Obsolescence Guaranteed

Most people will mention ENIAC or perhaps the IBM System/360 as examples of early computers. Room-filling machines fed by punch cards, producing payroll runs and scientific calculations. Yes, that was a big part of the story — but it is not our story. The computers we actually use in our daily lives escaped very early on from that path.

Alongside the mainframes, a different strand of computing was taking shape. Where you could use the computer as a tool you could sit down with and interact with. Where you didn't wait hours for a batch job to come back. This is the strand that led to Windows, Linux, Android, the smartphone in your pocket. And to the microcontroller in your dishwasher.

This page is an introduction to that story. It covers the key machines and some of the people. It wants to tell where our computers, and the fundamental ideas behind them, came from.

We admit: this is our perspective. We're Obsolescence Guaranteed, and our goal is to relive these machines hands-on, through the replicas we make of them. We apologise for any oversight or bias. We can't help it, we like our bias.

Prologue: Alan Turing

Before computers, there was the question: what is computation? In 1936, Alan Turing, then a 24-year-old Cambridge mathematician, answered this with a thought experiment. An imaginary machine that could read and write symbols on an infinite tape, following simple rules. The "Universal Turing Machine" could, in principle, perform any calculation that any computer ever could. It was a purely theoretical construct, but defined the boundary of what was possible.

A decade later, Turing was at Bletchley Park, building real machines to break the German Enigma cipher. The Bombes (and later the Colossus) were special-purpose devices, but proved that electronic computing was practical. Turing himself went on to design the ACE (Automatic Computing Engine) at the National Physical Laboratory, one of the first stored-program computers.

Turing's influence on the interactive computing story is fundamental. He showed that a machine could be general-purpose. Not just a calculator, but a tool for thinking.

Turing's ideas were not theoretical. The Pilot ACE, built at the National Physical Laboratory in 1950, was the fastest computer in the world: processing at around 1 MHz, a phenomenal speed for its time. And in 1950, Turing published another seminal paper — "Computing Machinery and Intelligence" — which asked a deceptively simple question: Can machines think? He proposed the Imitation Game, now known as the Turing Test. That question is still unanswered, but interactive computing is what made it worth asking in the first place.

1. Project Whirlwind — The First Interactive Computer

In 1945, the United States Navy asked MIT to build a better flight simulator. They wanted a machine that could model an aircraft in real time, calculating aerodynamic forces fast enough that a pilot could fly it by feel. This was a completely insane request. No computer had ever been built to work at human speed. It became a very successful failure.

Jay Forrester, the young MIT engineer put in charge, did not know this was impossible. He built Whirlwind. Running in 1951, was a revelation. It had a CRT display, you could watch it draw in vectors. It had a light pen, switches and buttons for direct input: the first computer designed to interact with a human being in real time. Every other computer of the era ran batch jobs: hand over your stack of punch cards and come back later. Whirlwind sat there and talked back to you.

Along the way, Forrester and his team invented magnetic core memory. Ferrite rings threaded with wires could hold data (actually, even without power). Core memory remained the dominant random-access memory technology for two decades, and is still used in some spacecraft today. It was born because vacuum tube memory kept failing in Whirlwind, and Forrester needed something reliable.

Whirlwind was a beast: 12,500 vacuum tubes, consuming 150 kW of power, yet managing only 20,000 operations per second. But within the project, something smaller and more significant was taking shape. The Memory Test Computer (MTC) was built to test Forrester's new core memory — and one of the engineers working on it was a young MIT graduate named Ken Olsen. The MTC taught Olsen everything he would later need to build DEC.

Whirlwind's child, the SAGE air defense system, pushed interactive computing even further. SAGE operators sat in front of large CRT displays and used light guns to point at aircraft blips — a direct ancestor of the mouse and every touchscreen that followed.

Whirlwind never succeeded flight simulator role; it was redirected to airspace defence, listening to rader stations and directing interceptor planes. That led directly to the SAGE air defense system. More importantly, it created a generation of engineers at MIT's Lincoln Laboratory who knew, from experience, that computers could do something far more interesting than batch processing.

2. TX-0 — The Transistor Arrives

Whirlwind proved that interactive computing was possible, but filled a largish building and consumed 150 kW of power. The next step was to make it smaller, cheaper, and more accessible. That meant replacing vacuum tubes with transistors.

The TX-0 ("Transistorized Experimental computer zero," affectionately called "tixo") was built at Lincoln Lab between 1955 and 1956. It was the first fully transistorized computer, using around 3,600 Philco surface-barrier transistors. It had a then-huge 64K of 18-bit words of core memory — and, crucially, it had a console. You could sit down at the TX-0 and interact with it directly.

This was the machine that MIT's Tech Model Railroad Club discovered. The TMRC members were already used to building complex signaling systems for their model trains, and when they got access to the TX-0, they found a new kind of playground. They wrote programs, experimented with the display, and pushed the machine in directions its designers had never imagined. They were the first "hackers" — the word already existed in the TMRC to describe an elaborate, creative prank or project — and they brought that ethos to computing.

The word "hacker" did not originate with computers. In the TMRC, a "hack" was an inventive, often playful solution to a technical challenge — wiring a complex signaling system for model trains was a hack. When the club members transferred that mindset to the TX-0, the word found its true home. Hacker still carries that original meaning: not a vandal, but a creative problem-solver who loves the machine.

In a way, the TX-0 was never meant to be more than a stepping stone. It was built as a test platform for the much larger TX-2, a 36-bit machine designed by Wesley Clark. When MIT got its first PDP-1 in 1960, the TX-0 was quietly retired — eventually given away to a museum. Its legacy was not the machine itself, but the generation of engineers and programmers it trained.

Ken Olsen, an MIT engineer who had worked on Whirlwind's memory test computer, directed the construction of the TX-0. He saw where this was going: computers could be personal tools.

3. Digital Equipment Corporation

In 1957, Olsen and Harlan Anderson, both from MIT's Lincoln Lab, decided to start a company. They had $70,000 in venture capital from Georges Doriot's American Research and Development Corporation; one of the first venture capital deals in history. They set up shop in an old nineteenth-century mill building at 146 Main Street in Maynard, Massachusetts.

The company name was a careful choice. In 1957, "computer" was what IBM made — giant, intimidating machines for corporate data centers. No venture capitalist would fund another computer company. So Olsen and Anderson called it Digital Equipment Corporation: "Digital" sounded safe, "Equipment" sounded industrial, and "Computer" was not in the name.

Their first products were logic modules for testing other systems. But Olsen had something bigger in mind. The TX-0 had been followed by the TX-2, a larger 36-bit machine designed by Wesley Clark. Olsen took the lessons of the TX-0 and TX-2 and asked: what if we could make a machine that brought interactive computing out of the lab and into the world?

Olsen would go on to become one of computing's most quotable figures. His most famous line — "There is no reason for any individual to have a computer in his home" (1977) — is almost always taken out of context. He was referring to computers programmed to control household appliances, not the personal computers that were just emerging. Olsen himself had a PDP-11 at home. The remark haunted him for the rest of his career, but it reflects how even the most visionary leaders struggled to predict where interactive computing would go next.

4. PDP-1 — The One That Started It All

In 1959, DEC announced the PDP-1 — the Programmed Data Processor-1. It cost $120,000, about a tenth of the price of an IBM mainframe. It had 4,096 words of 18-bit core memory, 27 instructions, and a central processor built from about 2,700 transistors. Compared to the room-filling IBM machines of the era, it looked like a toy. Only 54 were ever built, but those 54 machines really did change the world.

The PDP-1 had something no IBM machine had: a Type 30 display. It was a converted radar tube. The display took bits directly from the processor's I/O register as X/Y coordinates and drew them on the phosphor screen. It was the simplest possible graphics subsystem. It required no graphics memory, but that meant images had to be redrawn constantly. Still, it was exactly what interactive computing needed.

Spacewar!

In 1962, a group of MIT students — Steve Russell, Martin Graetz, and Wayne Wiitanen — programmed a game on the PDP-1. Two spaceships, controlled by separate sets of switches, fired torpedoes at each other while orbiting a star. They called it Spacewar! It was the first digital video game.

Spacewar! had gravity, thrust mechanics, and a star that would pull you in and destroy you if you weren't careful. It also had the first ever video game controller: a pair of custom switch boxes built by a fellow student, Alan Kotok, who got the parts list from DEC itself. The game was distributed for free, and every PDP-1 site eventually had a copy. It was the first viral piece of software.

The PDP-1 was also the first computer to perform music in public. At MIT's 1963 Science Open House, the Harmony Music Compiler played melodies through a radio speaker wired to the machine's output register. It was an early beginning of computer music — and proof that even with 4K of memory, you could make art.

Where "Hacker" Came From

The MIT PDP-1 community — many of them from the TMRC — took the word "hacker" from the model railroad club and applied it to programming. A hack was an elegant, inventive solution. The PDP-1 was the perfect machine for hacks: simple enough to understand entirely, interactive enough to give instant feedback, and limited enough to demand creativity.

At a time when IBM programmers wrote code, submitted it to a keypunch operator, and got results a day later, PDP-1 hackers sat at the console, toggled in instructions, and watched the results right now. This immediacy was revolutionary. It changed what people thought computers could be.

Firsts Too Numerous to List

The PDP-1 spawned an astonishing number of software firsts. The first text editors. Almost-the-first interactive debugger (DDT, the Dynamic Debugging Technique). The first time-sharing operating systems, allowing multiple users to work simultaneously. Peter Deutsch, a high school student, ported Lisp to the PDP-1 and along the way invented the read-eval-print loop — the REPL that is now the foundation of Python, Ruby, and every interactive programming environment. But the PDP-1, first and foremost, is the birthplace of democoding and gaming:

5. PDP-8 — Glorious Minimalism

The PDP-1 had proven that interactive computing was commercially viable. But DEC realized that to reach a broader market, they needed a machine cheap enough for a single lab or factory floor. So they stripped the PDP-1 concept down to its bare essence.

The PDP-8, introduced in 1965, was the PDP-1's 18-bit architecture shrunk to 12 bits. The instruction set went from 27 to just 8. Memory was 4,096 12-bit words, and the programmer could only directly address 128 words at a time — everything else required page-switching tricks. It cost under $20,000, making it the first computer that an individual department could buy without corporate approval.

DEC thought the PDP-8 would be an industrial controller. Open the valve when pressure gets too high. Stop the heater when temperature rises. They had no idea they were about to invent the personal computer.

DEC's chief engineer, Gordon Bell, called the PDP-8 "the Model T of computing" — and the comparison is apt. Just as Ford's assembly line had made cars affordable, the PDP-8 made computers accessible. And just as the Model T spawned an entire industry of parts and services, the PDP-8 created the minicomputer ecosystem.

The PDP-8's OS/8 operating system was the direct ancestor of CP/M, which in turn was the direct ancestor of MS-DOS. The software DNA of every PC traces back through this lineage. And OS/8 was running on a machine that fit in a rack cabinet, with 4K of memory, in 1968.

Remarkably, the PDP-8 also ran TSS/8, a time-sharing operating system that gave four simultaneous users access to the machine. On a computer with 8 instructions and 4K words of memory. If that sounds impossible, well — that is precisely the point. The PDP-8 forced programmers to be geniuses because it gave them nothing else to work with.

Over 50,000 PDP-8s were sold during its lifetime. It is the prototype/forefather of the personal computer, and of microcontrollers as well.

6. PDP-10 — The mainframe that was actually fun to use

While the PDP-8 was conquering the low end, DEC was working on something entirely different at the high end. The PDP-6 of 1964 had introduced a 36-bit architecture with a rich instruction set that now feels completely alien to modern programmers. But the PDP-6 was unreliable. When DEC got it right with the PDP-10 (KA10, 1968), the result was unlike anything else in computing.

The PDP-10 was a mainframe — it was big, it was expensive (over $100,000), and it was powerful. But unlike other mainframes of the era, it was designed for interactive use. No batch processing. You sat at a terminal and the machine answered you. That was what made the PDP-10 special.

The MIT AI Lab and ITS

At MIT's Artificial Intelligence Laboratory, the PDP-10 ran an operating system called ITS — the Incompatible Timesharing System. ITS was a hacker's paradise: no passwords, no security. But anyone on the early ARPANET could log in. And they did.

ITS was where Emacs was born (Richard Stallman, 1976). Where MacLisp grew up, pioneering much of modern programming language design. Where Shrdlu — the first AI program that could understand a three-dimensional world and obey English commands — dazzled visitors. Where Mazewar became the first networked multiplayer first-person shooter, connecting PDP-10s across the emerging ARPANET.

It is not an exaggeration to say that modern computer science was forged on the PDP-10. The free software movement began here, with Stallman's reaction to the commercialization of the hacker community. PDP-10s dominated the early ARPANET. And by way of trivia, Bill Gates and Paul Allen wrote the first Microsoft BASIC on a Harvard PDP-10, using an 8080 emulator they had written themselves. Open Source and Microsoft were born on the same system!

The PDP-10 also witnessed the very first computer worm. In the early 1970s, a program called Creeper crawled across the ARPANET, displaying "I'm the creeper, catch me if you can!" on connected terminals. Another program, Reaper, was unleashed to hunt it down.

Only about 1,500 PDP-10s were ever sold, but their cultural impact is immeasurable. DEC cancelled the line in 1983, killed internally by the success of its own VAX architecture. But the spirit of the PDP-10 lives on in hacker culture.

7. PDP-11 — Defining Normal

In 1970, DEC launched a machine that would change everything. The PDP-11 was a radical break. It was a 16-bit machine with an orthogonal instruction set that made it far easier to program than anything before it. It had eight general-purpose registers (later models had more). It introduced the Unibus, a standardized bus that made adding peripherals simple. And the design scaled up from a tiny 4K embedded controller to a system with 8 megabytes that could rival mainframes — a scale factor of 2,000.

The design was so elegant that every major microprocessor architecture that came after — the Intel x86, the Motorola 68000 — was visibly shaped by it. The 68000's register set, the x86's addressing modes, the very idea of what a CPU should look like: the PDP-11 defined it.

Where Unix Grew Up

In 1970, Ken Thompson and Dennis Ritchie at Bell Labs ported their experimental operating system — Unix — to the PDP-11/20. It was the first machine Unix ran on as a named operating system. Over the next decade, Unix and the PDP-11 evolved together. The C programming language was designed specifically to be efficient on the PDP-11's architecture. The abstract machine that C targets, to this day, is a PDP-11 in its fundamental assumptions.

The PDP-11's influence is even visible in the syntax of C itself. The close relationship between arrays and pointers in C — the fact that array[i] is equivalent to *(array + i) — is not an accident of language design. It reflects the PDP-11's architecture, where memory addressing was uniform and efficient. C was not designed for an abstract machine; it was designed for the PDP-11.

The Unix source code for the PDP-11 contains one of the most famous comments in programming history. In the I/O driver for the 11/20, Ken Thompson wrote: "You are not expected to understand this". It was a joke about the complexity of the interrupt system, but it became legendary.

When you open a terminal on a modern Linux or macOS system, you are in a direct line of descent from a programmer sitting at a PDP-11 in 1971. The file system hierarchy, the pipe operator, the shell itself — all of it was invented on the PDP-11.

The Road Not Taken — RSX-11

But Unix was not the only operating system on the PDP-11, and in fact it was DEC's least favorite. The company's own RSX-11M was a groundbreaking multi-user, real-time OS that feels utterly different from Unix. But its integrated development environment, its seamless multitasking, its approach to resource management — RSX was an alternative vision of computing that could have won. Easily.

Here is where the story takes a strange turn. When Microsoft decided to break with the MS-DOS past in the late 1980s, Dave Cutler, the lead architect of RSX-11M (and later of DEC's VMS), was tasked with building a new operating system from scratch. That became Windows NT. Its kernel design, its I/O model, and its structured exception handling all bear the marks of the PDP-11 RSX heritage. So: the other half of the modern desktop, the non-Unix half, also traces back to the PDP-11.

Over 600,000 PDP-11s were sold. They were used in laboratories, factories, universities, telephone exchanges, and spacecraft. The top-of-the-line PDP-11/70 pushed performance further with a second MASSBUS interface, capable of transferring data at 4 megabytes per second — breathtaking speed in 1975. The architecture remained in production for over 25 years. When you use any modern computer — PC, Mac, or Linux workstation — you are running on decisions made by the PDP-11's designers in the late 1960s.

8. The ARPANET — Where It All Connected

By the late 1960s, the problem was not computing power — it was isolation. Research labs across the United States had powerful interactive computers, but each was an island. If you worked on a PDP-10 at MIT's AI Lab and wanted to use a program running on a machine at Stanford, you got in your car.

The answer came from the Advanced Research Projects Agency (ARPA), a Pentagon agency that had been funding much of this computer science research. In 1968, ARPA issued a request for proposals to build an experimental packet-switched network. The contract went to Bolt, Beranek and Newman (BBN), a Cambridge, Massachusetts firm founded by two MIT professors. BBN had deep DEC expertise — they owned several PDP-1s and PDP-10s, and knew exactly how to make computers talk to each other.

The device that made it work was the IMP — the Interface Message Processor. It was a rugged minicomputer (initially a Honeywell DDP-516) that sat between a site's main computer and the telephone lines. Its job was to receive messages from the local machine, chop them into packets, and route them across the network to the destination. The IMP was the first purpose-built network router, and its software was written by a BBN team that included several former MIT hackers.

The first four ARPANET nodes went live in late 1969: UCLA, Stanford Research Institute, UC Santa Barbara, and the University of Utah. By 1971 there were fifteen. MIT joined early, and its PDP-10 running ITS became one of the network's most popular destinations — the magical place where the hacker ethos met the network.

In 1971, BBN engineer Ray Tomlinson invented email. He chose the @ symbol to separate the user's name from the machine's name because it was already on the keyboard and "made sense." Email would become the early ARPANET's killer app — and the @ sign is still with us.

1972: The Demo That Changed Everything

In October 1972, ARPA organized a public demonstration at the International Conference on Computer Communication in Washington, D.C. For three days, attendees could sit at one of 40 terminals connected to computers at 10 sites across the country — clattering Teletypes and sleek new CRT terminals — to connect with computers across the country. They could log into MIT's PDP-10, run programs on the Stanford AI Lab's machine, send messages to BBN's PDP-10 in Cambridge. The network worked. And people saw it work.

This was the moment the internet went from research project to an inevitability. The terminals at the 1972 demo were connected through IMPs that routed their traffic through leased telephone lines, entirely automatically. The technology was self-organizing — an IMP would announce its presence, learn about its neighbors, and within seconds the entire network knew about it. That design principle, refined over decades, is still how the internet works today.

The ARPANET grew rapidly through the 1970s. PDP-10s were its early backbone — the machines that ran the early network services, hosted the first email servers, and formed the kernel of what would become the internet. By 1977, the network had evolved into a web spanning the continent.

Currently, we're building a replica of the IMP as part of a larger ARPANET Reconstruction Project, aiming to recreate the 1972-73 network to surf on. It is the final piece of the puzzle, linking all the machines in this story together.