Earlier this week I posted Richard Stallman's recollections of the AI Lab at MIT, where he first encountered and came to love the hacker world and its spirit. That idyllic period came to an end as a result of the commercialisation of the AI Labs' computer system, called the Lisp Machine, which led to the destruction of the unique environment that created it in the first place, and to its re-birth as the GNU project.
[In 1979] a couple of companies were formed to manufacture computers that had been designed in the AI Lab – one was Lisp Machine Incorporated, and the other was Symbolics. Lisp Machine Incorporated [LMI] was founded by [Richard] Greenblatt who was also the person who had designed the Lisp machine more or less. He wanted to make what he called a "hacker company", which would mean that it would be less controlled by management which was unfriendly to hackers than most companies, and it would have nicer policies in some way than companies typically did.
And one of the ideas was that they wouldn't get outside investors, because outside investors would insist on imposing all the usual ways of doing things that would make it ugly. And another idea was that they wouldn't just hire away the hackers from the [AI] Lab, but would hire them part-time, so the Lab would still have its hacker culture, the community that we belonged to wouldn't be wiped out.
But then the other people said that they didn't trust Greenblatt to run a business, so Greenblatt found somebody who was a friend of his, somebody who used to be one of the administrators at the AI Lab, namely Russell Noftsker, who was in fact the person who had hired me, who'd given me my first job there the day I wandered over. He had left, he started some other business, I think, so he had business experience.
Greenblatt brought him in, and [Noftsker] then decided that they would go form a different company, which would try to get as much investment as possible and do all the usual ugly things, although a bunch of the hackers thought that it wouldn't really be ugly like the others, but they found out they were wrong. They made all the usual decisions, and got the usual consequences.
Greenblatt wasn't daunted by this, he went ahead and formed a company too. So there were two companies to make Lisp machines. And Greenblatt's plan was successful enough even without the help of all the other old Lab hackers, that it's clear they were mistaken in thinking that it was a hopeless idea. He found one customer willing to make progress payments – essentially pay in advance for the first two machines to be built. And with the profit from that, they were able to build three more machines and sell them to other people, and with the money from that they were able to build more machines, and so LMI got off the ground.
Well, Symbolics with its investors had more money and hired several of the best hackers from the Lab, and a year later they hired the rest of the hackers except for me and Greenblatt. And the result was that my community was wiped out. It felt like a ghost town. It was desolate, and I was grief-stricken. And at that time, the AI Lab was buying a new computer. And there were no hackers to port ITS to it – it was still a PDP-10, but it was a somewhat different model, and some changes would have been needed to make ITS run. And there was nobody to do it, and they decided to switch to a proprietary time-sharing system from Digital.
By then, Symbolics had given everyone at the AI Lab an ultimatum. They said: We demand that you choose a side. I had been neutral, I had been more favourable towards Greenblatt's company, but not particularly involved with it in any way. But at that point they demanded that I, like everyone else in the Lab, choose a side. And I said, in that case, my choice is obvious, I'm against you.
The neutrality as I saw it of the Lab, had been attacked. I compared it with Belgium, and it's obvious what you do in a situation like that: you join the war against the attacker. They were essentially asking people to choose either MIT's version of the software for the Lisp machine, which nominally was being shared with both companies, or Symbolics's version of the software. They said that they would let people at MIT run Symbolics's version but they would not let people at MIT use Symbolics's changes in the MIT version of the system which would have meant that LMI could get them.
And they expected that everybody would do the easy thing, which would have been to use Symbolics's version. They expected to herd everybody with little carrot and a little stick. They didn't expect somebody who says: No, I will not be herded, I will not be used by you as a pawn for you to hurt another company which has treated my community better than you have, and which didn't issue an ultimatum, the way you did, which didn't attack my community.
They'd won away part of the community and then they came back with an invading army was the way I saw it. And the only thing to do was to make sure that it didn't profit them, what they had done. And I effectively did that, because I kept the MIT version of the system going, for two more years, knowing full well that my community itself was dead, and that really this was to the benefit of LMI. But benefitting LMI was punishing Symbolics for what it had done.
In some ways it was very comfortable because I was doing almost nothing else, and I would go to sleep whenever I felt sleepy, when I woke up I would go back to coding, and when I felt sleepy again I'd go to sleep again, and I had nothing like a daily schedule. I'd sleep probably for a few hours one and half times a day, and it was wonderful, I felt more awake than I've ever felt. And I got a tremendous amount of work done – I did it tremendously efficiently.
And I thought of myself as being at war. I had myself taken off of all mailing lists, because I wanted nothing to distract me because I knew I had to compete with a much larger army and that I would have to do the work of several people. I'd have to work in the most focussed way I possibly could. It was exhilarating sometimes, sometimes it was terribly wearying.
When the people who had gone to Symbolics had stopped working on the MIT system sources, they had left them in an inconsistent state – it wouldn't build. So we had lots of problems to fix in a large system. What would happen was, I would do some work, and then when I got tired, I'd go to sleep and Greenblatt would take over, and then I'd wake up and see how he'd left things, I would look and see what bugs he was dealing with, and I would fix a bunch of them, and that would leave some more, and then he'd fix some – and that felt very good. But then when it was only me – he had other things to do, he was running LMI - so it was just me, and it was in some ways terribly lonesome, but I kept doing it, I wouldn't let anything stop me.
That started in March 1982. In fact it was on my birthday that we got the ultimatum from Symbolics that we were all going to have to choose a side. By 1983, LMI was getting bigger and was able to hire some programmers, so I could see that LMI was going to be able to do this work for itself. And I could also see that Symbolics had designed a new kind of machine and the AI Lab was starting to buy these machines, and there was no hope – and not much point, either – in making the MIT version of the system run on those machines, since LMI wouldn't have machines like that. They were also designing a new machine, but it was different.
I also could see that as the AI Lab switched over to those machines, I would lose my ability to do the job, because I couldn't do the job without users. Users had to find the bugs. I could write code, but I couldn't test everything myself, if I had to test it as well as write it that would be more than I could. So I depended on having users using the MIT version of the system. But when they switched over to these new Symbolics machines, they wouldn't be able to test my code any more. I would have been unable to continue.
But in any case, I saw that LMI was going to be able to do this on its own, and I decided I don't want to just continue punishing Symbolics forever. They destroyed my community, now I want to build something to replace. So I started looking for a way I could do that. This was in 1983, while I was still doing the work to keep the MIT system on a par with the Symbolics system. I could see that I wasn't going to do that indefinitely, and I was looking for what I could do after that, that would build a community to replace the one Symbolics destroyed.
The first announcement about the GNU project appeared in September of that year, beginning Stallman's long journey to recreate the community of hackers he had found then lost at MIT.