The Fight That Broke a Lab

By 1979, a small group of programmers at MIT had built something nobody else in the world had: a computer designed, from the circuitry up, to think in LISP. They called it the LISP Machine, and it worked so well that customers were already lining up to buy one. What happened next never made it into a market report. It happened in a room, between two men who disagreed about money — and it quietly determined the fate of the very hardware that would collapse, a decade later, in the second AI Winter.

The two men were Richard Greenblatt, the machine's chief architect and the closest thing the MIT AI Lab had to a resident purist, and Russell Noftsker, a former lab administrator who'd left years earlier to run his own company and had come back convinced the LISP Machine was worth a fortune. Greenblatt wanted to build the company the way the lab itself worked — funded by customer orders, controlled by the people who'd built the thing, no outside investors calling the shots. Noftsker wanted venture capital, real money, and the ownership structure that came with it. Neither man would bend. The disagreement went to the rest of the lab, and the lab split in half.

Most of the hackers sided with Noftsker. He walked away with the majority of the team and, in 1980, incorporated as Symbolics — the very company Episode 8 names as the dominant LISP Machine vendor before its market evaporated. Greenblatt, backed by a handful of loyalists and a CDC consultant who badly needed a machine built, scraped together his own venture: Lisp Machines, Incorporated. Two companies, born from the same lab, the same code, the same three years of work — now locked in direct competition before either had shipped a product.

One person from the lab wasn't recruited by either side: Richard Stallman. He watched the split as a betrayal of everything the lab had stood for — a culture where code was shared freely, where the point of building something was the building, not the owning. When Symbolics began keeping its software improvements to itself, Stallman spent two solid years, largely alone, rewriting equivalent features from scratch so the lab's machines wouldn't depend on a company he no longer trusted. It didn't save the lab. But the resentment it left behind became, within a few years, the founding grievance of the free software movement — the philosophy that still shapes how open-source code gets written today.

The video will walk you through the three mechanisms that ended the Expert Systems era on a balance sheet: maintenance costs, hardware economics, institutional confidence. What it won't tell you is that one of those companies — the hardware company whose collapse is charted as a clean crossing line on a graph — was fractured before it ever sold a single machine. Symbolics and LMI spent the 1980s competing against each other almost as fiercely as they competed against the cheaper workstations that eventually buried them both. A market doesn't just die from outside pressure. Sometimes it's already divided against itself, and nobody in the boardroom says so out loud.

That's worth sitting with the next time a "market collapse" gets explained purely in numbers. Behind the chart is usually a room full of people who disagreed about something long before the money did.

The post stops at the founding fracture. The video picks up where the fracture becomes a forecast — the exact three-part structure that took the Expert Systems industry from a $600-million-a-year market to a punchline, and the quiet 1986 result that was already pointing toward what comes next. Watch Episode 8 to see how a ceiling gets built into an architecture, and why the field's only real move was to remove it, not raise it.

Two companies. One lab. A ceiling nobody saw until the bill came due.