0.75. Re-running the check directly against each note's own
title text found neither note in the other's top-5 closest matches at all
(self-cosine 0.877 and 0.832 respectively) — the resemblance is thin even in
embedding space.

It is also thin structurally. Baydin et al. 2018 (Tier 1) describe ML and AD
as fields that "have largely been unaware of each other and, in some cases,
have independently discovered each other's results" — genuine isolation.
Virtual unwrapping's history is the opposite: Seales's own 2016 Science
Advances paper on the En-Gedi scroll explicitly engages a rival group's
concurrent work — "The work of Mocella et al. (12) claims that phase-contrast
tomography generates contrast at ink boundaries in scans of material from
Herculaneum" (source above, Tier 1). No priority dispute or citation gap
turned up; the teams were reading each other in real time.

The vault already has the right frame for this contrast:
[[observation-parker-and-labview-max-are-opposite-poles-of-multiple-discovery]]
names a "multiple discovery axis" running from real-independence-forgotten
(Parker/backprop) to apparent-independence-over-remembered (LabVIEW/Max).
Seales/Mocella is a third position: acknowledged, citation-based convergence
— never independent at all, and never forgotten either. That's the load-bearing
difference between the two seed notes, not a bridge between them.

## Why this was hop-worthy
A stated 0.75 cosine turned out to rest on shared genre language (decades-long
research arcs, belated recognition) rather than shared causal structure — and
chasing it down landed on a vault framework built for exactly this kind of
contrast.

## Further leads
- Vito Mocella's own account of the Seales exchange is unread; would sharpen whether "aware" also meant "amicable."

## Entity candidates
- Vito Mocella — person — physicist behind the 2015 rival phase-contrast tomography claim on Herculaneum papyri; no vault entity page, surfaced via the priority-dispute check that came up empty.

## Hop chain

Hop 1: claim-ml-and-ad-communities-mutually-unaware & claim-virtual-unwrapping-first-proved-on-2015-en-gedi-scroll (vault notes)
- Hook type: surprising claim (the seed's stated bridge itself)
- Hook: cosine 0.75 pairing flagged as unlinked
- Why followed: the prompt's explicit task — verify whether the bridge is real
- Key findings: read both notes; the AD/ML note's structure is "two literatures never read each other, so both reinvented the same algorithm." The virtual-unwrapping note's structure is "one lab's 20-year program paid off," a different shape on its face.

Hop 2: https://en.wikipedia.org/wiki/Virtual_unfolding + Seales/Vesuvius Challenge web search
- Hook type: mechanism question
- Hook: was virtual unwrapping's field also split into mutually-unaware communities?
- Why followed: the only way the bridge could be real is if the same isolation pattern existed on the archaeology side
- Key findings: coverage describes "technological convergence" — teams adapting known medical CT tomography together across French/British/American/Chinese groups — not isolated parallel reinvention.

Hop 3: claim-nakamoto-bitcoin-leaned-on-wei-dai-b-money / claim-linnainmaa-priority-not-paternity
- Hook type: cross-domain bridge
- Hook: a vault_novelty check on "Vesuvius Challenge funders" surfaced the Nakamoto/b-money note as closer than the virtual-unwrapping note itself
- Why followed: the vault's real "priority without paternity" cluster spans ML and cryptocurrency — worth checking before concluding archaeology doesn't belong to it
- Key findings: confirmed the cluster (Linnainmaa/Rumelhart, Amari/Hopfield, Nakamoto/Dai/Szabo) is defined by genuine non-transmission between named parties — a bar virtual unwrapping needed to clear and, per hop 4-5, didn't.

Hop 4: web search, Mocella 2015 vs. Seales priority dispute
- Hook type: surprising claim (testing for a hidden priority story)
- Hook: a 2015 Nature Communications paper (Mocella et al.) on Herculaneum tomography predates Seales's En-Gedi publication by months
- Why followed: this is the closest real-world analogue to Linnainmaa/Rumelhart if the awareness gap existed
- Key findings: no dispute surfaced; Seales and French collaborator Delattre had already co-published in 2013.
- Surprise: expected a plausible uncredited-precursor story similar to Linnainmaa/Rumelhart — found explicit mutual citation and prior collaboration instead.

Hop 5: https://pmc.ncbi.nlm.nih.gov/articles/PMC5031465/ (Seales et al., Science Advances, 2016)
- Hook type: mechanism question / primary verification
- Hook: does the primary paper itself cite Mocella?
- Why followed: zoom in to confirm hop 4's finding against the actual paper rather than secondary summaries
- Key findings: direct quote confirms Seales cites and characterizes Mocella's claim in the same paper — real-time, acknowledged engagement.
- Surprise: expected the primary source might soften or omit the rival citation — found it stated plainly and specifically.

Hop 6: 40-mocs/moc-backpropagation-origins.md
- Hook type: mechanism question (zoom out to the cluster's own self-definition)
- Hook: the MOC's opening line — "derived repeatedly, in different fields, by people who mostly didn't know about each other"
- Why followed: confirm the cluster's admission bar before ruling virtual unwrapping out of it
- Key findings: the cluster is explicitly about non-transmission; nothing in virtual unwrapping's history matches that bar.

Hop 7: observation-parker-and-labview-max-are-opposite-poles-of-multiple-discovery
- Hook type: cross-domain bridge
- Hook: vault_novelty on the synthesis claim (0.798, just under duplicate-risk) surfaced this note as the closest existing framework
- Why followed: highest-band frontier hit, and it names an axis this finding fits precisely
- Key findings: the vault already has a "multiple discovery axis" (Merton/Chandler) with two named poles; Seales/Mocella is a clean third point on that same axis (acknowledged convergence), giving the capture a real home instead of a forced link.

Saved hooks not followed:
- EduceLab's $14M NSF grant mechanics — from claim-virtual-unwrapping-first-proved-on-2015-en-gedi-scroll — interesting institutional-funding detail but not load-bearing for the bridge question.
- Nat Friedman / Daniel Gross as tech-VC funders of archaeology — from Vesuvius Challenge research — a genuine "person behind the thing" / cross-domain hook (Silicon Valley money → classics), saved for a future chain rather than this verification-focused one.
- Vito Mocella's own retrospective on the exchange — saved as an entity candidate above.

post-worthy: maybe — the finding is a clean negative result plus a genuine reuse of an existing vault framework (the multiple-discovery axis), but it's more "correcting a false positive" than a new discovery.