by Steven Levy
Tom Knight, who drifted up to the ninth floor as a startlingly tall and skinny seventeen-year-old freshman in 1965, went through that process, eventually earning winner status. To do that, he later recalled, “You have to pretty much bury yourself in that culture. Long nights looking over the shoulder of people who were doing interesting things that you didn’t understand.” What kept him going was his fascination with the machine; how it let you build complicated systems completely under your control. In that sense, Knight later reflected, you had the same kind of control that a dictator had over a political system. But Knight also felt that computers were an infinitely flexible artistic medium, one in which you could express yourself by creating your own little universe. Knight later explained: “Here is this object you can tell what to do, and with no questions asked, it’s doing what you tell it to. There are very few institutions where an eighteen-year-old person can get that to happen for him.”
People like Knight and Silver hacked so intensely and so well that they became winners. Others faced a long uphill climb, because once hackers felt that you were an obstacle to the general improvement of the overall system, you were a loser in the worst sense and should be either cold-shouldered or told to leave outright.
To some, that seemed cruel. A sensitive hacker named Brian Harvey was particularly upset at the drastically enforced standard. Harvey successfully passed muster himself. While working on the computer he discovered some bugs in the TECO editor, and when he pointed them out, people said, fine—now go fix them. He did, realized that the process of debugging was more fun than using a program you’d debugged, and set about looking for more bugs to fix. One day while he was hacking TECO, Greenblatt stood behind him, stroking his chin as Harvey hammered in some code, and said, “I guess we ought to start paying you.” That was the way you were hired in the lab. Only winners were hired.
But Harvey did not like it when other people were fingered as losers, treated like pariahs simply because they were not brilliant. Harvey thought that Marvin Minsky had a lot to do with promulgating that attitude. (Minsky later insisted that all he did was allow the hackers to run things themselves—“the system was open and literally encouraged people to try it out, and if they were harmful or incompetent, they’d be encouraged to go away.”) Harvey recognized that, while on the one hand the AI lab, fueled by the Hacker Ethic, was “a great intellectual garden,” on the other hand it was flawed by the fact that who you were didn’t matter as much as what kind of hacker you were.
Some people fell right into a trap of trying so hard to be a winner on the machine that they were judged instantly as losers: for instance, Gerry Sussman, who arrived at MIT as a cocky seventeen-year-old. Having been an adolescent electronics junkie and high school computer fan, the first thing he did when he arrived at MIT was to seek a computer. Someone pointed him to Tech Square. He asked a person who seemed to belong there if he could play with the computer. Richard Greenblatt said, go ahead, play with it.
So Sussman began working on a program. Not long after, this odd-looking bald guy came over. Sussman figured the guy was going to boot him out, but instead the man sat down, asking, “Hey, what are you doing?” Sussman talked over his program with the man, Marvin Minsky. At one point in the discussion, Sussman told Minsky that he was using a certain randomizing technique in his program because he didn’t want the machine to have any preconceived notions. Minsky said, “Well, it has them, it’s just that you don’t know what they are.” It was the most profound thing Gerry Sussman had ever heard. And Minsky continued, telling him that the world is built a certain way, and the most important thing we can do with the world is avoid randomness, and figure out ways by which things can be planned. Wisdom like this has its effect on seventeen-year-old freshmen, and from then on Sussman was hooked.
But he got off on the wrong foot with the hackers. He tried to compensate for his insecurity by excessive bravado, and everyone saw right through it. He was also, by many accounts, terrifically clumsy, almost getting himself flattened in a bout with the robot arm—which he had infinite trouble controlling—and once he accidentally crushed a special brand of imported Ping-Pong ball that Gosper had brought into the lab. Another time, while on a venture of the Midnight Computer Wiring Society, Sussman got a glob of solder in his eye. He was losing left and right.
Perhaps to cultivate a suave image, Sussman smoked a pipe, the utterly wrong thing to do on the smokeaphobic ninth floor, and one day the hackers managed to replace some of his tobacco with cut-up rubber bands of the same approximate color.
He unilaterally apprenticed himself to Gosper, the most verbally profound of the hackers. Gosper might not have thought that Sussman was much of a winner at that point, but he loved an audience, and tolerated Sussman’s misguided cockiness. Sometimes the wry guru’s remarks would set Sussman’s head spinning, like the time Gosper offhandedly remarked that “Well, data is just a dumb kind of programming.” To Sussman, that answered the eternal existence question, “What are you?” We are data, pieces of a cosmic computer program that is the universe. Looking at Gosper’s programs, Sussman divined that this philosophy was embedded in the code. Sussman later explained that “Gosper sort of imagined the world as being made out of all these little pieces, each of which is a little machine which is a little independent local state. And [each state] would talk to its neighbors.”
Looking at Gosper’s programs, Sussman realized an important assumption of hackerism: all serious computer programs are expressions of an individual. “It’s only incidental that computers execute programs,” Sussman would later explain. “The important thing about a program is that it’s something you can show to people, and they can read it and they can learn something from it. It carries information. It’s a piece of your mind that you can write down and give to someone else just like a book.” Sussman learned to read programs with the same sensitivity that a literature buff would read a poem. There are fun programs with jokes in them, there are exciting programs which do The Right Thing, and there are sad programs which make valiant tries but don’t quite fly.
These are important things to know, but they did not necessarily make you a winner. It was hacking that did it for Sussman. He stuck at it, hung around Gosper a lot, toned down his know-it-all attitude, and, above all, became an impressive programmer. He was the rare loser who eventually turned things around and became a winner. He later wrote a very complicated and much-heralded program in which the computer would move blocks with a robot arm; and by a process much like debugging, the program would figure out for itself which blocks it would have to move to get to the one requested. It was a significant step forward for artificial intelligence, and Sussman became known thereafter as more of a scientist, a planner. He named his famous program HACKER.
One thing that helped Sussman in his turnaround from loser to winner was a sense of what The Right Thing was. The biggest losers of all, in the eyes of the hackers, were those who so lacked that ability that they were incapable of realizing what the true best machine was, or the true best computer language, or the true best way to use a computer. And no system of using a computer earned the hackers’ contempt as much as the time-sharing systems which, since they were a major part of Project MAC, were also based on the ninth floor of Tech Square. The first one, which was operating since the mid-sixties, was the Compatible Time-sharing System (CTSS). The other, long in preparation and high in expense, was called Multics and was so offensive that its mere existence was an outrage.
Unlike the quiltwork of constantly improving systems programs operating on the PDP-6, CTSS had been written by one man, MIT Professor F.J. Corbató. It had been a virtuoso job in many respects, all carefully coded and ready to run on the IBM 7094, which would support a series of terminals to be used simultaneously. But to the hackers, CTSS represented bureaucracy and IBM-ism. “One of the really fun things about computers is that you have control over them,” CTSS foe Tom Knight would later explain. “When you have a bureaucracy around a com
puter you no longer have control over it. The CTSS was a ‘serious’ system. People had to go get accounts and had to pay attention to security. It was a benign bureaucracy, but nevertheless a bureaucracy, full of people who were here from nine to five. If there was some reason you wanted to change the behavior of the system, the way it worked, or develop a program that might have only sometimes worked, or might have some danger of crashing the system, that was not encouraged [on CTSS]. You want an environment where making those mistakes is not something for which you’re castigated, but an environment where people say, ‘Oops, you made a mistake.’”
In other words, CTSS discouraged hacking. Add to this the fact that it was run on a two-million-dollar IBM machine that the hackers thought was much inferior to their PDP-6, and you had one loser system. No one was asking the hackers to use CTSS, but it was there, and sometimes you just have to do some hacking on what’s available. When a hacker would try to use it, and a message would come on screen saying that you couldn’t log on without the proper password, he would be compelled to retaliate. Because to hackers, passwords were even more odious than locked doors. What could be worse than someone telling you that you weren’t authorized to use his computer?
As it turned out, the hackers learned the CTSS system so well that they could circumvent the password requirements. Once they were on the system, they would rub it in a bit by leaving messages to the administrators—high-tech equivalents of “Kilroy Was Here.” Sometimes they would even get the computer to print out a list of all current passwords, and leave the printout under an administrator’s door. Greenblatt recalls that the Project MAC-CTSS people took a dim view of that, and inserted an official MAC memo, which would flash when you logged in, basically saying, a password is your sanctity, and only the lowest form of human would violate a password. Tom Knight got inside the system and changed the heading of that memo from MAC to HAC.
But as bad as CTSS was, the hackers thought Multics was worse. Multics was the name of the hugely expensive time-sharing system for the masses being built and debugged on the ninth floor. Though it was designed for general users, the hackers evaluated the structure of any system in a very personal light, especially a system created on the very floor of the building in which they hacked. So MULTICS was a big topic of hacker conversation.
Originally, Multics was done in conjunction with General Electric; then Honeywell stepped in. There were all sorts of problems with it. As soon as the hackers heard that the system would run on teletype Model 33 terminals instead of fast, interactive CRT displays, they knew the system was a total loser. The fact that the system was written in an IBM-created computer language called PL/I instead of sleek machine language was appalling. When the system first ran, it was incredibly sluggish. It was so slow that the hackers concluded the whole system must be brain-damaged, a term used so often to describe Multics that “brain-damaged” became a standard hackerese pejorative.
But the worst thing about Multics was the heavy security and the system of charging the user for the time. Multics took the attitude that the user paid down to the last nickel; it charged some for the memory you used, some more for the disk space, more for the time. Meanwhile the Multics planners, in the hacker view, were making proclamations about how this was the only way that utilities could work. The system totally turned the Hacker Ethic around—instead of encouraging more time on the computer (the only good thing about time sharing as far as most hackers were concerned), it urged you to spend less time—and to use less of the computer’s facilities once you were on! The Multics philosophy was a disaster.
The hackers plagued the Multics system with tricks and crashes. It was almost a duty to do it. As Minsky would later say, “There were people doing projects that some other people didn’t like and they would play all sorts of jokes on them so that it was impossible to work with them . . . I think [the hackers] helped progress by undermining professors with stupid plans.”
In light of the guerrilla tendencies of hackers, the planners in charge of the AI lab had to tread very lightly with suggestions that would impact the hacker environment. And around 1967, the planners wanted a whopper of a change. They wanted to convert the hackers’ beloved PDP-6 into a time-sharing machine.
By that time, Minsky had turned many of his AI lab leadership duties over to his friend Ed Fredkin, Nelson’s boss at Triple-I who himself was easing out of full-time business and into a professorship at MIT. (Fredkin would be one of the youngest full professors on the faculty, and the only full professor without a degree.) A master programmer himself, Fredkin was already close to the hackers. He appreciated the way the laissez-faire attitude allowed hackers to be dazzlingly productive. But he thought that sometimes the hackers could benefit from top-down direction. One of his early attempts to organize a “human wave” approach toward a robotics problem, assigning the hackers specific parts of the problem himself, had failed ignominiously. “Everyone thought I was crazy,” Fredkin later recalled. He ultimately accepted the fact that the best way to get hackers to do things was to suggest them, and hope that the hackers would be interested enough. Then you would get production unheard of in industry or academia.
Time sharing was something that Minsky and Fredkin considered essential. Between hackers and Officially Sanctioned Users, the PDP-6 was in constant demand; people were frustrated by long waits for access. But the hackers did not consider time sharing acceptable. They pointed at CTSS, Multics, even at Jack Dennis’ more amiable system on the PDP-1, as examples of the slower, less powerful access one would be stuck with when one shared the computer with others using it at the same time.
They noted that certain large programs could not be run at all with time sharing. One of these was a monster program that Peter Samson had been working on. It was sort of an outgrowth of one of his first hacks on the TX-0, a program which, if you typed in the names of two subway stations on the MTA, would tell you the proper subway lines to take, and where to make the changes from one to another. Now, Samson was tackling the entire New York subway system . . . he intended to put the entire system in the computer’s memory and the full timetable of its trains on a data disk accessible by the computer. One day he ran the program to figure out a route by which a person could ride the entire subway system with one token. It got some media attention, and then someone suggested that they see if they could use the computer to actually do it, break a record previously set by a Harvard student for actually traveling to every stop on the New York subway system.
After months of hacking, Samson came up with a scheme, and one day two hackers made the run. A teletype was installed at the MIT Alumni Club in Manhattan, connected to the PDP-6. Two dozen or so messengers were stationed along the route, and they periodically ducked into pay phones, constantly updating schedule information, calling in late trains, reporting delays, and noting missed connections. The hackers at the teletype pounded in the information, and back in Cambridge the PDP-6 calculated changes in the route. As the travelers passed each station, Samson marked it off on a war-room map. The idea of these crew-cut madmen—stark contrast to the long-haired protesters making news in other sorts of activities—captured the imagination of the media for a day, and The Great Subway Hack was noted as one of the memorable uses of the PDP-6.
It underlined something that Greenblatt, Gosper, and the rest considered essential—the magic that could come only from programs using all of the computer. The hackers worked on the PDP-6, one by one, as if it were their own personal computer. They would often run display programs which ran in “real time” and required the computer to constantly refresh the screen; time sharing would make the display hacks run slower. And the hackers had gotten used to little frills that came from complete control of the PDP-6, like being able to track a program by the flashing lights (indicating which registers in the machine were firing). Those perks would be gone with time sharing.
At heart, though, the time-sharing issue was an esthetic question. The very idea that you could not cont
rol the entire machine was disturbing. Even if the time-sharing system allowed the machine to respond to you in exactly the same way as it did in single user mode, you would just know that it wasn’t all yours. It would be like trying to make love to your wife, knowing she was simultaneously making love to six other people!
The hackers’ stubbornness on this issue illustrated their commitment to the quality of computing; they were not prepared to compromise by using an inferior system that would serve more people and perhaps spread the gospel of hacking. In their view, hacking would be better served by using the best system possible. Not a time-shared system.
Fredkin was faced with an uphill political struggle. His strategy was to turn around the most vehement of the anti-time-sharing camp-Greenblatt. There was a certain affection between them. Fredkin was the only person on the ninth floor who called Greenblatt “Ricky.” So he courted. He cajoled. He told Greenblatt how the power of the PDP-6 would be improved by a new piece of hardware, which would expand its memory to a size bigger than any computer in the world. He promised that the time-sharing system would be better than any to date—and the hackers would control it. He worked on Greenblatt for weeks, and finally Ricky Greenblatt agreed that time sharing should be implemented on the PDP-6.
Soon after that, Fredkin was in his office when Bill Gosper marched in, leading several hackers. They lined up before Fredkin’s desk and gave him a collective icy stare.
“What’s up?” Fredkin asked.
They kept staring at him for a while longer. Finally they spoke.
“We’d like to know what you’ve done to Greenblatt,” they said. “We have reason to believe you’ve hypnotized him.”