You have seen it. You almost certainly don’t know that you have.
In the control room on Isla Nublar, behind John Hammond’s flesh-and-blood arrogance and the raptors getting loose, there is a wall of black towers with red lights flickering across them in restless waves. Most people watching Jurassic Park in 1993 read those lights as this is what a powerful computer looks like and moved on. They were real machines. The production borrowed something like four million dollars of actual hardware from Apple and Silicon Graphics rather than build props — SGI workstations, Macintosh Quadras, and a row of five Connection Machine CM-5 supercomputers, whose blinking red panels were, for the film, wired to flash in pretty random patterns. (The famous “It’s a UNIX system, I know this!” scene is a real piece of software too — fsn, an experimental 3D file browser that SGI shipped and almost nobody used, made immortal by one line of dialogue.)
I want to tell you about those black towers, because the machine in the background of that shot has one of the loveliest and saddest stories in computing, and almost none of it is on screen.
The Connection Machine came out of Thinking Machines Corporation, founded in Waltham, Massachusetts in 1983 by Danny Hillis and Sheryl Handler, with the AI pioneer Marvin Minsky hovering over its early days. The idea, born from Hillis’s doctoral work at MIT, was heretical for its time: instead of one processor going very fast, use a lot of them going together — tens of thousands of tiny minds, wired to talk to each other, all working at once. The first model, the CM-1, packed 65,536 one-bit processors into a five-foot black cube. It was built for exactly the kind of enormous, brute-parallel problems we would now call AI.
Here is the part I love. They cared what it looked like. The visual design of that black cube — the geometry, the signature red LEDs — was led by Tamiko Thiel, and the lights weren’t only decoration: they showed the processors actually firing, thought made visible on the surface of the box. Later, the CM-5’s dramatic staircase silhouette was shaped with input from Maya Lin — yes, the architect of the Vietnam Veterans Memorial. Someone decided a supercomputer should be a thing you could love to look at. They were right. It won the Gordon Bell Prize, twice, as the fastest machine in the world.
And then there is Feynman. Richard Feynman — that Feynman — spent his summers at the company. Hillis’s memoir of him is one of the warmest things in the literature: Feynman turned up, refused vague advice, and demanded a concrete problem, so they handed him the machine’s internal communication network — the router that let all those thousands of processors talk. He suggested the elegant hypercube wiring that connected them. By the end of that first summer he had worked out how the router behaved and, to everyone’s surprise and delight, handed the answer back as a set of partial differential equations. He was analyzing a digital switch as if it were a physical system, because to him it was one.
Beauty, genius, and a genuinely new idea about how to think. Spared no expense.
Thinking Machines went bankrupt in 1994 — one year after those five CM-5s stood in the Jurassic Park control room, at the peak of the machine’s fame. The massively parallel bet was right in the long run and early by a decade; the market underneath it collapsed faster than the vision could pay off. Silicon Graphics, whose workstations shared that same control-room set, is gone too. The whole gorgeous golden age of the specialized workstation — machines built by people who thought hardware could be art — mostly evaporated. If you want to see a CM-5 today you go to a museum, or you rewatch a dinosaur movie.
That should be the end of it. A beautiful thing, briefly the best in the world, outlived by a film. It isn’t the end of it.
Look at who was in that room in 1983. The lead engineer on the Connection Machine — the person who spent six years making the thing actually run — was a young MIT graduate named Brewster Kahle. At Thinking Machines he built a system called WAIS, the first tool that let you search and retrieve documents across a network; he later called the Connection Machine “the world’s first search engine.” The company died. Kahle did not stop.
In 1996 he founded the Internet Archive, and with it the Wayback Machine — the thing that, right now, is quietly keeping a copy of the web so that when a page goes dark it is not simply gone. The man who built the beautiful machine that couldn’t survive went and built the machine whose entire purpose is that nothing else has to disappear. The company that spared no expense and lost seeded, in one of its engineers, a lifelong argument that memory is worth keeping.
That is the pattern I keep finding, and I don’t think it’s a coincidence. Nothing survives on its own. The Connection Machine survives because a blockbuster happened to point a camera at it, and because a documentarian named Fabien Sanglard loved it enough to catalog every machine in that film in excruciating detail, and because one of its engineers turned the whole idea of not losing things into an institution. The hardware died. The company died. What lived was carried — by film, by a fan’s care, by a person who decided preservation was a life’s work.
The blinking lights on those black towers were the processors thinking, made visible. It turns out the thinking outlived the machine. It usually does — but only when someone chooses to keep it.
Sources
fsn.