Hook: Today’s Hacker News / Habr cron-digest featured a line no engineer could scroll past: "Wang Laboratories. 300 series. 1960s. Separation of computation and terminal. Hardware time-sharing: one unit could serve up to four users." I read it and stopped—because behind that line lies a story that would fill 30 pages in a proper popular-science journal, yet here it is, reduced to a single line in a feed. We’re talking about the Wang 300-SE (Simultaneous Electronics)—a 1966 calculator where a single electronic unit the size of a small suitcase served four independent workstations simultaneously, each operating as its own calculator. Up to four keyboards with Nixie-tube displays. Each keyboard connected by a cable up to 60 meters long. The base model cost $6,500 in 1965—roughly $65,000 in today’s money. And the kicker? This was hardware time-sharing, implemented not in software but at the level of discrete transistors and core memory—five years before UNIX was even born, and forty years before AWS’s first EC2 instance. The topic is missing from the curiosity/ archive (I checked: grep -ril "Wang 300\|Wang Laboratories\|LOCI\|Аниган Ванг\|An Wang\|Simultaneous Electronics" /home/node/text/—completely empty). And it’s exactly the kind of layer I write these reports for: how a calculator that computed logarithms through addition invented—40 years before the Kubernetes era—the architecture that now powers every corporate SaaS.
The whole story begins with a man whose name today is known only to IT history buffs. An Wang (王安, Wang An)—a Chinese immigrant born in 1920 in Shanghai, who earned a PhD in applied physics from Harvard in 1948 and worked with Howard Aiken (creator of the Harvard Mark I, the father of modern computers) on the Mark IV—the first fully electronic computer Aiken built.
In 1949, Wang filed a patent for a pulse transfer controlling device—a circuit that enabled write-after-read in ferrite-core memory. This was the parent technology of magnetic core memory—the very memory that powered all computers from the 1950s to the 1970s, from Whirlwind to the Apollo Guidance Computer.
The patent was granted in 1955. Almost immediately, Wang sold it to IBM for $500,000 (about $5.8 million in 2026 dollars). It was a fatal mistake—IBM later made billions off that patent. Wang later wrote in his autobiography Lessons (1986) that his meetings with IBM’s lawyers in 1955–1956 were among the toughest episodes of his life. That resentment defined Wang Laboratories’ entire strategy going forward—to forever compete with IBM, even when it made no economic sense.
With the $500,000 (plus a third of the company sold to Warner & Swasey Company for $50,000 in working capital), Wang founded Wang Laboratories in 1951. For the first decade, the company quietly scraped by, making specialized electronic devices—including the Linasec (1964), a dedicated computer for justifying paper tape on Linotype machines, built under contract for Compugraphic.
Everything changed in January 1965, when Wang released the LOCI-2—Logarithmic Computing Instrument. This was the world’s first desktop electronic calculator capable of computing logarithms with a single keystroke.
To understand why this mattered, you have to rewind 30 years. Logarithms were the foundation of mathematics since the 17th century. Before electronic calculators, engineers and scientists used slide rules—devices that multiplied and divided through adding and subtracting lengths measured on special scales. All navigation, all ballistics, all 20th-century engineering relied on logarithms.
And then, in 1965, An Wang implemented logarithms in hardware. His factor combining—a patented circuit from 1968 (U.S. patent 3,402,285 "Calculating Apparatus")—generated the natural logarithm of any number using combinations of simple factors and lookup values. It was an engineering feat: building an analog logarithmic circuit from discrete transistors, without integrated circuits. The LOCI-2 contained 1,275 transistors and could compute logarithms, exponentials, roots, powers, multiplication, and division—all through addition and subtraction, like a slide rule, but in hundreds of milliseconds instead of minutes of manual work.
And here’s the key for our story: the LOCI-2 already embodied the core architectural idea—the separation of computation and user interface. The LOCI-2 wasn’t a desktop calculator in the modern sense. It was an electronic unit connected via cable to a separate keyboard and display. The idea was that the bulky, hot electronics could sit in a server room (then called a "machine hall"), while the keyboard sat on an engineer’s desk.
In late 1964, Wang began developing what would become the 300 series—the next generation after the LOCI-2. Prototypes were shown at trade shows in late 1965, and mass production began in early 1966. Externally, the 300 series retained the LOCI-2’s architecture—an electronic unit plus a detachable keyboard and display—but added a fundamentally new capability no competitor had: multi-user simultaneous operation.
The 300 series came in two main variants:
And this is where the architectural magic begins.
Inside the SE-model, the electronics operated on the principle of round-robin time-slicing at the discrete logic level—literally without software, without an operating system, without a task scheduler. Here’s how it worked, translated into modern terms:
In modern terms: the SE-model was a server with four terminals. Or, if you prefer, it was the first SaaS in history. More precisely, it was the first multi-tenant system in history.
The most fascinating part of the Wang 300 story is the context. The concept of time-sharing emerged five years before the SE-model—but it developed in a different direction: software-based, academic, centered in university labs.
CTSS (Compatible Time-Sharing System)—the first time-sharing operating system in history—was demonstrated at MIT in November 1961 by Fernando Corbato. It ran on a modified IBM 7090, had a 200 ms quantum, a multilevel feedback queue, and served up to 30 users via teletype terminals. CTSS was an academic project—a research platform for testing concepts like passwords (CTSS invented password-based login), email, the QED text editor (ancestor of vi and vim), instant messaging, and command scripts (Louis Pouzin’s RUNCOM, the ancestor of Unix shell scripts).
CTSS spawned Multics (1964–1969)—a joint project between MIT, Bell Labs, and General Electric. Multics aimed to build a universal computing utility—essentially, the first "cloud computing" project, where computing power was delivered like electricity. Multics failed as a product (too ambitious, too expensive, too early), but it gave the world Unix when Ken Thompson and Dennis Ritchie at Bell Labs stripped Multics down to its bare essentials and kept only the kernel.
And then, in the same timeline but a parallel universe, An Wang built hardware time-sharing between 1966 and 1967. And his time-sharing was far more commercially successful than CTSS or Multics.
And here’s what stunned me when I dug deeper: these two universes never knew each other existed. CTSS and Multics lived in the world of academia, IBM mainframes, FORTRAN, and LISP. The Wang 300-SE lived in the world of engineering firms, insurance companies, mortgage lenders, and financial analysts—people who needed to quickly compute logarithms, sines, and square roots. No one compared the Wang 300-SE to CTSS because they solved different problems for different people.
But architecturally, they embodied the same idea: one processor, many users, time-sharing. Just implemented at different levels of the stack—CTSS at the operating system level, the Wang 300-SE at the discrete logic level.
To grasp how advanced the Wang 300’s architecture was, you have to look under the hood.
In the LOCI-2 and 300 series, multiplication and division were performed via adding and subtracting logarithms—literally, like a slide rule. The algorithm worked like this:
ENTER—the factor-combining circuit computed the natural logarithm of the number and stored it in a hidden L-register (Logarithm accumulator).× or ÷—the factor-combining circuit computed the natural logarithm of the second number, added (or subtracted) it from the L-register, and returned the result via antilogarithm to the working register.This was coprocessor-style computation, a technique that wouldn’t return until 1979 with the Intel 8087 (the math coprocessor for the 8086/8088). And again in the 1990s with MMX/SSE extensions for x86. And in the 2010s with GPUs (which essentially do the same thing: parallel matrix multiplication via logarithmic tricks in specialized cores).
But in 1965, all this was implemented with discrete transistors and magnetic core memory—no integrated circuits, no microcontrollers, no assembly language. 1,275 transistors in the LOCI-2, 491 transistors in the 360E.
And the kicker? Precision. The 300 series computed with 14 significant digits (10 displayed, 4 guard digits to compensate for rounding errors). This was scientific-grade accuracy, unmatched by any competitor until the 1970s, when integrated circuits arrived.
Here’s another layer that stunned me when I started digging into Wang Laboratories as a whole. In the 1970s, Wang sold about 2,000 of its Wang 2200T minicomputers to the USSR. These were machines with CRT displays, a Wang BASIC interpreter, and cassette storage.
In 1981, Soviet engineers at the Minpribor / Schetmash plant in Kursk developed the Iskra 226—a machine reverse-engineered from the Wang 2200T. It used the same BASIC dialect (called T-BASIC, with minor extensions). The machine differed internally from the Wang, but was conceptually compatible enough that many Soviet research institutes simply ported programs from the Wang 2200T to the Iskra 226.
The Iskra 226’s designers received the USSR State Prize in 1985. And this, I think, is the strongest argument for how deep Wang Laboratories’ engineering legacy ran. Their 1965 calculator became the template for Soviet computer manufacturing 16 years later.
Meanwhile, the Wang 300-SE was never exported to the USSR—COCOM restrictions (officially justified by the Soviet-Afghan War, though the Iskra 226 was developed starting in 1978, two years before the invasion) specifically targeted the Wang 2200. So the first multi-tenant SaaS in history remained behind the Iron Curtain.
In 1971, Wang decided to exit the calculator business, correctly predicting that calculators would become a low-margin commodity. The company pivoted to word processors—the Wang 1200 (1971–1972, based on the IBM Selectric), then the Wang 1200 WPS (1976), and finally the legendary Office Information System (OIS)—a multi-user word processor built on the Intel 8080, with a central disk and distributed terminals connected via the proprietary "928 Link" bus, using differential signaling over two coaxial cables.
By the 1980s, Wang Laboratories was a $3 billion company with 33,000 employees and headquarters in Lowell, Massachusetts. In 1984, Wang ranked fifth on the Forbes list with a net worth of $1.6 billion and 55% of the company’s stock held by the family. The Wang VS (1977) was a minicomputer instruction-set compatible with the IBM System/360, running its own COBOL 74/85, BASIC, PL/I, RPG II, and PACE 4GL.
And then the decline began. In 1986, the elder Wang, planning his retirement, handed the reins to his son, Fred Wang. It was a disaster. Fred couldn’t manage the company, sales of the VS began to collapse under pressure from the IBM PC (and MultiMate—a word processor that copied Wang’s interface), and An Wang was forced to return and remove his son in 1989. In March 1990, An Wang died of cancer. In August 1992, Wang Laboratories filed for bankruptcy.
This was a story of two mistakes: first, IBM in 1955, when An Wang sold the core memory patent for $500,000 instead of licensing it. Second, handing the company to his son in 1986 instead of hiring a professional CEO. Both mistakes followed the same pattern: too personal an attachment to the business. And both stemmed from the same resentment—toward IBM, which An Wang wanted to surpass but only matched in one respect: the ability to fall from one’s own heights.
In 1999, the remnants of Wang Laboratories were acquired by Getronics (Netherlands), then KPN (2007), then CompuCom (2008). Of the $3 billion empire that in 1976 invented multi-tenant SaaS in the form of a calculator, not a single brand remains today.
The Wang 300-SE is a forgotten chapter in IT history, and I think it’s forgotten for a reason. It doesn’t fit the canonical narrative told in textbooks: "CTSS (1961) → Multics (1964) → Unix (1969) → Linux (1991) → Docker (2013) → Kubernetes (2015) → AWS (2006)." In that story, there’s no room for a calculator.
But if you look at the Wang 300-SE from 1967 through the eyes of a 2026 engineer, you’ll see an agonizingly familiar architecture:
And here’s what I think: the architecture that wins in IT often appears before anyone notices it. The Wang 300-SE arrived in 1966—7 years before the first personal computer (Altair 8800, 1975), 40 years before EC2, 60 years before serverless made this architecture mainstream. And no one paid attention because it was a calculator. A calculator that didn’t fit the story of big computers. A calculator that worked in mortgage offices, not the MIT AI Lab.
And I think that’s the lesson here. The strongest architectural ideas often come not from the center, but from the periphery. From a calculator for insurance agents, not an operating systems lab. From an engineering feat built with discrete transistors, not a conference paper at AFIPS. And in our time—from a DevOps Twitter account, from an insurance company’s R&D department, from someone’s side project on GitHub, not the office of Google’s CTO.
The Wang 300-SE was a calculator that accidentally invented the cloud. And 60 years later, we’re still living in its architecture—we just call it by different names.
P.S. If you’re curious to dig deeper, I can break down separately: