Lisp Machines, Inc.

~ the obit ~

Lisp Machines, Inc. sold computers that ran one language all the way down. There was no operating system underneath written in something else. The microcode understood Lisp objects directly, and the entire environment from the window system to the compiler was source you could read and edit while it was running. Around 200 of the second-generation machines shipped. That is the whole commercial footprint.

The design came out of the MIT Artificial Intelligence Laboratory, where Richard Greenblatt and Tom Knight had built the CADR in the late 1970s. In 1979 Greenblatt formed a company in Cambridge, Massachusetts to sell it. Russell Noftsker, who had run the lab, wanted the same thing done with venture capital and a conventional board behind it. Greenblatt wanted to fund the company off customer deposits and keep control. They could not agree, so Noftsker started Symbolics and took most of the lab's hackers with him. Alexander Jacobson, a consultant at Control Data, eventually put money and a business plan behind what was left, and LMI got started anyway.

The first product was the LMI-CADR, essentially the lab machine with a nameplate on it. The LMI-LAMBDA followed in 1983 with reworked instruction fetch and better hardware, and it is the machine LMI is actually remembered for. Texas Instruments licensed the Lambda design and shipped its own version as the TI Explorer, which meant the architecture reached customers who had never heard of a company in Cambridge with fewer than a hundred people in it.

By the mid-1980s the Lambda was no longer competitive and LMI started over on a RISC-like successor called the K-machine. It never reached a customer. The company filed for bankruptcy in 1987 with the design unfinished, eight years after it opened, while the AI hardware market it had been built for was collapsing underneath everyone in it.

~ the rap sheet ~

~ the two companies ~

Until 1981 both companies took their changes to the Lisp Machine system software back to MIT, where the other one could pick them up. Symbolics stopped. Richard Stallman's account is that this was a deliberate move to starve a competitor that did not have the staff to maintain an operating system alone, and by any account it worked. Stallman spent the next two years reimplementing Symbolics' changes from the MIT side so LMI could keep shipping, then concluded that the arrangement he was defending had already ended.

What came out of that was GNU in 1984 and the Free Software Foundation a year later. Copyleft came out of those, and copyleft is the part that traveled. I have never touched a Lisp machine and never will. But the Linux boxes I set up in a university lab in Ankara in the 2000s, and every license header I scrolled past without reading, sit downstream of two Cambridge companies arguing about whether to share a window system. None of the hardware survived. The rule that argument produced is in the kernel of most devices sold today.

~ the machine that never shipped ~

The K-machine was a bet that the special-purpose Lisp processor could be rebuilt on RISC principles and stop losing ground to general workstations. The diagnosis was correct. Sun and the other workstation vendors were taking the AI market by being fast at everything rather than fast at one thing, and a tagged-architecture machine that could not keep up on ordinary code was going to lose regardless of how good its development environment was.

LMI ran out of money before the K-machine could be finished, which is the sentence that ends most of the companies on this site. The work was not entirely wasted. The Lisp environment itself got ported to ordinary hardware by people who wanted to keep using it, and the ideas about tagging and garbage collection kept circulating. What could not be ported was the reason to buy a dedicated machine in the first place.

~ the canadian money ~

The 1987 bankruptcy was a Chapter 11 reorganization rather than a liquidation, and the assets were bought out of it by a new company called GigaMos Systems. Greenblatt was an officer there. The money behind GigaMos came from Guy Montpetit, a Canadian businessman who had raised funds partly on the strength of government interest in a Quebec high-technology venture.

In 1989 Montpetit became the center of a Canadian political scandal, with court testimony in Quebec alleging that money earmarked for the venture had gone instead to a personal loan and a jet. The consequence for the people in Cambridge was blunt. Assets were frozen and the payroll could not be met. The last working remnant of the Lisp machine business stopped for reasons that had nothing to do with computers.

Symbolics outlived it by six years and then went through its own bankruptcy in 1995. Texas Instruments shut the Explorer line down. By the early 1990s no company anywhere was selling a computer whose native language was Lisp, and the people who had wanted one were running Emacs on a Unix workstation instead.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »