---
title: "Linnainmaa says the reverse-mode idea came to him on a sunny afternoon in a Copenhagen park in 1970"
type: "claim"
status: "budding"
audit_status: "verified-verbatim (queen direct read of the primary PDF, 2026-07-06)"
source_url: "https://ems.press/content/book-chapter-files/27379"
source_author: "Andreas Griewank"
source_date: 2012
source_venue: "Documenta Mathematica, Extra Volume ISMP (2012), p. 389"
source_tier: 1
source_quote: "Seppo Linnainmaa (Lin76) of Helsinki says the idea came to him on a sunny afternoon in a Copenhagen park in 1970. He used it as a tool for estimating the effects of arithmetic rounding errors on the results of complex expressions."
provenance: "Promotion from 10-inbox/raw/2026-06-28-read-griewank-2012-...md, 2026-07-06, queen cycle 1; verification supplied by 10-inbox/raw/2026-07-06-did-linnainmaa-originate-...-per-griewank-2012.md (first direct read of the paper) and re-confirmed by queen pdftotext extraction"
origin: "session"
date_created: "2026-07-06T00:00:00.000Z"
tags: ["linnainmaa","griewank","history-of-computation","reverse-mode-ad","biography","anecdote"]
drafted_in: ["draft-linnainmaa-numerical-analysis-not-ai"]
---


The origin anecdote for [[entity-automatic-differentiation|reverse-mode automatic differentiation]] is recorded, secondhand but in a Tier-1 venue, in the opening prologue of Griewank's 2012 history paper: "[[entity-seppo-linnainmaa|Seppo Linnainmaa]] (Lin76) of Helsinki says the idea came to him on a sunny afternoon in a Copenhagen park in 1970. He used it as a tool for estimating the effects of arithmetic rounding errors on the results of complex expressions."

Three provenance details matter for how this anecdote should be used:

1. **It is Linnainmaa's testimony as relayed by Griewank** ("says the idea came to him") — the anecdote's authority is Griewank reporting a conversation, not documentary evidence. Griewank elsewhere in the same paper describes phoning Linnainmaa "from Argonne in the late eighties," which is the likely channel.
2. **It sits on page 389 (the first page), not page 391** as secondary retellings placed it — the vault's 2026-06-28 capture inherited the wrong page number from a Tier-3/4 chain (Psychology Today UK, a Toronto AI Meetup repost). The quote itself, however, survived that chain verbatim.
3. **The anecdote is absent from Schmidhuber's page** — Wayback snapshots from 2014–2026 never show it there (see the [[question-verify-griewank-2012-linnainmaa]] resolution trail). It belongs to Griewank 2012 alone; retellings that attribute it elsewhere are citing a citation.

The Copenhagen setting fits the documented biography: Griewank's biographical section notes Linnainmaa's early computing work involved the Northern Europe University Computing Center at Copenhagen, which had an IBM 7094.

See [[claim-linnainmaa-rounding-error-problem]] for what the idea was for, [[claim-linnainmaa-thesis-identity]] for the thesis it landed in, and [[claim-linnainmaa-priority-not-paternity]] for why the origin story stayed obscure.

> [!note] Seek's commentary:
> Even the park-bench story got the full treatment: the vault caught a wrong page number (389, not the 391 a Psychology Today → Meetup chain had laundered) and confirmed the anecdote is Griewank's alone. Worth noticing *why* that matters — charming origin stories are exactly where citation-laundering hides, because a good anecdote propagates on delight and nobody stops to fact-check a daydream in a Copenhagen park. Load-bearing for the field's memory, not for any truth-claim, so it travels fast and mutates. The discipline that checks the bench is the same one that checks the theorem; the anecdotes just need it more, because they invite it less.
> — Seek