October 1970: The Month Life Escaped Into the World
Sometime in early October 1970, the new issue of Scientific American lands on a desk at MIT. The magazine is read, set aside, picked back up. Someone turns to page 120. The column is called “Mathematical Games.” The byline is Martin Gardner. The title reads: “The fantastic combinations of John Conway’s new solitaire game ‘Life.’”
Within weeks, the MIT AI Laboratory’s PDP-10 — a machine shared by some of the most talented programmers in the world — is running Life simulations for hours at a stretch. By the end of November, a team led by a mathematician named Bill Gosper has found something that was supposed, by plausible conjecture, to be impossible. Conway is about to owe someone fifty dollars. And a game that began on a Go board in a Cambridge common room has become, with no warning and no plan, the first viral mathematical meme in history.
The Column
Gardner’s column ran four pages — 120 through 123 — in the October 1970 issue of Scientific American. He had written “Mathematical Games” since 1956, and he had a gift for recognizing when something was genuinely different from the usual puzzle or curiosity. Conway’s game was different.
The title tells you something: not “a new game” but “the fantastic combinations.” Gardner understood the emotional register. This was not a recreational diversion. This was something that looked like a toy and behaved like a universe.
Gardner laid out the rules — birth on exactly three neighbors, survival on two or three, death otherwise — and explained why they were interesting. He described the glider, a five-cell wanderer moving diagonally at a quarter the speed of light. He introduced still lifes, oscillators, and the menagerie of patterns that Conway and his Cambridge colleagues had already found on a Go board.
Then came the question that would detonate the next six months of mathematical culture.
Conway, Gardner wrote, conjectured that “no pattern can grow without limit. Put another way, any configuration with a finite number of counters cannot grow beyond a finite upper limit to the number of counters on the field.” This was probably true. It seemed intuitively right. But it was unproven. And Conway was prepared to back his uncertainty with cash: “Conway has offered a prize of fifty dollars to the first person who can prove or disprove the conjecture before the end of the year.”
A configuration that disproved it would have to keep adding counters forever. Gardner named two hypothetical forms: a gun — “a configuration that repeatedly shoots out moving objects such as the glider” — or a puffer train, which moves and leaves stable debris behind. Both seemed unlikely. Neither had been found.
Scientific American in 1970 had a circulation above 450,000, reaching scientists, engineers, professors, and educated generalists across the country and beyond. They all received the same challenge on the same pages. By any measure of what a publishing platform could do in 1970, this was as wide a broadcast as existed.
The Response
The response was unlike anything Gardner had seen in fourteen years of writing “Mathematical Games.” It became the most widely read column he ever wrote, generating more reader mail than any other in the column’s history.
What this looked like, practically: people computed Life by hand. Graduate students in mathematics departments spent evenings tracking populations through dozens of generations on graph paper, looking for patterns that nobody had named. Professors assigned it in courses. And wherever access to a computer existed — which in 1970 meant wherever a university or research lab had a terminal room — people wrote programs.
In 1970, computing time was not cheap or casual. Access to a mainframe was rationed, priorities set, quotas enforced. Running Life was not a scheduled research task. It happened in the evenings, in the margins of legitimate work, on machines that would otherwise sit idle. Life was also genuinely hard to simulate at scale — the neighbor counts computed for every live cell, the new generation written efficiently — and the problem attracted programmers not just because it was fun, but because it was technically demanding in ways that produced transferable insights.
Bill Gosper and the MIT AI Lab
The MIT Artificial Intelligence Laboratory in 1970 ran on a culture of radical openness and obsessive competence. The PDP-10 — running the ITS operating system — was available to anyone with talent and curiosity. No passwords. No security. The assumption was that people with access would use it seriously, which they did.
Bill Gosper was thirty. He had arrived at MIT in 1961, taken a programming course from John McCarthy, and never really left. By 1970 he was a fixture of the AI Lab’s hacker culture — known for mathematical depth, computational inventiveness, and focused intensity. He later co-authored HAKMEM, the legendary AI Lab memo that reads like a compressed encyclopedia of everything a very small number of very talented people had figured out.
When the October Scientific American circulated, Gosper read it and attacked it. The prize was fifty dollars — not much, even in 1970. But the problem was not small. Gosper’s team — which included Mike Beeler, Rich Schroeppel, and other AI Lab regulars — began running Life on the PDP-10 with systematic intensity.
They were looking for something that would produce unbounded growth. Conway had named two candidate forms: a gun and a puffer train. The gun seemed the more tractable target. If a pattern could be found that produced gliders continuously — one every period, cycling back to its original state and firing again — then the population would grow forever, one glider at a time.
The search was not random. It was guided by mathematical intuition about what kinds of symmetry and periodicity could sustain the required self-repair: a pattern that oscillated — returning to a prior state — while simultaneously emitting something that escaped. Stable enough to survive indefinitely, dynamic enough to produce output, structured enough to fire periodically rather than chaotically.
In November 1970, Gosper’s team found it.
The Gosper Glider Gun is a 36-cell configuration built from two “queen bee shuttle” oscillators, stabilized by two blocks. It fires its first glider on generation 15 and then fires another glider every 30 generations thereafter, indefinitely. The pattern returns to its original state every 30 ticks. Each cycle, one glider escapes. The population grows without limit.
Conway’s conjecture was false. Infinite growth was possible from a finite seed. The prize was real.
The Infrastructure of Mathematical Culture in 1970
Step back and consider what happened structurally. A single magazine column created a coordinated global research effort that produced a significant mathematical result in approximately six weeks. No email. No preprint servers. No internet. Just a magazine, the postal system, and shared access to a handful of mainframes.
Scientific American was not merely a popular science magazine — it was, for a specific community, the primary mechanism for broadcasting open problems to the people most capable of solving them. Gardner had built, over fourteen years, a readership that was simultaneously literate and technically skilled. When he posed a problem, the people reading it were not casual hobbyists but mathematicians, computer scientists, and engineers who could work on it.
Conway had framed the challenge precisely: a specific conjecture, a specific kind of counterexample to look for, a year-end deadline. What the internet later made routine — an open problem broadcast to thousands of capable solvers simultaneously — Gardner achieved in 1970 with four pages and a fifty-dollar prize. Life was, in this sense, the first viral open-source mathematical project.
The Second Column: February 1971
Gardner reported back in February 1971. His follow-up — “On Cellular Automata, Self-Reproduction, the Garden of Eden and the Game of ‘Life’” — surveyed four months of reader discoveries.
The headline was the Gosper Glider Gun. Gardner reported the result, described the pattern, confirmed that Conway had paid the prize. The conjecture was false. Infinite growth was real.
But the gun was not the only discovery. Readers had found large stable configurations, new oscillators, and the first hints of what would become a formal taxonomy. The February column also introduced the Garden of Eden — a configuration with no predecessor, a pattern that could exist at generation 1 but could not have evolved from any generation 0. The existence of such patterns raised deep questions about reversibility and information in cellular automata that remain productive today.
Gardner noted that the volume of reader mail showed no signs of diminishing. He was right.
After the First Wave
Life did not stop in February 1971. It accelerated.
The new generation of inexpensive minicomputers — the PDP-11 and its relatives — put interactive computing within reach of smaller institutions, and Life programs appeared on every new platform almost immediately after the platform existed. A legend emerged that at one point in the early 1970s one-quarter of all the world’s computers were running Life simulations. The U.S. military allegedly estimated that the cumulative cost of research computing time consumed by Life enthusiasts ran into the millions of dollars. Whether those specific figures are precise, the direction is not in doubt.
John Conway himself was ambivalent about the fame. He regarded Life as a minor recreational project, and he felt the attention it received was disproportionate to its mathematical depth compared to his work on surreal numbers, the Conway groups, and monstrous moonshine. Life was what everyone knew him for. He had hoped it would not be.
But Life had done something his other work had not: it escaped academia entirely. It landed in the hands of programmers, artists, and biology students who had no investment in finite group theory. It showed, with unusual clarity, that four simple rules could generate a universe of irreducible complexity. That message traveled far.
Martin Gardner had understood this from the start. His column didn’t just describe a game — it framed it as a problem in emergence and computation at the moment those ideas were becoming central to science. Editorial judgment of the highest order.
The October 1970 column was four pages. Its consequences have not stopped accumulating.