notes.husk.org/likes images.

Whirlwind 75th Anniversary (1951 - 2026)

“We’re still living with it. There was only one Whirlwind, but it’s still with us.” — Guy Fedorkow

I don’t think any computing project changed as much between concept and delivery as Whirlwind, making its history somewhat challenging to follow. This is also the reason I’m as captivated with its story as I am, and find it miraculous that it developed the way it did, and impacted every computer that succeeded it.

Whirlwind’s origins go back to 1943, with the Office of Naval Research (ONR). U.S. Navy Captain Luis de Florez pitched the project to his alma mater, MIT, as a universal flight simulator. This description caused me a bit of confusion - for a long time I assumed it was a flight simulator for training pilots (like a link trainer), and while that may have been Florez’ initial concept, Project Whirlwind’s objective was a flight simulator for modeling aircraft behavior.

The Whirlwind Program, Part 1, Volume 1, p. 7 and 1. (Source)

Nevertheless, the task of simulation would be carried out by an analog control system modeled on the link trainer, which had been in use since the late 1920s. By the year 1945, work on Project Whirlwind was underway at MIT’s Servomechanisms Laboratory, overseen by Jay Forrester, Robert Everett and Perry Crawford.

Robert Everett with the original Whirlwind flight simulator. (source)

It soon became apparent there was no analog solution to the engineering problem posed. Having a control system that could realistically respond to the aircraft’s reactions necessitated high speed input/output, something that could only be achieved through the use of an electronic, digital computer, which was out of MIT’s wheelhouse in 1945. (These were the early days of such technologies, and MIT was particularly married to analog methods). Aware of the nascent ENIAC and EDVAC, Forrester sought out information from personnel involved in the project, and wrote to the Pentagon for technical reports relating to EDVAC.

The Whirlwind Program, Part 1, Volume 1, p. 8

The course of the project was re-charted in 1946. The heart of Project Whirlwind would be a digital electronic computer, construction on which began in 1948.

Like ENIAC, Project Whirlwind far outlived the end of the war that had borne it. While ENIAC was kept in use as an artillery computer even in the absence of active combat, Whirlwind was separated from its original purpose, the flight simulator aspects of the project scrapped completely in the late 40s. The ONR continued to fund Whirlwind as a high-speed, general-purpose computer, but grew dissatisfied with delays and a growing price tag amidst budget cuts, and dropped support for the project in 1950.

In 1950, MIT physicist and Radiation Laboratory alum George E. Valley recognized an unmet need for an automated air defense network and identified Whirlwind as the ideal computer for the task. After Forrester demonstrated Whirlwind’s ability to process radar data via telephone line and display results on a cathode ray tube, The U.S. Air Force stepped in with funding in November of that year.

Thus Whirlwind was repurposed as the heart of Project Charles, a prototype air defense network. Air traffic data was transmitted to Whirlwind from a radar installation in Cape Cod. In 1952, Whirlwind was able to successfully auto-pilot a B-26 bomber (cast in the role of the interceptor) to its rendezvous point, serving as proof-of-concept for the SAGE network.

Chronology: From the Cambridge Field Station to the Air Force Geophysics Laboratory, 1945-1985, p. 10 (source)

There are a few implications I want to dig into. The most obvious being the role of computers in air defense & air traffic control. But with its data links to the radar station & B-26, this event also presaged the development of computer networks, possibly the first instance of a computer receiving and transmitting data remotely. The 1958 Bell 101 for the AN/FSQ-7 (SAGE) is cited as the first commercial modem, but Whirlwind was sending and receiving data over phone lines earlier.

This is the first instance of real-time computing. Until Whirlwind, the dominant mode of input/output was punched cards; it was the first computer that was interactive. It was the first computer with a “monitor” in any manner of speaking. I think there’s irony in the primary mode of input being a light gun, a primitive forerunner of the stylus and touchscreen computing that would become ubiquitous about 60 years later. (Whirlwind had a keyboard, again the first computer to do so, but it was originally a numeric pad only.)

The third important point is Whirlwind’s flexibility with what it could connect to. Radar antennae and airplanes were among its devices. So were light guns, CRTs, Flexowriters, radar scopes for the Project Charles experiments, paper tape, magnetic tape, magnetic drums and others. Guy Fedorkow referred to Whirlwind as something like “the first Arduino” (paraphrasing) and that comparison sticks with me. It was prescient to the ways we use computers today.

While Forrester had been experimenting with magnetic core memory since the late 40s, Whirlwind launched with electrostatic tube (CRT) memory, similar to the Williams tube memory used by other computers of the time. However it was expensive and unreliable. By 1953 its tubes had been replaced with a magnetic core memory stack, initially developed and tested on the Memory Test Computer, a small, special-purpose version of Whirlwind developed by Ken Olsen.

Left: Forrester, Youz & Dodd with two of Whirlwind’s CRT memory tubes. (Source) Right: Unidentified technician weaving a pane of Whirlwind’s core memory. (Source)

Here we see two other tributaries between Whirlwind and the wider world of computing. The popularization of core memory, which would become a dominant technology for the next quarter century, and Digital Equipment Corporation, which Ken Olsen would found after exiting the TX-0 and TX-2 computer projects, themselves influenced by Whirlwind.

Given Whirlwind’s influence on the development of both large (IBM) and small (DEC) systems, Robert Everett’s quote that “The millions and billions of computers that are around, and there are more and more all the time, in my opinion, really, descended from Whirlwind” rings true to me.

Whirlwind operated at MIT through the spring of 1959, at which point it was leased by William Wolf, who dismantled and moved it. Back in operation by 1963, it ran at Wolf Research and Development for the next decade, and was finally dismantled in 1976.

Countdown to shutdown, 1959. (Source)

As a physical object, Whirlwind computer survives as mere fragments. But it lives on through software recovered from original paper and magnetic tapes and a hardware simulator (with a PiDP-style physical simulator upcoming). Today, anyone can have their own Whirlwind computer, run original software, and even write new programs.

Left: “Bouncing ball” program, ca. 1949. (Source [PDF]) Right: The same program today on the Whirlwind simulator. (Own image)

This is only a brief introduction to Whirlwind, and I encourage anyone who’s interested to investigate the wealth of available documentation. There’s so much that I couldn’t cover here, including the political intrigue surrounding its development; the story of its first hired operator, Joseph Thompson, an African American high school student; the saga of its afterlife and efforts to preserve its software.

The thing that most stands out to me is that no one ever gave up on Whirlwind. There are many heartbreaking stories in computer history, wasted potential, ideas too early or late for their time, but Whirlwind’s feels like one of the triumphant ones. Its success was never certain, but there was always someone willing to fight for it. Despite immense change those involved with the project maintained a clarity of vision that carried it to completion. And even when it was destroyed, people refused to let it be lost to history. It’s with us not only in the technology we use today, but the memory and imaginations of those who know it.

Opening image descriptions & sources:

Layout of the Whirlwind I digital computer. MIT Museum. (1)
With Dodd, Forrester, Everett and Ramona Ferenz. DECworld (2)
With operator Joe Thompson (seated)
With Jack Gilmore and Joe Thompson
Core memory stack. Bit by Bit blog. (3 - 5)
Vacuum tube closeups. Wikimedia Commons. (6)
Circuitry closeups. Wikimedia Commons. (7)