In the late 1970s the MIT Artificial Intelligence Laboratory built specialized computers optimized to run Lisp, the language of choice for AI research. Around 1980 the commercialization effort split the lab in two. Daniel Weinreb, who became one of the founders of Symbolics, recounts that Richard Greenblatt proposed a Lisp machine company on terms his colleagues found unworkable - no outside investment, and Greenblatt as CEO. As Weinreb puts it, “the other members of the Lisp machine project were extremely dubious of Greenblatt’s ability to run a company. So Greenblatt and the others went their separate ways and set up two companies.” The result was Symbolics and Lisp Machines Inc. (LMI).
For a few years the bet looked sound. AI was fashionable, expert systems were selling, and a workstation tuned for Lisp was a genuinely fast way to do the work. Symbolics grew on the strength of that demand.
Then the floor moved. In Weinreb’s firsthand account, “the new ‘workstation’ category of computer appeared: the Suns and Apollos and so on.” General-purpose workstations got fast enough and cheap enough to run Lisp acceptably, and the premium for dedicated hardware evaporated. Weinreb also describes internal causes - a corporate culture in which questioning the strategy meant you “just didn’t get it,” management conflict that drove out much of the top team, and costly long-term real-estate commitments for growth that never came. He left in 1988 to start a different company, having concluded Symbolics was not adapting.
The lesson the episode teaches is not that the technology was bad. The Lisp machines were, by accounts of the people who built them, excellent. The lesson is that specialized hardware lives or dies by the gap between it and commodity hardware, and that gap closed faster than the industry’s business model could survive.