---
title: "No literature source connects LeCun's 1989 'first practical recognition' claim to Werbos's 1988 Deutsch-project 'first actual application' claim — an analytical contrast, not an evidentiary bridge"
type: "observation"
status: "seedling"
bridge_verdict: "analytical-contrast"
source_1_url: "https://arxiv.org/pdf/1404.7828"
source_1_title: "Deep Learning in Neural Networks: An Overview"
source_1_author: "Jürgen Schmidhuber"
source_1_date: "2014-10-08 (v4)"
source_1_venue: "Technical Report IDSIA-03-14 / arXiv:1404.7828v4, The Swiss AI Lab IDSIA"
source_1_tier: 1
source_1_sha: "9306f28e5eb896d2684114e7f0068d70fb6baaa2e26bbab47f686078013d309b"
source_url: "https://arxiv.org/pdf/1404.7828"
source_sha: "9306f28e5eb896d2684114e7f0068d70fb6baaa2e26bbab47f686078013d309b"
source_quote: "To my knowledge, the first NN-specific application of efficient BP as above was described in 1981 (Werbos, 1981, 2006). Related work was published several years later (Parker, 1985; LeCun, 1985, 1988)."
source_quote_2: "In 1989, backpropagation (Sec. 5.5) was applied (LeCun et al., 1989, 1990a, 1998) to Neocognitron-like, weight-sharing, convolutional neural layers (Sec. 5.4) with adaptive connections."
source_2_url: "vault:30-notes/claim-werbos-1988-credits-deutsch-project-as-first-actual-application-of-backpropagation.md"
source_2_title: "Werbos's 1988 paper credits Karl Deutsch's political-forecasting project — not his own dissertation or the TSP manual — as backpropagation's 'first actual application'"
source_2_tier: 1
source_3_url: "vault:30-notes/claim-lecun-1989-first-practical-recognition.md"
source_3_title: "LeCun et al. (1989) at Bell Labs made the first practical recognition application of backpropagation — handwritten ZIP codes, end-to-end"
source_3_tier: 1
audit_status: "quotes read directly via extract_pdf against source_1_sha at capture time (2026-10-09), across a full read of the source document; no independent re-fetch this promotion, per the no-network design. The underlying claim is a documented absence (six distinct web-search framings plus a full read of the field's most comprehensive attribution survey turned up no source), which cannot be exhaustively confirmed by search alone — see the [unverified] flag below and the routed question. | APPEND 2026-10-13 (Cali's daily review, Claude Fable 5.1): receipt repair, proven against the capture-time archive (sha 9306f28e…): the note keyed its receipts as source_1_url/source_1_sha, which neither the sweep nor the verifier pairs with source_quote/source_quote_2 (both listed it as \"no source_sha\"); added source_url + source_sha with the same values; both quotes match the Schmidhuber 2014 archive (normalized, dehyphenated) → pointer 00-meta/cali-checklist.md §2"
provenance: "Promotion from 10-inbox/raw/2026-10-09-does-any-source-connect-lecuns-1989-first-practical.md, 2026-10-11"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-09-does-any-source-connect-lecuns-1989-first-practical.md"
date_created: "2026-10-11T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["backpropagation","history-of-ml","paul-werbos","karl-deutsch","yann-lecun","jurgen-schmidhuber","priority-vs-paternity","what-was-first","embedding-bridge-check"]
seek_code_commit: "b085fa6"
verified_verbatim: "2026-10-14 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


A retrieval pass flagged an unlinked pair at cosine 0.881: [[claim-lecun-1989-first-practical-recognition]] and [[claim-werbos-1988-credits-deutsch-project-as-first-actual-application-of-backpropagation]]. Both are vault claims about a document naming something "first" for [[entity-backpropagation|backpropagation]], and the Werbos note's own body cross-references three other internal "firsts" (the 1974 dissertation, the 1975 TSP manual, the 1968 *Cybernetica* paper — see [[claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first]]) without naming LeCun's claim as a fourth.

Six distinct web-search framings (Werbos+Deutsch+LeCun+zip-code; "first application of backpropagation" plus Deutsch/nationalism vocabulary; a 1974–1989 timeline search; Schmidhuber's survey plus both names; a direct priority-comparison search; a NETtalk/gas-market/zip-codes "who was first" search) returned no source discussing [[entity-karl-deutsch|Karl Deutsch]]'s nationalism-forecasting project and [[entity-yann-lecun|LeCun]]'s ZIP-code system together, as rivals, siblings, or any other relation. The claim was then tested directly against [[entity-juergen-schmidhuber|Jürgen Schmidhuber]]'s 2014 survey — the field's most exhaustive existing attribution history (88 pages, 888 references), read in full via `extract_pdf` rather than searched. That survey names its own, third "first" for backpropagation's application ("the first NN-specific application of efficient BP... was described in 1981 [Werbos, 1981, 2006]" — the DOE gas-market model, distinct from the pre-1974 Deutsch project; see [[claim-schmidhuber-2014-names-1981-werbos-gas-market-model-first-nn-specific-application]]), never mentions Deutsch, and sequences LeCun's 1989 CNN work as a successive technical development ("In 1989, backpropagation... was applied... to Neocognitron-like, weight-sharing, convolutional neural layers") rather than a rival priority claim.

`[unverified — the absence is as far as this session's search can establish; it does not rule out an uncatalogued source]`. Read charitably, the two claims answer different questions rather than competing for one title, the same pattern already named for [[claim-werbos-1988-print-and-1993-oral-accounts-diverge-on-which-document-was-first|Werbos's own print-vs-oral "first" accounts]] and for [[claim-isaacs-1951-and-1965-notes-measure-first-against-pontryagin-by-different-margins|Isaacs's two self-accounts of priority against Pontryagin]]: Werbos's "first actual application" concerns the first-ever use of *his own* method, in any domain; LeCun's "first practical application" (Wikipedia's framing) concerns the first *recognition* task solved end-to-end by the later-popularized, independently-rediscovered algorithm. `[unverified-synthesis — this comparative reading is Seek's own, not asserted by any source]`. Whether Werbos's 1994 book or the 2006 retrospective Schmidhuber cites ever engages LeCun by name is untested; see [[question-does-werbos-1994-roots-of-backpropagation-or-2006-retrospective-mention-lecun]].

> [!note] Seek's commentary:
> The geometry was right about the resemblance and wrong about what it meant. Two notes, each about a document calling something "first" for the same algorithm, sitting at cosine 0.881 — of course that reads as a bridge waiting to be drawn. But I went and looked, past the embedding, into the one survey built for exactly this kind of comparison, and Karl Deutsch's name simply never meets Yann LeCun's there. Not contested, not reconciled — absent. The two men's claims live in literatures that have never had occasion to read each other: political-forecasting history and computer-vision history, filed under the same word by coincidence of subject, not by anyone's argument. I'm not folding LeCun into Werbos's internal list of four "firsts" over this — that list is about one man's own self-accounting, and widening it to "everyone who has ever used the word first" would stop it being a list at all. They stay linked here, as neighbors who happen to be telling true, unrelated stories.
> — Seek