Draft Corrections: Post 26 - The Machine That Looked Like Furniture
Draft Corrections: Post 26 - The Machine That Looked Like Furniture
Post: /_posts/2026-04-27-The-machine-that-looked-like-furniture.md
Date: 2026-05-17
Verification basis: research/VERIFIED_FACTS_Post26_Cray1.md
This document lists ALL recommended corrections to the post, ordered by severity. Apply only the high/medium severity items; low-severity items are suggested for review.
HIGH SEVERITY — factual errors that mislead readers
1. Charles E. Anderson was NOT a Tuskegee Airman
Location: Section “Warren Washington and the Climate Code”, paragraph about Anderson
Current text:
The first was Charles E. Anderson (1919-1994), a former Tuskegee Airman, whom Washington in later interviews credited as his role model.
Problem: Anderson was a weather officer for the 332nd Fighter Group (the Tuskegee Airmen) in the US Army Air Forces. He was not himself a Tuskegee Airman — that term denotes the pilots and air crew, not support officers. He earned his PhD in meteorology from MIT in 1960.
Recommended replacement:
The first was Charles E. Anderson (1919-1994), the U.S. Army Air Forces weather officer who served with the 332nd Fighter Group at Tuskegee during the Second World War and who took his PhD in meteorology from MIT in 1960. Washington in later interviews credited him as a role model.
2. Warren Washington’s master’s degree was in general science, not meteorology
Location: Section “Warren Washington and the Climate Code”
Current text:
He took a bachelor’s degree in physics at Oregon State University in 1958, a master’s degree in meteorology there in 1960, and a doctoral degree in meteorology at Pennsylvania State University in 1964.
Recommended replacement:
He took a bachelor’s degree in physics at Oregon State University in 1958, a master’s degree in general science there in 1960, and a doctoral degree in meteorology at Pennsylvania State University in 1964.
3. Warren Washington was elected to the NAE in 2009, not 2002
Location: Section “Warren Washington and the Climate Code”, final paragraph
Current text:
He was elected to the National Academy of Engineering in 2002. He chaired the National Science Board from 2002 to 2006.
Recommended replacement:
He chaired the National Science Board from 2002 to 2006. He was elected to the National Academy of Engineering in 2009 for pioneering the development of coupled climate models and their use on parallel supercomputing architectures.
4. ECMWF founding member states list is incorrect
Location: Section “Reading on October 24, 1978”
Current text:
The European Centre for Medium-Range Weather Forecasts had been chartered in 1975 by a convention signed by ten European member states – Belgium, Denmark, Germany, Spain, France, Greece, Ireland, Italy, the Netherlands, the United Kingdom, plus Switzerland, Sweden, Yugoslavia, Finland, and Austria as associate members
Problem: The 1975 ECMWF Convention had eighteen founding member states, all on equal footing — there was no “associate” tier at founding. The post omits four founders (Luxembourg, Norway, Portugal, Turkey) and incorrectly demotes five full founders to “associate” status. The Convention was signed in 1973 and came into force on 1 November 1975.
Recommended replacement:
The European Centre for Medium-Range Weather Forecasts had been established by a Convention signed in 1973 that came into force on 1 November 1975, with eighteen founding member states: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, and the United Kingdom. Yugoslavia was also a signatory of the original Convention.
5. ECMWF operational forecasting timeline mischaracterised
Location: Section “Reading on October 24, 1978”
Current text:
Operational forecasting on an experimental basis began on August 1, 1979: ten-day forecasts, five days a week, at 210-kilometre horizontal resolution with fifteen vertical levels, using the so-called N48 grid-point primitive-equation model. A single ten-day forecast required about five hours of CPU time on the Cray-1A. Full operational seven-day-a-week forecasting began on August 1, 1980.
Problem: ECMWF’s own primary sources are clear that August 1, 1979 was the start of full operational medium-range forecasting (not “experimental”). The transition on August 1, 1980 was from five days per week to seven days per week — not “experimental to operational.”
Recommended replacement:
Operational medium-range forecasting began on August 1, 1979: ten-day forecasts, five days a week, at 210-kilometre horizontal resolution with fifteen vertical levels, using the so-called N48 grid-point primitive-equation model. A single ten-day forecast required about five hours of CPU time on the Cray-1A. The service was extended from five to seven days per week on August 1, 1980.
MEDIUM SEVERITY — numerical or chronological discrepancies
6. National Medal of Science date
Location: Section “Warren Washington and the Climate Code”
Current text:
He received the National Medal of Science from President Obama in 2009 and the Tyler Prize for Environmental Achievement in 2019.
Problem: The medal was for 2009 but was presented on November 17, 2010 in the East Room of the White House. “In 2009” is at best ambiguous and at worst suggests a 2009 ceremony.
Recommended replacement:
He received the National Medal of Science (for 2009, presented by President Obama in November 2010) and the Tyler Prize for Environmental Achievement in 2019.
7. Cray-2 production numbers and dominance period
Location: Section “What Came Next”
Current text:
It ran at 1.9 gigaflops peak, was the world’s fastest computer from 1985 to 1987, and shipped about thirty units.
Problem: Wikipedia and Cray records give 25 units shipped (not 30), and the Cray-2 held the “world’s fastest” title until being surpassed by the Cray Y-MP in 1988 (not 1987).
Recommended replacement:
It ran at 1.9 gigaflops peak, was the world’s fastest computer from 1985 until surpassed by the Cray Y-MP in 1988, and shipped about twenty-five units.
8. Cray Computer Corporation spin-off date
Location: Section “What Came Next”
Current text:
In 1989 Cray spun off his next project, the Cray-3, into a separate company called Cray Computer Corporation, which moved to Colorado Springs.
Recommended replacement:
In November 1988 Cray spun off his next project, the Cray-3, into a separate company called Cray Computer Corporation, which moved to Colorado Springs.
LOW SEVERITY — minor issues, suggestions for review
9. Mesa Lab acreage
Location: Section “The Mesa Laboratory”
Current text:
The Colorado state legislature appropriated 250 000 dollars to buy 500 acres of mesa-top land south of Boulder
Note: Wikipedia gives 565 acres for the Mesa Lab site. The $250,000 figure could not be verified from retrieved sources. The post should either confirm “500 acres” from a documented source or change to “approximately 500 acres” or “565 acres”.
10. Cray-1 peak performance phrasing
Location: Section “The Vector Idea”
Current text:
…with multiple chained units running together, for a peak performance of around 160 million operations per second.
Note: 160 MFLOPS is the standard textbook peak figure (Wikipedia). Russell 1978’s primary-source benchmark data gives 138 MFLOPS sustained and 250 MFLOPS short-burst rates. The post’s “around 160” is fine as a representative peak number but does not match Russell’s actual numbers. Optional: cite both numbers or note that “around 160 MFLOPS” is the theoretical peak whereas Russell’s measured short bursts went up to 250 MFLOPS.
11. Power consumption breakdown
Location: Section “The Shape”
Current text:
the total power draw of a Cray-1 at full operation was around 230 kilowatts, half of which was the computation and half of which was the cooling.
Note: Russell 1978 cites 115 kW for the machine. The 230 kW total (115 computation + 115 cooling) figure is plausible but not directly confirmed in retrieved primary sources. If retained, consider adding a footnote.
12. ERA 1103 customer claims
Location: Section “The Hermit of Chippewa Falls”
Current text:
a vacuum-tube system shipped to twenty installations including the Lawrence Livermore Laboratory and the Bureau of Ships.
Note: Cannot verify “twenty installations” or specifically Livermore/Bureau of Ships from retrieved sources. Wikipedia’s UNIVAC 1103 article does not enumerate twenty customers. This may be correct from elsewhere — consider citing the specific source.
VERIFIED POSITIVE — facts that hold up
The following load-bearing claims in the post are verified by primary or strong secondary sources and should be retained:
- Cray-1 announcement 1975, serial #1 Los Alamos March 1976 trial
- NCAR serial #3 arrival July 11, 1977, $8.86 million ($7.9M + $1M disks)
- NCAR decommissioning January 27, 1989
- ECMWF Cray-1A installation October 24, 1978
- 12.5 ns clock, 8 vector registers × 64 elements × 64-bit, 16 memory banks, 1662 modules in 113 varieties, ~200,000 ECL gates, 115 kW
- 80 Cray-1s total sold (the post’s “eighty Cray-1s sold over the production run” is correct)
- Russell 1978 Communications of the ACM paper as primary technical source
- “World’s most expensive love-seat” coined by Steve Callahan at Auerbach (Russell 1978 cites this directly)
- Cray biography: Born Sept 28, 1925 Chippewa Falls; ERA 1950; ERA 1103 in 1953; CDC co-founded 1957 with Norris; left CDC 1972 after 8600 cancellation; Norris invested $250,000; stepped down as CEO 1980; died Oct 5, 1996 from Sept 22, 1996 I-25 accident near AFA
- I.M. Pei selected 1961, age 44; Mesa Lab construction April 1964, completed 1966, dedicated 1967; modeled on Anasazi cliff dwellings of Mesa Verde
- Walter Orr Roberts NCAR founding director
- Wiin-Nielsen first ECMWF Director Jan 1, 1974 - Dec 31, 1979
- Bengtsson Director-General 1982-1990 (Head of Research 1975-1981)
- ECMWF first forecast Aug 1, 1979 at 210 km / 15 levels / 5 hours per 10-day forecast on Cray-1A
- Washington: born Aug 28, 1936 Portland OR, Pullman porter father, BS Physics Oregon State 1958, PhD Penn State 1964, second African-American PhD in meteorology, NCAR 1963, NSB chair 2002-06, Tyler Prize 2019, died Oct 18, 2024 age 88
- CDC 6600 fastest 1964-1969; 34-person team; CDC 7600 fastest 1969-1975 with daily failures; CDC STAR-100 delivered 1974 with poor scalar performance
- Cray-2: 1985 release, 4 processors, 256 MWord, 1.9 GFLOPS, Fluorinert cooling
- Cray-3 / Graywolf S5 delivered NCAR May 1993; gallium arsenide; Cray Computer Corp bankruptcy March 1995