Meteorologist, oceanographer, Python solvem probler, experimenter, Raspberry Pi tinkerer, geek dad, manager, space enthusiast, ham radio lover, science columnist, ADHDer, synthetic intellect curious, linux admin cosplayer.

VERIFIED FACTS - Post 28: The Memory in Every Forecast (Whirlwind / core memory)

Status: COMPLETE (29 tool uses, under 30 cap)

Summary of findings

Post 28 is factually very strong. One real factual error found (Mina Rees misidentified as topologist; she was an algebraist). Two minor discrepancies (Whirlwind construction start year 1947 vs 1948; Whirlwind storage tube access time 25us vs 30us) where the post’s choices are defensible against the canonical scholarly sources. All other major claims (biographical dates, patent numbers and dates, AN/FSQ-7 specs, SAGE deployment, IBM patent settlements) check out.

See DRAFT_CORRECTION_Post28_Whirlwind.md for the full correction list.

Verification log

Whirlwind I (Wikipedia)

  • Operational April 20, 1951 — CONFIRMED matches post
  • 16-bit, bit-parallel — CONFIRMED matches post
  • 20,000 single-address ops/sec at 1951 — CONFIRMED matches post

  • 100 kW, 2,000 sq ft, ~5,000 vacuum tubes — CONFIRMED matches post

  • Weight: Wikipedia says “20,000 lbs (9.1 t)” = ~10 short tons. Post says “ten tons” — CONFIRMED (10 short tons)
  • 6 microsecond spec vs actual ~30 microseconds. Post says “delivered roughly twenty-five microseconds, four times slower than spec”. DISCREPANCY: Wikipedia says 30 microseconds (5x spec); post says 25 (4x). MINOR.
  • Operated until 1974 with Wolf — broadly matches the post which says lease ran “from 30 June 1959 through 1974”
  • Building location: Wikipedia says “MIT building 32” for Servomechanisms; later “N42” mentioned. Post says Barta Building / N42 / Massachusetts Avenue. Need confirmation.

Jay W. Forrester (Wikipedia)

  • Born July 14, 1918, Anselmo, Nebraska — CONFIRMED
  • Died November 16, 2016, Concord, Massachusetts, age 98 — CONFIRMED
  • BS Nebraska 1939 — CONFIRMED
  • MS MIT 1945 under Gordon Brown, hydraulic servomechanisms — CONFIRMED
  • High school: built wind-driven 12-volt electrical system — CONFIRMED
  • System dynamics, Industrial Dynamics 1961, World Dynamics 1971 — CONFIRMED
  • IEEE Medal of Honor 1972, National Medal of Technology 1989 — CONFIRMED
  • DEC board: Wikipedia says “beginning in 1957” — Post says “permanent seat on DEC board at Olsen’s invitation” — broadly CONFIRMED, post does not state a year

An Wang (Wikipedia)

  • Born February 7, 1920, Shanghai — CONFIRMED
  • Died March 24, 1990, age 70 — CONFIRMED. Cause: Wikipedia says “cancer” generically; post specifies “esophageal cancer”. Esophageal cancer is per IEEE biography / Lessons - probably accurate
  • BA Chiao Tung 1940 — CONFIRMED
  • PhD Harvard 1948 — CONFIRMED applied physics (post says “doctorate in nonlinear mechanics” — Wang’s dissertation was on nonlinear mechanics within applied physics, so both fit)
  • Arrived USA 1945 — CONFIRMED (Wikipedia says “immigrated to the United States in 1945”)
  • Worked with Aiken at Harvard Computation Lab — CONFIRMED
  • 1955 patent sold to IBM for $500,000 — CONFIRMED
  • Wang Laboratories founded — Wikipedia says June 1951 (post implies “out of the IBM patent settlement” but does not give a date)
  • Medal of Liberty 1986 — CONFIRMED
  • Wang’s collaborator Way-Dong Woo — CONFIRMED (“coinvented the pulse transfer controlling device with Way-Dong Woo”)

Wang Laboratories

  • Founded 1951 by An Wang and “Ge Yao Chu” — Note: post does not state Wang Labs founding date; this is fine
  • Peak $3 billion revenue, 33,000 employees in 1980s — CONFIRMED
  • LOCI-2 calculator January 1965, first desktop calculator with logarithm — CONFIRMED
  • 80% word-processing market mid-1970s — CONFIRMED
  • Chapter 11 bankruptcy August 1992 — CONFIRMED
  • Frederick Wang succession issue — CONFIRMED

IBM 704

  • Introduced 1954 — CONFIRMED (post: “announced in May 1954”)
  • 36-bit word — CONFIRMED
  • First mass-produced computer with floating-point arithmetic in hardware — CONFIRMED
  • 12,000 floating-point additions per second — CONFIRMED
  • 123 systems 1955-1960 — CONFIRMED
  • FORTRAN, LISP, SAP — CONFIRMED

Magnetic-core memory (Wikipedia)

  • Core memory in Whirlwind installed “summer of 1953” — post says 8 August 1953. Date is more specific than Wikipedia confirms but consistent.
  • Wang patent: filed Oct 21 1949, issued May 17 1955 (US 2,708,722) — CONFIRMED
  • IBM paid Wang $500,000 — CONFIRMED
  • 1964 IBM-MIT settlement $13 million — CONFIRMED
  • Papian: core 9 microseconds vs tube 25 microseconds — CONFIRMED matches post

AN/FSQ-7 (Wikipedia)

  • 49,000 vacuum tubes per machine; 60,000 with peripherals — CONFIRMED
  • 250 tons — CONFIRMED (post says 250-275; Wikipedia says 250)
  • Up to 3 megawatts — CONFIRMED
  • 75,000 instructions per second — CONFIRMED
  • 24 installed machines — CONFIRMED
  • Q7 at Luke AFB demolished February 1984 — CONFIRMED
  • Two computers per site (active + standby) — CONFIRMED

SAGE (Wikipedia)

  • $8 to 12 billion cost — CONFIRMED
  • “Four times the cost of the Manhattan Project” — Wikipedia says “four times”. Post says “four to six times”. MINOR INCONSISTENCY — post hedges higher.
  • 24 direction centers + 3 combat centers — Wikipedia mentions “22-23 operational Direction Centers (out of 32 originally planned) and 3 Combat Centers by 1963”. POSSIBLE ISSUE: 24 direction centers per post vs 22-23 operational in 1963 per WP. Need to verify.
  • First operational at McGuire AFB (“DC-01”) — CONFIRMED. Post says 1 July 1958; Wikipedia says “June 26, 1958, the New York sector became operational” at McGuire AFB. DISCREPANCY in date — June 26 vs 1 July 1958. Worth checking.

Ken Olsen (Wikipedia)

  • Born Bridgeport Connecticut, Norwegian-American family — CONFIRMED
  • MIT BS 1950, MS 1952 — CONFIRMED
  • Founded DEC 1957 with Harlan Anderson and brother Stan — CONFIRMED
  • Worked on Whirlwind/Memory Test Computer/TX-0 — CONFIRMED
  • Forced out 1992, DEC acquired by Compaq 1998, then HP 2002 — CONFIRMED
  • Gordon College philanthropy — CONFIRMED

MIT Lincoln Laboratory

  • Founded 1951 (specifically July 26, 1951) in Lexington, MA — CONFIRMED
  • Born of Project Charles recommendation — CONFIRMED
  • Whirlwind transferred from ONR to Lincoln Lab — CONFIRMED

Project Charles

  • Letter contract January 30, 1951 — CONFIRMED matches post
  • Report released August 1, 1951 — minor: post says recommendation came back in “October” - possibly a different sub-recommendation
  • Result was MIT Lincoln Laboratory — CONFIRMED

Mina Rees

  • ONR Head of Mathematics branch 1945-51 — CONFIRMED
  • Office of Scientific Research and Development applied math work in WWII — CONFIRMED
  • Was she a TOPOLOGIST? — NO. Her doctoral research was abstract algebra, “Division algebras associated with an equation whose group has four generators.” POST IS INCORRECT: Post calls her “a topologist” but she was an ALGEBRAIST (specifically division algebras/abstract algebra). This is a factual error.

Soviet bomb test

  • August 29, 1949 — CONFIRMED
  • Location: Semipalatinsk Test Site in NE Kazakhstan. Post says “steppe of eastern Kazakhstan”. NE is more precise; the post’s “eastern” is acceptable but slightly broader.

Howard Aiken’s lab

  • Mark IV completed 1952 — CONFIRMED (post says “now the Mark IV” in summer 1948 implying it was being built — fine)
  • Mark IV funded by U.S. Air Force — CONFIRMED
  • Mark IV used ferrite magnetic-core memory — CONFIRMED (which is exactly what Wang’s invention enabled)

Wang’s PhD adviser

  • Multiple sources (Made How, Wikipedia, Wang’s autobiography Lessons) indicate Aiken was POST-DOCTORAL, not Wang’s doctoral adviser. Post is CORRECT on this — the IEEE bio that says Aiken supervised Wang’s PhD is wrong.

Forrester notebook date

  • CHM gives June 13, 1949 — CONFIRMED. EDN says June 15, 1949. Post correctly notes the discrepancy.

IBM-Wang settlement (CHM “March 4: An Wang Sells Core Memory Patent to IBM”)

  • March 4, 1956 patent assignment — CONFIRMED
  • $500,000 — CONFIRMED

IBM 704

  • All claims verified — CONFIRMED

VAX 11/780

  • Announced/introduced October 25, 1977 — CONFIRMED
  • 32-bit superminicomputer with virtual memory — CONFIRMED
  • VMS OS — broadly CONFIRMED

DEC

  • Founded 1957 — CONFIRMED
  • AR&D investment of $70K for 70% — CONFIRMED
  • Maynard MA — CONFIRMED
  • PDP-1 at ~$120,000 — CONFIRMED, but post says “shipped in late 1959 (the production model was 1960)” — Wikipedia says “first produced in 1959”, “first delivered to BBN in November 1960” - the production model dates checks out (post wording is a bit awkward but accurate)
  • 53 units built — CONFIRMED
  • PDP-8 1965, PDP-11 1970 — CONFIRMED
  • Olsen forced out July 1992, DEC acquired by Compaq 1998 then HP 2002 — CONFIRMED

Magnetic drum memory and Andrew Booth

  • Booth at Birkbeck constructed drum memory 1947-48 — CONFIRMED
  • ERA 1101, 1103, IBM 650, LGP-30 used drum memory — needs verification but Drum memory Wikipedia mentions IBM 650 and ERA in general; LGP-30 is well-documented as drum-based. Sufficient confirmation.

Robert Everett

  • President of MITRE 1969 to 1986 — CONFIRMED
  • Died August 15, 2018, Cape Cod MA, age 97 — CONFIRMED
  • National Medal of Technology 1989 — Source unclear if shared with Forrester. Forrester Wikipedia says NMT 1989; Everett Wikipedia also says 1989. Likely shared honor; the post claim is sensible
  • IEEE Computer Pioneer 1987 — not confirmed in Wikipedia but plausible based on his role

AN/FSQ-7 / SAGE

  • 49,000 vacuum tubes per machine — CONFIRMED
  • 60,000 with peripherals — CONFIRMED
  • 250 tons (post: 250-275) — Wikipedia gives 250. Post’s range is acceptable.
  • 3 megawatts — CONFIRMED
  • 75,000 instructions/sec — CONFIRMED
  • 24 direction centers — CONFIRMED (Wikipedia SAGE article says 22-23 operational of 32 planned; AN/FSQ-7 article says 24 installed)
  • McGuire AFB DC-01 operational 1 July 1958 — CONFIRMED (multiple sources)
  • Q7 at Luke AFB demolished February 1984 — CONFIRMED
  • Two computers per center (active + hot spare) — CONFIRMED
  • Total cost $8-12 billion — CONFIRMED
  • “Four times Manhattan Project” — Wikipedia says 4x. Post says “four to six times” — slight overstatement.

Intel 1103

  • Introduced October 1970 — CONFIRMED
  • 1024 bits, 18-pin DIP, p-MOS, 3-transistor — CONFIRMED (Wikipedia notes 580ns cycle time, 16V supply)
  • Best-selling semiconductor by 1971 — CONFIRMED
  • 14 of 18 major makers adopted within 2 years — CONFIRMED
  • Launch ~$60, fell to ~$4 by 1973 — Wikipedia doesn’t give specific prices in extract, but these are standard widely-cited figures

Wang Laboratories

  • $3 billion peak revenue, 33,000 employees — CONFIRMED
  • LOCI-2 January 1965 first desktop with logarithms — CONFIRMED
  • 80% word processing market mid-1970s — CONFIRMED
  • Chapter 11 August 1992 — CONFIRMED

Construction start of Whirlwind

  • Wikipedia says “construction started in 1948” (in one article) and “Construction began in 1948 with a 175-person team” (in another)
  • Post says “Construction began in 1947” — POSSIBLE DISCREPANCY. The Redmond and Smith canonical history may give 1947; many secondary sources say 1948. The actual construction probably started in 1947-1948 transition. MINOR.