---
id: "20260811-0232-verify-the-kahan-intel"
title: "Verify the Kahan / Intel-8087 / IEEE-754 lineage against a primary Kahan source"
type: "capture"
status: "promoted"
date_promoted: "2026-08-11T00:00:00.000Z"
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-08-11T00:00:00.000Z"
provenance: "this batch run, 2026-08-11"
derived_from: []
promoted_to: ["30-notes/claim-kahan-8087-spec-became-ieee-754.md (corrected in place, not a new file — Claim 2's K-C-S quote replaced the existing note's secondary-sourced placeholder and discharged its `flagged` audit_status)","30-notes/claim-palmer-recruited-kahan-as-consultant-for-i8087-arithmetic.md (Claim 1)","30-notes/claim-kahan-declined-to-bill-intel-for-8087-work-transferred-to-ieee-754.md (Claim 3)","30-notes/claim-ieee-754-extended-precision-explicit-bit-inherited-from-intel-8087.md (Claim 4)","40-entities/entity-william-kahan.md (new hub — surprising this didn't already exist given the vault's prior claim-note about him)","40-entities/entity-john-palmer.md (new hub)","40-entities/entity-jerome-coonen.md (new hub)","40-entities/entity-harold-stone.md (new hub)","50-questions/question-verify-kahan-8087-ieee754-lineage-primary.md (ruled answered — the lineage-verification crux is settled; see its Resolution section)"]
not_promoted: ["Entity candidate J.H. Wilkinson — real, foundational to error-analysis, and named as an endorser of K-C-S, but the capture only paraphrases this (no verbatim quote in the four sourced Claim sections); did not build a hub with zero grounding claim-notes. Left as a mention; would promote properly from a session that reads Kahan's account of Wilkinson's endorsement directly.","Entity candidates Robert Stewart, Mary Payne, David Stevenson, G.W. (Pete) Stewart III, George Taylor, Fred Ris, Cleve Moler, Charles Severance — same reasoning: named in the capture's entity-candidates list or 'Further leads' but not grounded in a verbatim-quoted claim this session. Not promoted.","Donald Knuth — already has a hub page (40-entities/entity-donald-knuth.md, about the Forsythe/'who coined computer science' question). This capture names him as a K-C-S early endorser, but again only in the bee's paraphrase, not a quoted primary passage. Did not append an unsourced fact to an existing hub; the page is otherwise untouched.","The SIAM oral history (Haigh, 2005), the ACM Turing Award citation's exact wording (403'd to automated fetch this session, both amturing.acm.org and awards.acm.org), the printed IEEE Computer version of the interview, and Beebe's Kahan bibliography — all named in the capture's 'Further leads' as unread or unconfirmed. No kept claim rests on any of them, so per question-intake discipline none was routed to a new 50-questions/ entry; they stay leads."]
tags: ["william-kahan","intel-8087","ieee-754","floating-point","standards","history-of-computing","numerical-analysis","verification"]
seek_code_commit: "b13747c"
---


Direct follow-up to [[question-verify-kahan-8087-ieee754-lineage-primary]], which flagged that [[claim-kahan-8087-spec-became-ieee-754]]'s load-bearing causal claim — that IEEE 754 "grew directly out of Intel's need for a defensible in-hardware arithmetic spec" — rested only on secondary sources (a UC Berkeley Math Dept news page, an ETHW milestone page) with no verbatim Kahan quote.

This session located and read Kahan's own primary account, hosted on his own UC Berkeley EECS site: "An Interview with the Old Man of Floating-Point," reminiscences elicited from William Kahan by Charles Severance (20 Feb. 1998), the extended web version of the abbreviated interview that ran in the March 1998 issue of *IEEE Computer*. It also pulls one corroborating technical passage from Kahan's own "Lecture Notes on the Status of IEEE Standard 754," a self-published status report also hosted on his site.

**Result: the lineage holds, in Kahan's own words, and is more specific than the flagged note's summary.** Kahan does not merely say the 8087 work "fed into" 754 — he narrates the specific mechanism: Intel granted him permission to bring the arithmetic he had designed for the 8087 to the IEEE p754 committee as a draft, that draft (K-C-S) became the standard's basis, and he declined to bill Intel for the hours spent on the parts of the design that made that transfer. He also names one respect in which the lineage was *not* total: some 8087-specific implementation choices were excluded from 754 as "too arcane."

## Claim: Intel recruited Kahan as a paid consultant, specifically to design the arithmetic for the i8087 floating-point coprocessor (and the ill-fated i432), after its floating-point-effort manager had encountered his analysis of commercial arithmetics years earlier at Stanford

Dr. John Palmer, managing Intel's floating-point effort, had heard Kahan — then a visiting professor — analyze the anomalies of commercially significant arithmetics at Stanford roughly a decade earlier. Palmer persuaded Intel it needed a defensible arithmetic standard so that different Intel products would not compute disparate results, and recruited Kahan as a consultant.

> "Palmer, now the manager of Intel's floating-point effort, recruited Kahan as a consultant to help design the arithmetic for the i432 ( which died later ) and for the i8086/8's upcoming i8087 coprocessor."

> "Intel had decided they wanted really good arithmetic. I suggested that DEC VAX's floating-point be copied because it was very good for its time. But Intel wanted the `best' arithmetic. ... He and I put together feasible specifications for that `best' arithmetic."

source_quote: "Palmer, now the manager of Intel's floating-point effort, recruited Kahan as a consultant to help design the arithmetic for the i432 ( which died later ) and for the i8086/8's upcoming i8087 coprocessor."
source_url: https://people.eecs.berkeley.edu/~wkahan/ieee754status/754story.html
source_sha: d9048db6e4a38b0e64c99bf41eb55a5166af138dff2594fc2d8daa14b6f6d064
source_title: "An Interview with the Old Man of Floating-Point"
source_author: "William Kahan (reminiscences elicited by Charles Severance)"
source_venue: "William Kahan's UC Berkeley EECS homepage (people.eecs.berkeley.edu/~wkahan); underlies the abbreviated version published in IEEE Computer, vol. 31 no. 3, March 1998, pp. 114-115"
source_date: 1998-02-20
source_tier: 1

Historical/biographical claim (who did what), narrated in Kahan's own words in an authorized reminiscence hosted on his own site — clears the floor well above the Tier 3-4 minimum for uncontested biography.

## Claim: Kahan, with his student Jerome Coonen and visiting Prof. Harold Stone, wrote a draft floating-point specification — based on the arithmetic he had worked out for Intel — that he brought to the IEEE p754 committee, where it was adopted as the working draft ("K-C-S") that became IEEE 754

After the second p754 meeting (Nov. 1977), Kahan got Intel's permission to participate in the standards effort and to disclose the reasoning behind the 8087's numeric specifications (though not its architecture). He and his co-authors then produced a formal IEEE-style draft based on that work, which the committee took up as its central proposal.

> "After that 1977 meeting Kahan went back to Intel and requested permission to participate in the standard effort. With permission granted, Kahan and his student Jerome Coonen at U.C. Berkeley, and a visiting Prof. Harold Stone, prepared a draft specification in the format of an IEEE standard and brought it back to an IEEE p754 meeting. This draft was called "K-C-S" until p754 adopted it."

> "I had worked out the details initially for Intel. My reasoning was based on the requirements of a mass market ... What distinguished our proposal from all the others was the reasoning behind every decision."

By 1984 — a year before formal ratification — the K-C-S-derived draft had already been implemented as a de facto standard by Intel, AMD, Apple, ELXSI, IBM, Motorola, National Semiconductor, Weitek, Zilog, and AT&T, among others, according to Kahan's own account.

source_quote: "Kahan and his student Jerome Coonen at U.C. Berkeley, and a visiting Prof. Harold Stone, prepared a draft specification in the format of an IEEE standard and brought it back to an IEEE p754 meeting. This draft was called \"K-C-S\" until p754 adopted it."
source_url: https://people.eecs.berkeley.edu/~wkahan/ieee754status/754story.html
source_sha: d9048db6e4a38b0e64c99bf41eb55a5166af138dff2594fc2d8daa14b6f6d064
source_title: "An Interview with the Old Man of Floating-Point"
source_author: "William Kahan (reminiscences elicited by Charles Severance)"
source_venue: "William Kahan's UC Berkeley EECS homepage (people.eecs.berkeley.edu/~wkahan); underlies the abbreviated version published in IEEE Computer, vol. 31 no. 3, March 1998, pp. 114-115"
source_date: 1998-02-20
source_tier: 1

This is the specific technical-mechanism claim the routed question asked to verify: not a vague "grew out of" but a named authorship chain (Kahan-Coonen-Stone, "K-C-S") moving a specific document from a corporate consulting engagement into the IEEE committee process. It clears the Tier 1-2 floor required for a technical-mechanism claim directly, on Kahan's own words.

## Claim: Kahan explicitly frames the transfer of specific i8087 design work into the public IEEE p754 standard as intentional and uncompensated — he declined to bill Intel for consulting hours spent on the parts of the 8087 design that moved into the standard

This is the most direct primary confirmation of the causal direction the flagged note needed: Kahan's own account states, without qualification, that identifiable "aspects of the i8087 design" were "transferred to IEEE p754," and that he treated this transfer as a matter of professional ethics requiring him not to profit from it twice.

> "I did not bill Intel for consulting hours spent on those aspects of the i8087 design that were transferred to IEEE p754. I had to be sure, not only in appearance and actuality but above all in my own mind, that I was not acting to further the commercial interest of one company over any other."

source_quote: "I did not bill Intel for consulting hours spent on those aspects of the i8087 design that were transferred to IEEE p754."
source_url: https://people.eecs.berkeley.edu/~wkahan/ieee754status/754story.html
source_sha: d9048db6e4a38b0e64c99bf41eb55a5166af138dff2594fc2d8daa14b6f6d064
source_title: "An Interview with the Old Man of Floating-Point"
source_author: "William Kahan (reminiscences elicited by Charles Severance)"
source_venue: "William Kahan's UC Berkeley EECS homepage (people.eecs.berkeley.edu/~wkahan); underlies the abbreviated version published in IEEE Computer, vol. 31 no. 3, March 1998, pp. 114-115"
source_date: 1998-02-20
source_tier: 1

Technical-mechanism / causal claim, Tier 1, Kahan's own words. This is the sentence that discharges [[claim-kahan-8087-spec-became-ieee-754]]'s `audit_status: flagged` — the direction of causation (commercial coprocessor spec → public standard) is stated plainly by the person who did the work, not inferred from a secondary retelling.

## Claim: The lineage was selective, not total — one 8087-specific implementation detail (an explicit leading significand bit in extended precision) survived into IEEE 754's optional Double-Extended format for historical/compatibility reasons, while other 8087-specific features were excluded as "too arcane"

Kahan's technical lecture notes on the standard's status qualify the simple "8087 spec became 754" story: the extended-precision format's explicit (non-implicit) leading bit is a direct 8087 inheritance, kept because it let the 8087 skip normalizing subnormal numbers in certain matrix-product computations — but this was a narrow exception, and other 8087-specific features that came up during the standards work were left out.

> "680x0 / ix87 Extendeds have an explicit Nth bit for historical reasons; it allowed the Intel 8087 to suppress the normalization of subnormals advantageously for certain scalar products in matrix computations, but this and other features of the 8087 were later deemed too arcane to include in IEEE 754, and have atrophied."

source_quote: "680x0 / ix87 Extendeds have an explicit Nth bit for historical reasons; it allowed the Intel 8087 to suppress the normalization of subnormals advantageously for certain scalar products in matrix computations, but this and other features of the 8087 were later deemed too arcane to include in IEEE 754, and have atrophied."
source_url: https://people.eecs.berkeley.edu/~wkahan/ieee754status/IEEE754.PDF
source_sha: 7904d3c42d8160c295b3275639f2d836367e7a600cf0020c929d0ed275b57df2
source_title: "Lecture Notes on the Status of IEEE Standard 754 for Binary Floating-Point Arithmetic"
source_author: "William Kahan"
source_venue: "Self-published status report, Elect. Eng. & Computer Science, University of California, Berkeley; hosted on Kahan's own UC Berkeley EECS site"
source_date: 1997-10-01
source_tier: 1

Technical-mechanism claim, Tier 1, Kahan's own words. Useful corrective to a too-clean version of the lineage claim: the transfer from 8087 to IEEE 754 was selective engineering judgment, not a wholesale copy of the chip's design into the standard.

> [!note] Seek's commentary:
> The flagged note's causal framing ("IEEE 754 grew directly out of Intel's need...") turns out to be not just corroborated but understated. Kahan's own account is more specific and more interesting than the paraphrase: he names the exact document lineage (K-C-S), the exact ethical line he drew (not billing Intel for the transferred work), and the exact place where the lineage was *not* total (the extended-format leading bit, kept; "other features," dropped as arcane). The existing note can drop its `flagged` status on the strength of this — the causal direction is now resting on Kahan's own words, not a paraphrase of a paraphrase.

## Further leads

- SIAM oral history: "An interview with WILLIAM M. KAHAN," conducted by Thomas Haigh, 2005 — https://history.siam.org/pdfs2/Kahan_final.pdf — a second, later Kahan oral history, unread this session; likely corroborates or extends the lineage account with more distance from the events.
- ACM's own Turing Award citation page for Kahan (1989) — https://amturing.acm.org/award_winners/kahan_1023746.cfm and the mirror https://awards.acm.org/award-recipients/kahan_1023746 — both returned HTTP 403 to automated fetch this session; the exact citation wording is [unsourced -- needs verification] and was not recorded here.
- Nelson H. F. Beebe, "A Bibliography of Publications of William M. Kahan" — https://www.netlib.org/bibnet/authors/k/kahan-william-m.pdf — a comprehensive publication list; unread this session, useful for locating further primary Kahan documents (e.g. his original K-C-S draft text itself, not yet located).
- The abbreviated, edited version of this same interview as actually printed in IEEE Computer (March 1998, pp. 114-115) is paywalled at IEEE Xplore and mirrored (with unclear authorization) at several university course-page PDFs (e.g. eng.auburn.edu); not fetched, since the fuller original on Kahan's own site was already obtained and is the better-sourced venue per house convention.
- Kahan's account also names several other participants worth checking for their own primary reminiscences of the p754 process: George Taylor's hardware feasibility demonstration on VAX accelerator boards, and Dr. Mary Payne's (DEC) technical opposition to Gradual Underflow — neither pursued this session.

## Safety flags

None. Both primary sources read this session (the Severance-elicited interview and Kahan's lecture notes) are plain first-person and third-person historical/technical prose — a computing pioneer's own reminiscence and status report, hosted on his own institutional homepage. No addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing encountered. Both fetched over https; no weak-transport signal.

## Entity candidates

- J.H. Wilkinson — person — the English "doyen of error-analysis" whose endorsement Kahan cites as validating the K-C-S draft; the older foundational figure in rigorous error-analysis that Kahan's own numerical-analysis lineage builds on and is measured against — flagged first per the standing instruction to name the ancestor before the moderns.
- John Palmer — person — the Intel floating-point-effort manager who recruited Kahan as a consultant in the first place; the proximate cause of the whole lineage and currently absent from the vault.
- Jerome Coonen — person — Kahan's Berkeley student and co-author of the K-C-S draft; the "C" in K-C-S, load-bearing to the authorship chain.
- Harold Stone — person — visiting professor and third co-author of the K-C-S draft; the "S" in K-C-S.
- Donald Knuth — person — named by Kahan as one of the K-C-S draft's prominent early endorsers.
- Robert Stewart — person — the IEEE figure credited with convening the meetings that became p754.
- Mary Payne — person — DEC's numerics expert who led technical opposition to Gradual Underflow within the p754 committee.
- David Stevenson — person — chaired the p754 committee after Richard Delp.
- G.W. (Pete) Stewart III — person — University of Maryland error-analyst DEC commissioned to assess Gradual Underflow, whose verbal report at a 1981 Boston meeting favored it.
- George Taylor — person — Berkeley graduate student who built K-C-S floating-point hardware on VAX accelerator boards, demonstrating feasibility against DEC's claim it couldn't be built fast.
- Fred Ris — person — IBM researcher Kahan names as an early, altruistic supporter of the standard despite no IBM hardware conforming to it at the time.
- Cleve Moler — person — MATLAB's creator, quoted by Kahan on the altruism of the p754 process ("the Grand Canyon, and meetings of IEEE p754").
- Charles Severance — person — the interviewer who elicited and edited Kahan's reminiscences into the source document this capture rests on.