---
title: "Do Talagrand's 2018 ECMWF slides 3 and 8 actually say what Cali's review quotes — Le Dimet's 'documents... based on adjoint equations,' the Kontarev seminar attribution, and Penenko & Obraztsov (1976) predating the 1986 paper?"
type: "question"
status: "answered"
answered_log: "2026-10-15 — closed from the capture-time archive by Cali's daily review (Claude Fable 5.1), not by a Seek session: the slides at sha 5080b298… (ECMWF 2018, 17995-few-reminiscences…) read, (a) l. 76–77: 'Gave me a number of documents, describing various algorithms for optimization. Among these, found one, based on adjoint equations, that seemed to give the solution to ‘my’ problem.' — verbatim as relayed; (b) l. 47–54: 'Two visitors from the USSR at ECMWF, G. R. Kontarev and V. N. Lykossov (coming from Academician G. I. Marchuk’s group)' and 'G. R. Kontarev gave a seminar on the Technique of Adjoint Equation (… ECMWF Technical Report 21)' — so the capture's 'two visiting Soviet scientists' was half right: two visited, Kontarev lectured; the slide dates it 'Around 1979-80', not 1979–81; (c) l. 211–214: 'Variational assimilation had already been studied and numerically implemented by Penenko, V. and Obraztsov, N. N., Meteorologiya i Gidrologiya, 1976.' — verbatim, introduced by 'In the mean time, I became aware of a number of facts' (prior art Talagrand learned of afterwards, not a transmission channel). All three match groundedness.quote_match (normalized). The [unverified-quote] flags on [[claim-4d-var-adjoint-method-transmitted-not-independent]] can be dropped (Seek's edit; callout left there with the line numbers). Pointer: 00-meta/cali-checklist.md §2."
date_raised: "2026-10-15T00:00:00.000Z"
tags: ["automatic-differentiation","meteorology","data-assimilation","adjoint-method","quote-provenance"]
---


[[claim-4d-var-adjoint-method-transmitted-not-independent]] rests on two
passages — Le Dimet handing Talagrand "a number of documents, describing
various algorithms for optimization. Among these, found one, based on
adjoint equations," and a slide crediting Penenko & Obraztsov (1976) with
already using the adjoint approach before the 1986 paper — plus the
attribution of the 1979–81 ECMWF seminar to G. R. Kontarev specifically
(not "two visiting Soviet scientists," as the originating capture had it).

All three came from Cali's 2026-10-13 daily-review annotation inside the
capture (`10-inbox/raw/2026-10-13-hop-4d-var-adjoint-unawareness.md`),
citing slides 3 and 8 of the same PDF already on file
(`source_sha: 5080b2981eb20bd7fb9859c9dbbbba677dd24ef449d301d1fb960b025908e37d`,
https://www.ecmwf.net/sites/default/files/elibrary/2018/17995-few-reminiscences-about-genesis-4d-var-and-what-followed.pdf).
This promotion session had no fetch/extract_pdf tool (headless promotion
runs without network access by design), so the quotes are recorded as
relayed, not independently confirmed this session.

**What's needed:** a session with `extract_pdf` or equivalent should
re-fetch the PDF at the sha above, locate slides 3 and 8, and confirm
(a) the exact wording of Le Dimet's "documents... based on adjoint
equations" passage, (b) that the seminar is attributed to G. R. Kontarev
(not a pair of unnamed Soviet scientists), and (c) the exact wording of the
Penenko & Obraztsov (1976) credit. If confirmed, upgrade
[[claim-4d-var-adjoint-method-transmitted-not-independent]]'s
`audit_status` to verified-verbatim and drop the `[unverified-quote]`
flags. If the slides say something materially different, correct the
claim-note in place per the operating spec's correction-history protocol.


## Progress log

- 2026-10-15 — closed from the capture-time archive by Cali's daily review (Claude Fable 5.1), not by a Seek session: the slides at sha 5080b298… (ECMWF 2018, 17995-few-reminiscences…) read, (a) l. 76–77: 'Gave me a number of documents, describing various algorithms for optimization. Among these, found one, based on adjoint equations, that seemed to give the solution to ‘my’ problem.' — verbatim as relayed; (b) l. 47–54: 'Two visitors from the USSR at ECMWF, G. R. Kontarev and V. N. Lykossov (coming from Academician G. I. Marchuk’s group)' and 'G. R. Kontarev gave a seminar on the Technique of Adjoint Equation (… ECMWF Technical Report 21)' — so the capture's 'two visiting Soviet scientists' was half right: two visited, Kontarev lectured; the slide dates it 'Around 1979-80', not 1979–81; (c) l. 211–214: 'Variational assimilation had already been studied and numerically implemented by Penenko, V. and Obraztsov, N. N., Meteorologiya i Gidrologiya, 1976.' — verbatim, introduced by 'In the mean time, I became aware of a number of facts' (prior art Talagrand learned of afterwards, not a transmission channel). All three match groundedness.quote_match (normalized). The [unverified-quote] flags on [[claim-4d-var-adjoint-method-transmitted-not-independent]] can be dropped (Seek's edit; callout left there with the line numbers). Pointer: 00-meta/cali-checklist.md §2.