Project Cybersyn

On 30 December 1972, Fernando Flores brought Chilean President Salvador Allende and General Carlos Prats, the commander of the army and minister of the interior, up to a hexagonal room in Santiago. The walls were wood-paneled. Seven white fiberglass chairs faced inward in a circle. Allende sat down in one and pressed a button on the armrest, and a screen on the wall changed. Eden Medina, who reconstructed the scene from the participants and their papers , records what happened behind the wall: a slide projector advanced a carousel. The images on the screens were back-projected 35mm slides, many of them drawn by hand by graphic designers, made to look like the flat-panel displays that did not yet exist.

That room was the most photographed part of Project Cybersyn and the least used. The part that did the work was a few hundred telex machines in factories around Chile, typing numbers into a network once a day. The room was a promise about what the network would become. Nine months after Allende sat in that chair, he was dead, the project was over, and a soldier had taken a knife to the slides, one at a time.

The Pitch

In 1971 Chile’s government owned a rapidly growing share of the country’s industry and had no good way to know what was happening inside it. Allende’s Popular Unity coalition was nationalizing copper, coal, textiles and manufacturing faster than it could find managers to run them. The state development agency, CORFO, sat on top of a pile of firms whose output, inventories and shortages arrived in Santiago as paper reports weeks or months late.

The pitch came from Stafford Beer, a British management consultant who had spent the 1960s arguing that a company, a nation, or any organization that wants to survive should be designed like a nervous system. His Viable System Model divided an organization into five nested functions, from the units that do the work up to the level that sets policy, and insisted that each level should handle its own problems and pass upward only what it could not handle. The model’s promise was autonomy with coordination: the factory runs itself, and the center hears about it only when something is going wrong.

Applied to a socialist economy, that promise was unusual. The Soviet model of planning ran on targets sent down from above. Beer’s model ran on signals sent up from below. Each factory would report a handful of indicators daily. Software in Santiago would watch each series and flag any that started to move outside its normal range. The factory’s own management would get the warning first and have a set time to fix it. Only if the problem persisted would it climb to the next level. Beer called these escalating alarms “algedonic” signals, from the Greek for pain and pleasure. The system was meant to tell the government where the economy hurt, quickly, without telling the factories what to do.

Beer liked to say the design served freedom, and he meant it structurally. The less the center had to know, the less it would be tempted to control. When he first explained the five levels to Allende, he drew System 5, the policy level, expecting the president to see himself there. By Beer’s own account , Allende looked at it and said: “Ah! System 5. At last, the people.” The story is Beer’s, and he told it more than once; it may be polished. But the government did hold him to its substance. According to Medina , Allende’s one condition for the project was that it increase worker participation.

That was the pitch: a country with a few mainframes and a hostile superpower to its north would build, in two years, an information system that let the state manage industry without crushing it. It would be cheap, built on borrowed equipment, and designed so that the people running it were the workers, not a technocracy in Santiago. If it worked, it would be a third way between the market and Gosplan.

The People

Fernando Flores was 28 and the technical director of CORFO when, on 13 July 1971 , he wrote to Beer. He had read Beer’s books as an engineering student and wanted help applying them to the nationalized sector. The two met in London in August. Flores was the project’s political protector and its fixer, and he rose fast: after the crisis of October 1972 he joined Allende’s cabinet. On the day of the coup he was at La Moneda with the president. He spent the next three years in military prison camps , until Amnesty International helped negotiate his release in 1976.

Stafford Beer was a big, bearded man who had made a lot of money telling British steel and publishing companies how to organize themselves. He had a house in Surrey and a Rolls-Royce. His terms for Chile, as Americas Quarterly reports , were $500 a day plus a steady supply of chocolate, wine and cigars. He first flew to Santiago in November 1971, met Allende at the presidential palace on 12 November, and within eight days had the team agreed on a plan. Over the following two years he shuttled between Britain and Chile. The project changed him. The Chilean biologist Humberto Maturana said Beer “arrived in the country a businessman and left a hippie.”

Raúl Espejo, a young CORFO engineer, ran the day-to-day work of turning Beer’s diagrams into a functioning system. On 10 September 1973 he wrote a letter rebutting a journalist who had described Cybersyn as a secret Marxist plot. He planned to mail it the next morning. That night, as Americas Quarterly tells it , tanks came into Santiago. Espejo left for Britain, became a consultant and an academic in the cybernetics Beer had taught him, and never went back to the room.

There was also Gui Bonsiepe, a German designer from the Ulm School, who ran the state industrial design group that built the operations room, and whose team spent weeks working out viewing angles and how much text a person could read on a wall from a reclining chair. The room’s look came from him. The room’s purpose came from Beer.

What Got Built

Cybersyn had four parts, of which one did nearly all the work.

Cybernet was the network. Chile did not have the computers or the data lines for anything resembling real-time computing; it had, by most accounts, two mainframes worth mentioning. What it did have was telex. The team found a stock of telex machines sitting unused in a warehouse, bought by a previous government. The commonly repeated figure is 500; Damn Interesting puts it at about 400; and later research summarized on Wikipedia suggests the number actually installed was far smaller. Factories typed their daily figures into the machines, and the messages went to Santiago. Some plants were remote enough that, according to the Beer archive at Metaphorum , the figures had to go by horse to the nearest telex station.

Cyberstride was the software that read those figures. It was written by British engineers from Arthur Andersen and implemented in Santiago by programmers at ECOM, the state computing company. It applied Bayesian time-series filtering to each indicator and flagged the ones drifting toward trouble. Medina’s account has it starting on an IBM 360/50 and moving to a less heavily used Burroughs 3500; newer research says it may have run on the IBM throughout. By early July 1972, around 65 state-owned sites were feeding data to the prototype.

CHECO, the Chilean Economic simulator, was meant to model the whole economy and let planners test policies before trying them. It was built on Jay Forrester’s DYNAMO language from MIT, and its design cost £5,000 . It was the one component that never got past early testing.

The Opsroom was the operations room, in what Americas Quarterly describes as an unassuming eighth-floor room at CORFO’s headquarters. It was a hexagon because a circle had proved too hard to build. One wall opened onto a small kitchen. Another held the Datafeed, four screens in fiberglass cabinets, one large above three small. There was a magnetic board for flowcharts, a diagram of the Viable System Model the team called “Staffy,” and lights for the economic sectors. The chairs had buttons in the right armrest, and next to the buttons, according to 99% Invisible , an ashtray and a place for a whiskey glass. There were no keyboards. Typing was considered secretaries’ work, and the room was designed for ministers and, in theory, for workers’ representatives, none of whom were expected to type.

There was also a fifth piece that existed only as a prototype. Beer’s son Simon built a small device from old radio parts with two dials, meant to let television viewers register approval or disapproval of a policy in real time. The project called it Cyberfolk. It was tested in two cities and went no further.

By May 1973, Medina found, citing the project’s own reports as summarized on Wikipedia , the system had been extended to some degree to 26.7 percent of the nationalized enterprises, which accounted for about half of the sector’s revenue. “To some degree” is doing a lot of work in that sentence, and it is the closest thing to a final measurement the project ever got.

How It Died

The project’s real test came early and was not the one it was designed for. In October 1972, the truck owners’ confederation went on strike against the government. It was backed by business associations, and the United States, by many accounts, helped pay for it through the CIA’s funding of the opposition. Somewhere between 40,000 and 50,000 trucks stopped, in a long, narrow country that moved nearly everything by road. Food, fuel and raw materials stopped with them.

The government took over the telex network and set up a command center next to the communications room. It was not the Opsroom, which was not finished. People sat at telex machines around the clock tracking which roads were open, where fuel was, which drivers were still working and which factories were short. Beer estimated the network was carrying about 2,000 messages a day . Using roughly 200 loyal trucks, the government kept Santiago supplied. One official later told Medina the government would have collapsed that night without the network. The strike ended in early November. Flores went into the cabinet.

It was the high point, and it taught the government something slightly different from what Beer had intended. What saved the day was not algedonic filtering or simulation or worker participation. It was a fast, centralized communication network run by a small team of trusted people under emergency conditions. That is a much older idea.

The Opsroom prototype was finished at the end of December, which is when Allende came to sit in it. Beer drafted a speech for the president to give at a formal inauguration, promising to “hand to the people the power that science commands.” The inauguration never happened.

On 7 January 1973, the British Observer published a story by Nigel Hawkes under the headline “Chile Run by Computer.” It described the system as something close to an electronic dictatorship. Beer and Flores wrote to correct it, explaining that Cybersyn gave no computer any control over anything and only supplied information to people. Medina calls the article “both damaging and untrue,” and it spread: an American newspaper reprinted it with a cartoon, other outlets compared the system to Nineteen Eighty-Four, and in March the Chilean magazine Qué Pasa ran a hostile piece suggesting it would be used for political control. The Chilean opposition had its slogan: Popular Unity controls us by computer. Parts of the left, for their own reasons, distrusted it too. A system built to protect autonomy was now known, at home and abroad, as a machine for surveillance.

Meanwhile the Chilean economy was coming apart. Inflation ran into triple digits. Through 1973 there were strikes, shortages, street battles and, in August, sabotage of pipelines and the power grid. The worker participation that Allende had made the project’s one condition mostly did not materialize; the system was being used, where it was used at all, by managers and officials. Beer, who by mid-1973 felt the project had been left to drift, later remarked that “a cynic could declare that I was left to sink or swim.” In August he pressed Allende directly about worker control. A month later there was no one to press.

The coup came on 11 September 1973. Allende died in the palace that afternoon. Beer was in London, where he had been lobbying for the Chilean government, and learned the news, as Damn Interesting tells it , from a newspaper billboard. The military took over CORFO and questioned the team. Medina reports that a member of the military took a knife and stabbed each of the Opsroom’s slides. The junta’s interest in the network faded quickly. By the Guardian’s 2003 account, the officers found it intact and did not care for its open, egalitarian aspects. The room was dismantled. The economic policy that replaced it came from the Chicago Boys, and it ran on prices.

The Autopsy

The verdict here is Killed. A coup ended the project while its core was working, and there is no version of the story in which Cybersyn would have continued under Pinochet. That much is not in dispute. What is in dispute is whether the thing that was killed was viable, and the strongest case against this verdict deserves a full hearing.

That case says the verdict should be Wrong. It comes from the long argument, going back to Ludwig von Mises and Friedrich Hayek, that central planning fails because the knowledge needed to run an economy is dispersed among millions of people and cannot be gathered at the center, however good the wires. On this reading Cybersyn was a doomed attempt to beat the price system with telexes, and the coup only spared it from the embarrassment of failing in public. The economic record supports the pessimism. Chile in 1973 was in a crisis that no information system could fix. Medina herself is clear that the project as a whole never left the prototype stage, that it reached perhaps a quarter of the nationalized enterprises, and that its signature feature, worker participation, barely happened. The October strike, its one triumph, was won with a technique any wartime government would recognize.

That reading is right about the economy and wrong about the machine. Cybersyn did not try to set prices, allocate resources or calculate a plan. It was an exception-reporting system: a set of daily indicators per factory, a statistical filter, and an escalation rule. Its premise was close to Hayek’s own. Beer designed it because he believed the center could not and should not know everything, and he built it to pass up as little information as possible. Whether that kind of system could work was a reasonable question in 1973 and is not a controversial one now; every large company runs dashboards on the same principle. The narrower question, whether it could work for a nationalized economy in a poor country under a blockade, never got an answer. The project was 22 months old when it ended. Its telex network had already done one thing that mattered, under pressure, in the real world. The rest was being installed.

There is also a genuine secondary cause, and it explains why the project was so exposed when the end came: it had lost the argument about what it was. Between the Observer story and the Qué Pasa piece, Cybersyn spent its last nine months being described as the opposite of its design, and its own team never got the worker participation that would have made the description false. A system meant to distribute power had been built, in practice, by and for a small group of engineers in an eighth-floor room. That does not make it Wrong. It makes it unfinished in the way that mattered most to the people who commissioned it.

So: Killed, while viable as a tool and unproven as a theory of government. The coup destroyed an experiment before it could produce a result, and the result it was heading toward was smaller and more ordinary than either its admirers or its enemies supposed.

What Survived

The Viable System Model. Beer’s framework outlived the project and is still taught and applied in management cybernetics. Espejo spent his career on it after leaving Chile, and the model has a small, continuous community of practitioners; the Metaphorum group maintains an archive of his work. Beer’s best-known line, “the purpose of a system is what it does,” from The Heart of Enterprise (1979), has become a stock phrase of systems thinking .

Fernando Flores’s second career. After his release, Flores went to California, earned a PhD in philosophy at Berkeley, and worked with the Stanford computer scientist Terry Winograd. Their conversations persuaded Winograd that the AI research of the period rested on a mistaken theory of cognition, and the two wrote Understanding Computers and Cognition (1986), an influential book in human-computer interaction. Flores then founded Action Technologies, which built workflow software, including The Coordinator, on the idea that organizations run on conversations and commitments. The line from Cybersyn runs through Flores himself: a man who had tried to run an economy through communication networks spent the next decade designing software for coordinating work. Flores returned to Chile and was elected to the Senate in 2001.

Beer’s own account. Beer went back to Britain, gave away most of what he owned, and moved to an isolated cottage in the Welsh hills . What survived of Cybersyn in his hands was mostly the record: the 1973 CBC Massey Lectures, published as Designing Freedom, and a long section on Chile in the second edition of Brain of the Firm (1981), which is the primary source for much of what is known about the design. The political scientist Kevin Munger argues that after Platform for Change (1975) found few readers, Beer stopped correcting course and his commitment hardened past the point of responding to feedback. He died in 2002.

The design. Bonsiepe’s group made the Opsroom one of the most reproduced images of Chilean design of the period, and the room has become a reference point for designers and historians rather than for planners. Medina’s 2006 article on the project won the IEEE’s prize for electrical history, and her 2011 book turned a half-forgotten episode into a standard case study.

The Ghost Today

The room got built again. In September 2023, for the fiftieth anniversary of the coup, the Centro Cultural La Moneda in Santiago opened How to Design a Revolution, an exhibition of design from the Allende years curated by Hugo Palmarola, Pedro Ignacio Alonso and Eden Medina. Its centerpiece was what the curators called the first comprehensive and functional reconstruction of the Opsroom. The cultural center sits beneath the plaza in front of the presidential palace. The room had always been meant to move to La Moneda, and fifty years late, as a museum piece, it finally did. The exhibition traveled to Barcelona’s Disseny Hub in 2025.

The idea itself is walking around in less flattering places. In 2014 Evgeny Morozov argued in the New Yorker that Cybersyn’s real heirs were the data-driven platforms, citing Uber’s use of real-time demand data. In 2019 Leigh Phillips and Michal Rozworski’s The People’s Republic of Walmart made the matching left-wing argument: that the largest retailers already plan enormous economies internally, with information systems far beyond anything Beer had, and that the socialist calculation debate looks different in light of that. Both arguments point to the same irony. The thing Cybersyn was built to do, watch thousands of production units daily and escalate the exceptions, is now routine. It is done by private companies, at scale, for the purpose of running supply chains. The part that never got built is the part Allende asked for: the workers in the chairs.