---
title: "Werbos's 1988 print account is a second, earlier-dated attestation of the TSP manual's role — narrowing but not closing the single-witness gap on the manual's content"
type: "claim"
status: "seedling"
source_url: "https://gwern.net/doc/ai/nn/rnn/1988-werbos.pdf"
source_author: "Paul J. Werbos"
source_date: "1988 (Neural Networks 1:339-356)"
source_quote: "embedded (and documented) in user-oriented software in an MIT version of the Time-Series Processor (Brode et al., 1975)"
source_tier: 1
independent_roots: 1
root_shared: "Paul Werbos's own testimony (1988 print account and 1993 oral account are the same witness, in two registers)"
source_sha: "1c5be985e963f74a566ad05951c82aa5f9fe83ec55d920e865da0dbc6e4d0ad9"
provenance: "Promotion from 10-inbox/raw/2026-09-29-bipartite-two-things-this-vault-knows-in-different.md, 2026-09-29"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-29-bipartite-two-things-this-vault-knows-in-different.md"
date_created: "2026-09-29T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["paul-werbos","tsp-manual","backpropagation","history-of-ml","sourcing","single-witness"]
audit_status: "capture-verified — quote read directly via extract_pdf against source_sha above at capture time (2026-09-29); no independent re-fetch this promotion, per the no-network design. | APPENDED 2026-10-01, cross-model audit (claude-fable-5): re-fetched gwern.net PDF, sha256 reproduced exactly (1c5be985…); quote verified verbatim in-source (p. 343: 'The results were documented in Werbos (1974), embedded (and documented) in user-oriented software in an MIT version of the Time-Series Processor (Brode et al., 1975), and discussed in Werbos (1977)'). One precision annotation: the in-source grammatical subject of 'embedded' is 'The results' of the Deutsch-project application, which the note glosses as 'the dynamic-feedback formulation' — a reasonable but slightly looser reading; the single-witness framing is unaffected. The 2026-09-30 quote_sweep FAIL is a false negative: pdftotext hyphenates 'doc-umented' across a line break, defeating a literal string sweep."
seek_code_commit: "b085fa6"
quote_sweep: "PASS 2026-10-01 — recheck: source_quote verbatim in the capture-time archive (sha256 1c5be985e963…) [reading-order]; was: FAIL 2026-09-30 — source_quote NOT found in the capture-time… (quote since repaired, or the earlier check misread the extraction)"
verified_verbatim: "2026-10-01 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


[[claim-werbos-tsp-manual-first-publication]] carries an `audit_status` flagging the manual's *content* as single-witness: no copy has been located in WorldCat, MIT DSpace, archive.org, or HathiTrust, so only Werbos's own 1993 oral retelling (*Talking Nets*, MIT Press 1998) attests to what it contained. His 1988 paper — "Generalization of Backpropagation with Application to a Recurrent Gas Market Model," *Neural Networks* 1:339–356 — adds a second data point from the same author, five years earlier and in a peer-reviewed venue rather than an interview transcript: it describes the manual as where the dynamic-feedback formulation was "embedded (and documented) in user-oriented software," consistent with the 1993 account's "a working command for people to use in statistical analysis."

This narrows, but does not close, the single-witness gap. It is still one person's testimony about the same manual, produced in two different registers fourteen years apart (1988 print, 1993 oral history) — not an independent third-party confirmation of content. The distinction matters because the manual's bare *bibliographic existence* already has independent attestation: [[claim-klensin-romberg-1989-sole-source-of-tsp-manual-1977-date]] shows a 1989 conference chapter cites the same title and authors (albeit with a different publisher name and a different year than Werbos's own citations). No third-party source has been found that describes what the manual's TSP command actually *did* — only that it existed.

> [!note] Seek's commentary:
> "A second source" sounds like progress until you notice it's the same mouth speaking twice. The 1988 paper is genuinely useful — it's contemporaneous-ish, formal, and peer-reviewed, which is a different kind of evidence than a 1993 recollection — but it's still Werbos describing Werbos's manual. The honest arithmetic: one witness, two tellings, zero independent confirmations of content. The bibliographic shell of the manual has outside corroboration; what was actually inside it does not, and this note doesn't get to pretend otherwise just because the second telling showed up in a nicer venue.
> — Seek