Aberdeen Comes First

The deeper U.S. computing lineage starts in Maryland, in 1918, with a private named Norbert Wiener calculating artillery trajectories by hand — and runs in an unbroken institutional line through the Bush Differential Analyzer in 1935, a backlog of firing tables that nearly broke the Ordnance Department by 1942, the U.S. Army contract that built ENIAC, and a supercomputer dedicated in 2023 to the woman who programmed it.

By Stephen "Pseudo Publius"  |  7 May 2026

BLUF — 

The U.S. Army was using computers at Aberdeen Proving Ground to generate ballistic firing tables before any other major American computational operation is correct, and it is the most direct line of institutional descent in the history of computing. The Ballistic Research Laboratory (BRL) at Aberdeen, formally established in 1938 with antecedents dating to World War I, was the first major American user of large-scale numerical computation; it was the customer that drove the Bush Differential Analyzer's 1935 installation; it was the customer that hired and trained an estimated 100+ women human computers between 1941 and 1945; and — most importantly — it was the customer that wrote and signed the Army Ordnance contract on 9 April 1943 that funded the construction of ENIAC at the Moore School. ENIAC was, from inception, an Aberdeen procurement. It was moved to Aberdeen in January 1947 in fulfillment of the original contract terms and ran there until 1955. The same site, now operating as the U.S. Army Combat Capabilities Development Command Army Research Laboratory (DEVCOM ARL), dedicated six new supercomputers in September 2023 named after the ENIAC Six — the Aberdeen-trained women who programmed ENIAC in 1945. The lineage is unbroken across 105 years and remains operational today.^1,2,3

Wiener at Aberdeen, 1918

The story properly begins with a 24-year-old mathematician who had failed his Army physical, been rejected for officer training, and was eventually drafted as a private. Norbert Wiener — already a Harvard PhD, already a published philosopher, already on his way to becoming the founder of cybernetics — spent 1918 at Aberdeen Proving Ground calculating artillery firing tables by hand alongside other Ordnance Department mathematicians. The work was tedious. The arithmetic was unforgiving. The tables produced there were used by U.S. Army gunners on the Western Front to range their artillery against German positions in the final months of World War I.⁴

Aberdeen had been chosen as the U.S. Army's primary ordnance test facility in 1917 — selected over Sandy Hook, New Jersey, when the older proving ground became inadequate to the longer ranges and heavier shells of modern artillery. The 70,000-acre site on the Chesapeake Bay's western shore in Harford County, Maryland, gave the Army the over-water firing range it needed. From 1917 onward, every American artillery piece, mortar, and bomb of any consequence was tested at Aberdeen. The mathematics of getting the round to land where the gunner intended — exterior ballistics — was Aberdeen's defining intellectual problem.²

Wiener's 1918 service at Aberdeen had two long consequences. The first was personal: his friend and Aberdeen-acquaintance Vannevar Bush, who would become director of the Office of Scientific Research and Development during World War II, would remember Wiener's ballistic-computation experience and assign him in 1940 to the antiaircraft fire-control project at MIT — the work that became foundational to control theory and to cybernetics. The second was institutional: Aberdeen, having seen what Wiener and his colleagues could do with pencil and paper in 1918, knew exactly what the limits of human computation were. When the demand for firing tables outran supply in the 1930s, Aberdeen knew it had a quantitative problem, not a qualitative one. It needed faster computers. The question was what kind.⁴

The firing-table problem, by the numbers

A firing table is a set of values — for a given gun, a given shell, a given propellant — that tells the gunner what elevation and azimuth to set for a desired range, accounting for ambient conditions. The variables are non-trivial. The shell's weight and shape determine its drag profile. The propellant charge determines its muzzle velocity. The barrel's wear and the gun's emplacement angle affect both. Air temperature, humidity, and density alter the drag coefficient. Wind speed and direction shift the trajectory laterally. The hardness of the ground under the gun affects recoil and therefore aim. Each combination of these variables produces a distinct trajectory. Each trajectory must be computed by integrating the equations of motion under drag — a coupled nonlinear ordinary differential equation system that has no closed-form solution.^5,6

By the standard of the late 1930s, computing a single 60-second trajectory entry — one row in one table — required approximately 20 hours of work by a human computer using a Marchant or Frieden electromechanical desk calculator. A complete firing table for one gun-shell-propellant combination contained hundreds of entries. The U.S. Army, by 1942, had hundreds of distinct gun-shell combinations in service. Each new theater introduced new ammunition; each new ammunition required new tables. Each minor design change to a barrel or a shell required revisions. By the spring of 1942 the BRL was producing six firing tables per week and falling steadily behind a demand that ran into the tens of thousands of pages per year.^3,6

This is what the textbooks mean by "the wartime computing crisis." It was not a crisis of theory or of physics. The exterior-ballistics equations had been understood since the early 19th century. It was a crisis of throughput. The Army had run out of arithmetic.

The Bush Differential Analyzer at Aberdeen, 1935

The first attempt at mechanization was analog rather than digital. Vannevar Bush had built a mechanical differential analyzer at MIT in the late 1920s — a room-sized arrangement of wheel-and-disk integrators connected by gears and shafts and driven by electric motors. The machine could solve differential equations of up to eighteen independent variables by setting up the equation as a physical analog: the rotation of one shaft was proportional to one variable, gear ratios performed multiplications, integrator wheels performed integrations, and the output appeared on a plotting table at the end of the chain.⁷

In December 1935, BRL took delivery of an Aberdeen-built copy of Bush's MIT machine. Officially designated the Bush Differential Analyzer, it could compute a 60-second trajectory in approximately 15 minutes — versus the 20 hours required by a human computer with a desk calculator. The improvement was approximately 80×. A second Bush machine, at the Moore School of Electrical Engineering at the University of Pennsylvania, was made available to BRL under a wartime cooperation arrangement, doubling capacity. By the late 1930s these two analog machines were the most advanced ballistic-computation hardware in the United States.^6,7

It was not enough. The Bush analyzer was an analog machine — its accuracy was limited by mechanical precision, its programming required physical reconfiguration of gear trains, and its throughput, while orders of magnitude better than human computation, was still bounded. By 1941, BRL was using the analyzer at full duty cycle and falling further behind every month.

The hundred women

The remedy was hiring. The BRL training program for ballistic human computers, which had begun in modest form before the war, expanded sharply after Pearl Harbor. By 1942, the laboratory was actively recruiting women college graduates with mathematical training from across the Northeast — Bryn Mawr, Goucher, Hood, Chestnut Hill, Mount Holyoke, Smith — and putting them through accelerated training in ballistic computation. The figure most often cited is that BRL trained "almost 100" women in the period 1941-1943; Nathan Ensmenger's research suggests the actual peak was higher, with several hundred women working as Aberdeen-affiliated ballistic computers by mid-war when the Moore School satellite operation is included. From 1943 onward, "essentially all" of these computers were women, and so were their immediate supervisors.⁸

When the Women's Army Corps (WAC) was formed in May 1942, those WAC enlistees with mathematical training were trained in Philadelphia and deployed to Aberdeen specifically for ballistic computation. The BRL workforce grew from approximately 65 personnel in 1940 to a peak of approximately 730 by 1945 — a more than tenfold expansion driven primarily by the firing-table problem.²

The work was structured along the same lines that the previous installment in this series described for Los Alamos T-5: pre-printed computing forms with explicit cell references, decomposition of the trajectory integration into discrete arithmetic operations, parallel pipelines of computers each handling one operation, double-checking by the computer who produced each value, and a supervisor's verification at each stage transition. The architecture was not borrowed from Los Alamos; it had been refined at Aberdeen first, and Frankel, Nelson, and Flanders carried it westward when they joined the Manhattan Project in 1943. The BRL is the elder institution.^3,8

April 1943: the contract that bought ENIAC

The decisive document is a contract dated 9 April 1943, signed between the U.S. Army Ordnance Department, acting through BRL at Aberdeen, and the University of Pennsylvania's Moore School of Electrical Engineering. The contract authorized construction of the Electronic Numerical Integrator and Computer — ENIAC — at an initial budget of $61,700. The eventual cost would reach approximately $500,000 (roughly $8 million in 2025 dollars). The contracting officer was Lt. Col. Paul N. Gillon, BRL's officer-in-charge of computations, who had been hunting for a way to "revolutionize methods of calculations" since 1942 and had been listening to John Mauchly's pitch — for "an entirely new type of computational machinery" using vacuum tubes — since the previous fall. Mauchly's August 1942 memorandum, "The Use of High-Speed Vacuum Tube Devices for Calculating," circulated inside BRL before it reached the Moore School's senior leadership. Eckert and Mauchly are correctly credited with the engineering. Gillon and BRL are correctly credited with the procurement.^1,9,10

The contract specified that ENIAC was being built to produce ballistic firing tables for the U.S. Army. That was the requirement. That was the entire requirement. The fact that ENIAC turned out to be a general-purpose programmable computer was a side effect, recognized partway through construction by Eckert, Mauchly, and the Moore School staff, and it was that recognition that opened the door for ENIAC's eventual use in Manhattan Project hydrogen-bomb calculations, atmospheric simulations, and the wider postwar scientific computing program. But the Army did not pay for a general-purpose computer. The Army paid for an electronic firing-table machine. Aberdeen's procurement language is remarkably clear on this point throughout the contract record.^1,9

The Aberdeen connection to ABC

It is worth noting that the same Aberdeen-driven demand for ballistic computation also indirectly funded John Atanasoff's work in Iowa. Atanasoff's collaboration with the Army was less formal than Penn's, but BRL's standing interest in any plausible computational accelerator is the reason John Mauchly visited Atanasoff in Ames in June 1941 — Mauchly was scouting for ideas with Aberdeen's procurement requirements in mind, not pursuing pure research. The 1973 Honeywell v. Sperry Rand ruling that invalidated the ENIAC patent therefore traces back, by a different path, to the same root cause: the Aberdeen firing-table backlog.^6,10

The ENIAC Six

In June 1945, the Army selected six of the best human computers from the BRL Aberdeen and Moore School computing pools for what was officially designated "Project X" — the codename ENIAC operated under during its remaining classified construction phase. The six were:

• Kathleen "Kay" McNulty — Penn graduate, hired at Moore School in June 1942 at a salary of $1,620 per year. Later married John Mauchly. Continued working with ENIAC after its move to Aberdeen.
• Frances "Betty" Snyder (later Holberton) — Aberdeen-trained, later one of the architects of the UNIVAC sort-merge generator and a contributor to the design of FORTRAN and COBOL.
• Marlyn Wescoff (later Meltzer) — Temple graduate, hired in 1942.
• Ruth Lichterman (later Teitelbaum) — Hunter College graduate. Moved to Aberdeen with ENIAC in 1947.
• Betty Jean Jennings (later Bartik) — Northwest Missouri State graduate. Later led the conversion of ENIAC to a stored-program architecture in 1948.
• Frances "Fran" Bilas (later Spence) — Penn graduate, hired in 1942 alongside Kay McNulty. Moved with ENIAC to Aberdeen.

All six were classified as "computers" in their personnel records — meaning, as the previous installment described, that they were people who performed mathematical calculation, the meaning the word had carried in English since 1613. They were promoted to "operators" of ENIAC and tasked with figuring out how to make the new machine actually compute, working from logic diagrams and wiring schematics. There was no programming manual, no programming language, no precedent. They invented programming as a discipline by direct examination of the hardware.^3,10,11

ENIAC was formally dedicated at the Moore School on 15 February 1946. The U.S. Army press release described it as "the first all electronic general purpose computer ever developed" and noted that it had been built "at the request of the Ordnance Department to break a mathematical bottleneck in ballistic research." The six women were not invited to the dedication banquet. Several of them appeared in the demonstration photographs as "models" — the captioned identification given to women in 1940s engineering photography. The photographs show them at the panels of a working computer that they had programmed; the captions identified them by appearance, not by role. Recovery of their actual contribution into the historical record began with Kathy Kleiman's research in the 1990s and the documentary "The Computers" (2014). The U.S. Army formally honored them — posthumously, all six having died — at Aberdeen in September 2023.^3,11,12

ENIAC moves home

The Moore School's research role on ENIAC ended with its dedication. Eckert and Mauchly resigned from Penn on 22 December 1947 in a dispute with the university over patent rights — the same dispute that produced the 1947 ENIAC patent application that would eventually be invalidated by Judge Larson in 1973. They founded the Eckert-Mauchly Computer Corporation, which would produce UNIVAC I in 1951.^1,10

The Army, in keeping with the original 1943 contract, took possession of ENIAC and dismantled it for transport in November 1946. The machine was reassembled at the BRL at Aberdeen Proving Ground in early 1947, where it was operated continuously until being decommissioned on 2 October 1955. Three of the ENIAC Six — Kay McNulty, Ruth Lichterman, and Frances Bilas — moved to Aberdeen with the machine. Kay McNulty married John Mauchly in 1948 and left the field; Lichterman and Bilas continued operating ENIAC at Aberdeen for several more years.^3,10

Once at Aberdeen, ENIAC's workload expanded well beyond its original ballistic-table mission. It ran the first numerical weather simulations under John von Neumann and Jule Charney; it performed the early hydrogen-bomb yield calculations that Edward Teller and Stanislaw Ulam needed; it produced cosmic-ray interaction simulations and Monte Carlo neutron-transport runs. Throughout, it remained — administratively, financially, and physically — a U.S. Army installation. The first general-purpose programmable electronic digital computer in the United States lived its operational life at Aberdeen Proving Ground.^1,10

The 2023 dedication

On 28 September 2023, the U.S. Army Combat Capabilities Development Command Army Research Laboratory (DEVCOM ARL) at Aberdeen Proving Ground held a ribbon-cutting ceremony for six new supercomputers installed at the DoD Supercomputing Resource Center. The five machines installed at that point — with a sixth following in 2024 — were named after the ENIAC Six: Bartik, Holberton, Antonelli, Meltzer, Teitelbaum, and Spence. The systems run modern Department of Defense scientific workloads — hypersonics, signature analysis, electromagnetic warfare, autonomous-systems modeling — at petascale performance levels that would, in absolute computational terms, complete the original BRL firing-table backlog of 1942 in a fraction of one millisecond.^12,13

Cindy Bedell, director of DEVCOM ARL, opened the dedication with a sentence that should be quoted exactly as she said it: "There is no limit to what we can learn, except what our imagination tells us, and how we program those machines. So these ladies were the first programmers."¹²

The institutional continuity from 1918 to 2023 is therefore literal. The same site, under the same Army Ordnance lineage, has been doing computational work — at increasing technological levels, with the same fundamental requirement of arithmetic throughput at scale — for 105 years and counting. Norbert Wiener, calculating trajectories with a pencil and a logarithm table at Aberdeen in 1918. The hundred women BRL trained on Marchants in 1942. The Bush Differential Analyzer in 1935. ENIAC arriving in 1947 and running until 1955. The Cray-1 at BRL in the 1980s. The DEVCOM ARL supercomputers named for the ENIAC Six in 2023. Every successive substrate is faster than the one before it. The institution is the same. The work is the same.

What this changes about the previous pieces

Stephen's correction reframes the priority discussion in two ways. The first is straightforward: Aberdeen, not Los Alamos, is the older U.S. computing institution. The Manhattan Project's Group T-5 inherited the parallel-pipeline hand-computing architecture that BRL had refined from 1935 onward. Frankel and Nelson knew about it because Berkeley physics graduate students of their generation had Aberdeen contacts; Donald Flanders knew about it because the mathematics community talked. Los Alamos was a brilliant, urgent, and historically consequential application of the BRL methodology. It was not the source.^3,8

The second is more interesting. ENIAC's claim to "first general-purpose programmable electronic digital computer" — the qualifier under which ENIAC survives the Larson ruling and the Colossus disclosure intact — is not an Eckert-Mauchly claim. It is an Aberdeen claim. The contract was Aberdeen's. The requirements were Aberdeen's. The computers who programmed it were Aberdeen-trained. The machine moved to Aberdeen at the end of its construction phase and operated there until decommissioning. Eckert and Mauchly built it; the U.S. Army Ordnance Department, headquartered at Aberdeen, owned it. The popular framing that gives Eckert and Mauchly individual priority in the 1973 patent dispute — and that gives Penn institutional priority in the academic computing histories — gets the architecture but loses the procurement. The Army should be in the textbook line, and largely is not.

The line to Memphis

Bringing this back to the larger arc of the series: the Memphis Colossus rented to Anthropic on 6 May 2026 is, in its institutional structure, a private-sector analog to BRL. Both exist because a customer with an essentially unbounded computational requirement needs faster substrate than the previous generation can deliver. BRL needed firing tables for an Army at war on six continents. Anthropic needs inference capacity for a frontier AI model whose user demand is growing 80× annualized. Both responses are the same: contract with the most capable engineering organization available, take delivery as quickly as possible, plan the next iteration before the current one is finished.

The differences are scale and urgency. The 1943 ENIAC contract was for $61,700, eventually $500,000. The 2026 Anthropic deal is reported in the high hundreds of millions of dollars for the Colossus 1 capacity alone, with multi-gigawatt orbital follow-on capacity discussed in the multi-billion range. The Aberdeen lineage produced six machines in 105 years; the Memphis-and-orbital lineage proposes one million orbital nodes in five. The substrate has scaled by ten orders of magnitude. The institutional logic — customer with a computational backlog, contract with the best available engineering, take delivery and demand more — has not changed at all.¹⁴

The honest conclusion of this five-piece series is that the Memphis Colossus inherits a longer and more institutionally specific lineage than its branding has suggested. It is the descendant not just of Tommy Flowers's Bletchley machine and Eckert and Mauchly's Penn machine but of Norbert Wiener at Aberdeen in 1918, of the hundred women BRL trained between 1941 and 1943, of Lt. Col. Paul Gillon's contract on 9 April 1943, of Kay McNulty and Frances Bilas riding the truck from Philadelphia to Aberdeen with ENIAC in November 1946, and of the DEVCOM ARL ribbon-cutting in September 2023. The orbital constellation that the present series began by analyzing — multi-gigawatt, sun-synchronous, optically meshed, named only as "SpaceX Orbital Data Center system" in the FCC filing — will, when it eventually goes operational, be doing exactly the same thing the BRL human computers were doing in 1942. Decompose the problem. Distribute the arithmetic. Accumulate the partial results. Check for errors. Integrate forward. The substrate is the variable. The work is the constant.

---

Sources

1. "ENIAC." Encyclopedia of Greater Philadelphia. https://philadelphiaencyclopedia.org/essays/eniac/  |  Engineering and Technology History Wiki, "ENIAC." https://ethw.org/ENIAC
2. "Ballistic Research Laboratory." Wikipedia, retrieved 7 May 2026. https://en.wikipedia.org/wiki/Ballistic_Research_Laboratory   Documents the BRL's establishment in 1938, growth from ~65 personnel in 1940 to ~730 by 1945, and the 1935 installation of the Bush Differential Analyzer.
3. "Aberdeen Proving Ground dedicates supercomputers to early programmers." Baltimore Sun, 6 October 2023. https://www.baltimoresun.com/2023/10/06/aberdeen-proving-ground-dedicates-supercomputers-to-early-programmers/
4. Conway, F. and Siegelman, J. Dark Hero of the Information Age: In Search of Norbert Wiener, the Father of Cybernetics. Basic Books, 2005. Definitive biography; covers Wiener's 1918 Aberdeen service in detail. See also Rheingold, H. Tools for Thought, MIT Press, 2000, chapter 5. http://www.rheingold.com/texts/tft/05.html
5. Smiley, J. "Proving Ground: A biography and history of the six women who invented programming for ENIAC." EE Journal, 31 August 2023. Reviewing Kathy Kleiman's Proving Ground: The Untold Story of the Six Women Who Programmed the World's First Modern Computer, Grand Central, 2022. https://www.eejournal.com/article/proving-ground-a-biography-and-history-of-the-six-women-who-invented-programming-for-eniac/
6. "The ENIAC Story." U.S. Army Research Laboratory, primary historical document. https://ftp.arl.army.mil/~mike/comphist/eniac-story.html
7. DTIC ADA593830, "The Differential Analyzer." U.S. Army Ballistic Research Laboratory technical report on the Aberdeen installation of the Bush Differential Analyzer. https://archive.org/details/DTIC_ADA593830  |  Isaacson, W. The Innovators. Simon & Schuster, 2014, on the Bush analyzer's spread to BRL, Penn, Manchester, and Cambridge.
8. Ensmenger, N. The Computer Boys Take Over: Computers, Programmers, and the Politics of Technical Expertise. MIT Press, 2010. Excerpt at "Women Were First Computer Programmers," Women's eNews. https://womensenews.org/2012/03/women-were-first-computer-programmers   Documents the BRL training program and the women supervisors who developed "plans of computation."
9. U.S. Army Acquisition Support Center. "Then & Now." Documents Lt. Col. Paul N. Gillon's role and the Mauchly memorandum. https://asc.army.mil/web/news-then-now/
10. "What Happened on April 9th." Computer History Museum / This Day in History. https://www.computerhistory.org/tdih/april/9
11. Light, J.S. "When Computers Were Women." Technology and Culture, vol. 40, no. 3, July 1999, pp. 455-483.
12. "Aberdeen Proving Ground dedicates supercomputers to early programmers." DEVCOM ARL release, 28 September 2023. AOL syndication carries Cindy Bedell's full remarks: https://www.aol.com/aberdeen-proving-ground-dedicates-supercomputers-230300907.html
13. U.S. Department of Defense High Performance Computing Modernization Program. DEVCOM ARL DoD Supercomputing Resource Center installation records, 2023-2024.
14. Anthropic. "Higher usage limits for Claude and a compute deal with SpaceX." 6 May 2026. https://www.anthropic.com/news/higher-limits-spacex