Advent of Computing - Episode 72 - Viatron Topples IBM

Episode Date: December 27, 2021

Viatron's System 21 was the computer of the 1970s! ...At least that's what their marketing claimed. Started in 1967 Viatron was set to be one of the most exciting companies of the coming decade. They were offering a desktop sized computing machine, the System 21, that promised to break IBM's domination of the office. The System 21's heart, the so-called "micro-processor", was slated to be built from cutting edge LSI chips. It could automate data processing, replace bulky IBM hardware, and do away with the punch card. And this marvel could be yours for just $39 a month. Sounds like a good deal, right? Maybe too good. According to some Viatron was strait up stock fraud. Selected sources: http://bitsavers.trailing-edge.com/pdf/viatron/ViatronSystem21Brochure.pdf - 1969 Viatron Brochure http://vintagecomputer.ca/viatron-system-21-model-2111-restoration/ - The beast itself https://archive.org/details/CIA-RDP80-01794R000100200043-2/mode/2up - CIA review of System 21

System 21 is unique. You saw the ads that introduced this low-cost system to prepare, communicate, input, process, retrieve, output, display, and print data, anywhere, anytime. A system that would greatly increase your data processing capabilities. A system that would put data processing within the reach of every business. A system that wasn't too expensive or too complex for people to use. You can rent or buy Viatron System 21 data processing equipment for half, a third, a quarter of what you're presently paying for equipment with similar capabilities.

That has a real call-today kind of vibe to it, huh? That has a real call-today kind of vibe to it, huh? That's from this beautiful, if a little bit pushy, 1969 brochure for the Viatron System 21 thing. It's a little bit unclear from that passage, so let's do an exercise. From what I've given you, what do you think the System 21 is exactly? What is Viatron selling here? Go ahead, pause the podcast, take a minute, and try to figure it out. Back?

All right. Well, if you answered terminal, then you'd be partly correct. Same goes if you said computer. And, well, you'd also win if you said microprocessor. Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 72, Viatron topples IBM. I decided that for the last episode of this year, we're going to be taking a look at a fascinating disaster. Now, this topic sort of fell into my lap while I was working on my recent 6502 coverage. One of my main sources for that episode was an oral history interview of Chuck Petal, conducted by the Computer History Museum.

Now, I use oral histories as sources pretty often. It's a great way to get details about an event from the people who were actually there. You can find out what was important to the actual players, and best of all, it's in their own words. It's a really valuable resource, there's no other way I can put that. But oral histories can be a bit of a double-edged sword. They usually aren't edited, and I find that a lot of computer nerds aren't really the best at public speaking or staying on topic. Interviews can range from really cogent and really just pinpoint accurate descriptions to meandering streams of unrelated stories and broken fragments of consciousness.

Petal's oral history lands somewhere in the middle of that spectrum. The golden lining here is that you stumble upon some interesting leads mixed in with the digressions and side notes. While discussing his time at Motorola, Petal related this anecdote. Quote, he, one of Petal's co-workers, was working on a thing for a company called Viatron, and I don't know if you ever mentioned Viatron in your computer history museum. Viatron is the company that announced they could build intelligent CRTs and lease them for $10 a month. There was a great big stock hype,

and they went to all the semiconductor companies and said, we can't do this unless you do this LSI for us. And people like Motorola said, oh, expletive, we finally found somebody who can do it. And they'd also read all the hype. So this company? Perhaps stock fraud in the industry? This is one of those weird stories that I actually knew nothing about and I just had to look into. Petal only gives the barest outlines and the details that are provided are a little bit sloppy here. Viatron was proposing to lease intelligent terminals for $39 a month, so a little bit more pricey, but still a great bargain. They claimed their ultimate goal was to quote-unquote topple IBM to get more Viatron hardware in circulation than the biggest computer manufacturer

in the world. That's a little more wild, sounds like a computer on every desk kind of promise. But here's what finally hooked me all the way. Viatron was calling their cheap machine a micro-processor. This may have been the first time that phrase was used. So what time period are we talking here? Well, would you believe that Viatron was founded in 1967? That's, depending on how you cut it, three or four years before the first commercial microprocessor as we know it. When I got to this point, I had one crucial question. Obviously, Viatron didn't exactly topple IBM. Not even close. I like to think I have a pretty well-rounded understanding of computer lore,

and I've never heard of Viatron before this random chance encounter. So Viatron, was it really a stock fraud, or was it just a poorly conceived idea? Was it just a business that went under, or was it a business designed to extract money from unwitting suspects? The central promise of Viatron was to deliver a range of cheap commodity computers and terminals that were easy to install and use. The trick, which was touted in all of their advertisements, was to use LSI, or large-scale integration. That was the newest buzzword.

Basically, it just meant denser and more complex integrated circuits. It's the same technology that would eventually make chips like Intel's 4004 possible, you know, real microprocessors. But by the time System 21 started shipping, there wasn't anything like a microprocessor proper. There were just chips. The bottom line here is that we're entering some really weird territory that I've never come across before. This episode will be looking at Viatron's early rise and their very quick demise. Along the way I want to try and figure out if they were really just a cynical grift or if Viatron's plans were just a little bit too early to succeed. Either way, I think we're in for a bit of a wild ride.

How do you even really start this kind of story? I guess to kick things off, I want to put up a disclaimer first and foremost. The paper trail for Viatron isn't like what I usually deal with. I'm fine working with scarce sourcing, that's a fun adventure. It's like putting together a puzzle with a few missing pieces. But with Viatron, well, it's just a little outside my usual scope. I don't have any modern interviews from major players, which kinda saps out some of the flavor of things. What I do have mainly come from stray mentions like the Chuck Petal interview, newspaper and magazine articles, plus some documentation and marketing material from

Viatron itself. A lot of the articles about Viatron from the contemporary period are mainly from business and financial magazines. Mix in some post-disaster analysis pieces and a handful of more technical write-ups, and we can start to kind of form a picture. But, like I was saying, the distinct lack of interviews kind of sucks here. I wish there was something a little bit more ironclad to get a flavor of things. Anyway, the issue I've run into here is that most contemporary coverage of Viatron either focuses on its early explosion onto the scene or its bankruptcy. Information on its very early beginnings and the people behind the company are a little bit

lacking. The same goes for the period between its wild claims coming public and its inevitable demise. So, when it comes to how Viatron actually survived in the market, I have very few sources to work off. This isn't exactly like a Commodore we're talking about. That said, one name does come up in the same breath as Viatron almost always. Dr. Ed Bennett. Now, just to be 100% clear, the good doctor seems to hold no relation to Tom Bennett of 6800 fame. No, Ed Bennett is a much more enigmatic character. Oddly enough, he was not a doctor of computer science.

The first graduate degrees in Comp Sci weren't awarded until 1965, and by then, Bennett was already into his active years. Instead, he held a degree in psychology. This may not be as weird as it initially sounds, though. The earliest paper trail shows Bennett working on user interfaces in connection with the SAGE project. Some researchers doing similar work also had degrees in psychology, so there is a precedent here. In early 1960, our enigmatic doctor was working on an equally enigmatic project called ASOP, or an evolutionary system for online planning. I've also seen it abbreviated as an experimental system for online planning. I've also seen it abbreviated as an experimental system for online planning. In the first strange twist of many, this takes us back to my beloved hypertext. Well, kinda at least. ASOP was a program designed for traversing information in a database.

From a 1965 progress report on the project, quote, That's a bit of extra flowery language, but makes for a reasonable summary. ASOP was meant to be used by a wide range of operators, from technicians up to higher level managers. by a wide range of operators, from technicians up to higher-level managers. The core problem ASOP was trying to solve was that of data access. This was initially in the context of Sage, but over time ASOP became more generalized. Now, I mentioned the S-word. Sage is something of a specter lurking in the background here, and it must be at least winked at.

something of a specter lurking in the background here, and it must be at least winked at. Put as simply as possible, SAGE was an early defense network developed by the US Air Force. It was huge, complicated, and pioneered a laundry list of new technology. This ranged from, well, networking, it is a defense network, all the way down to improved user interfaces. defense network, all the way down to improved user interfaces. I'm working off two main reports here, one from 1965 and the other one from 1968, so I don't have super finely detailed chronologies to work with. Best as I can tell, ASAP grew out of a need for better data handling within the larger SAGE project. Most data on SAGE was related to air traffic monitoring. The network was designed to keep tabs on essentially anything in American airspace, and in practice that worked out to something like a really big, really dynamic database. Of course, that had to be coupled

with static information about air carriers, airports, you know, general intelligence data points. The ASOP reports show example screenshots from the system which list air traffic information, so that's why I'm making this leap in logic. I don't think it's too much of a chasm to jump. If my train of thought here is right, then what we're seeing is another example of the information problem at work. SAGE had a whole lot of data to deal with, too much for us flesh folk to process alone, so Aesop was designed to help us digest the information stream. Sage already had its own solutions for handling data, so Aesop appears

to have been more of an experiment than an actual planned replacement. To just further complicate things just a little bit more, I gotta mention who was actually working on ASAP. SAGE was an Air Force project, since, you know, it's air defense. They're kind of the air defense people in the states. That makes sense. Actual development of SAGE was handled by MIT's Lincoln Lab. IBM also helped, as well as a number of other non-governmental agencies. In 1958, part of Lincoln Lab spun off to form the MITRE Corp. That's spelled M-I-T-R-E. It's one of those public-private partnership research outfits that we hear so much about. ASOP was a MITRE Corp. project. Bennett was a MITRE employee.

So once again, this is part of Sage, but it's not affiliated in exactly a straight line. Okay, so that's the weird admin side of things. What was ASOP actually? What was it in practice? In general, ASOP was a database management interface designed to be used with a light pin. The main features of this interface were, once again, inherited from Sage, and the Sage terminals that were being used were really cool. They looked straight out of a sci-fi movie. Each had a large circular screen that looked something like a radar display.

It was used via a light gun. To select something on the screen, the operator moved the gun to point at the intended item, and then quite literally pulled a trigger. To complete the look, these terminals also had a number of lighted buttons, sometimes a rotary telephone was built in, and some even had ashtrays. rotary telephone was built in, and some even had ashtrays. So, you know, definitely Cold War era sci-fi at its best. Now, do you notice something missing in that list? There's no keyboard here. The terminals used in SAGE were designed primarily for viewing and tagging incoming radar data. These aren't the kind of things that you would program on, at least not in a normal sense. They were part of a command and control system. In that capacity,

a SAGE terminal was a window into intelligence data, not necessarily what we'd traditionally think of as a computer terminal. The ASOP project was trying to add another layer of complexity on top of that system. The simplest way I can think to describe ASOP is a souped-up spreadsheet editor. It was able to take a data source, presumably from somewhere in the depths of Sage, and present it as a grid of rows and columns. From there, a user could manipulate their view, edit data points, generate reports, and presumably do some kind of simple collation of data. So far, that's pretty mundane.

What's interesting is the interface, more specifically, how ASAP presented data to the user. This was all designed for a pretty general audience. Reports emphasize how ASAP was meant to be used by those without much computer experience. This meant that a lot of work went into the user interface. Accessibility was really the name of the game here. But crucially, this all had to be accomplished without a keyboard. Everything in ASOP was driven by point and click, or depending on if you use light guns, point and shoot. Most of the screen was taken up by the actual tabular data you were viewing, with a menu on the rightmost

margin. Each file in Aesop could have multiple tables of different formats, with each table contained in a so-called section. So the menu had an item for next and previous section. You could also have tables that were larger than the display. These were broken up into pages with appropriate menu items available. The tables had a neat twist to them. Tables could be connected into a tree structure with each line item linking out to another table. For instance, if you're viewing a table that showed the number of airports in each state, you could then link to a table of airports for its respective state with a simple click. These connections were followed using a display line command from the menu, so not exactly

point and click, more like point, click, and then point, click again. The language here is very important because, well, this is starting to all sound close to hypertext, and I want to be totally clear that Aesop has some features of hypertext, but that's as far as the connection goes. This is much more of a database system than anything else. There are definitely similar shades, but Aesop doesn't have the generality of a full-on hypertext system. The whole display line feature here is closer to a database pivot than an actual link. ASOP is definitely dealing with similar problems that, say, the team working on NLS dealt with,

but Bennett and his team reached a pretty different solution. The other core component of ASOP's interface was a series of tree-based menus. Trees were big all throughout the system. When you wanted to run a command, like accessing a file for instance, you would be dropped down to this command tree. This would list the current command at the top, and all the subcommands and sub-subcommands and arguments and so on down below. Commands were built by going down a branch of the tree. For example, if you wanted to try and open a file, then you would start with git. From there, you'd select

file name and then be prompted for a name, then you would choose display to open the file and, optionally, which page to display. AOP also included a programming environment that followed the same tree-based approach. A series of commands and operations would be entered via tree, thus allowing you to build up a program for manipulating data. The system even had a fun punny name. It was called OakTree. The real meat and potatoes of ASOP was, of course, generating high-level reports. It was part of a larger intelligence system, after all. The project would continue until at least 1968. That's

the latest report I could find. Either by chance or by design, this year corresponds to Ed Bennett's departure from MITRE. Viatron was founded in 1967 by Dr. Joseph Spiegel, one of Bennett's colleagues on the ASOP project. Very shortly thereafter, Bennett joined as president of the company. Now, Viatron was founded explicitly to turn Project ASOP into a commercial product. The main offices of this new company opened directly across the street from MITRE. Bennett and Spiegel would start off their endeavor by siphoning off as many MITRE employees as possible. So, from the start, maybe this was a bit of a scummy way to do things. Despite MITRE being a private sector company, ASOP was developed in conjunction with a tax-funded project. So maybe turning a publicly

funded project into a private, profit-seeking endeavor isn't the coolest move to kick things off. So the initial business plan was to adapt ASOP, but in such a manner that the hardware could be made available at rock-bottom prices. At least, that's what most post-mortem coverage of Viatron claims. If you look at contemporary coverage, contemporary to the early days of Viatron, a slightly different picture comes out. A 1968 article in Computer World makes it clear that Viatron's initial offering, called the System 21, would be a lot closer to a basic, very bare-bones data entry machine. The article describes the terminal more in terms of punch-card technology than high-flying military intelligence systems. At this point, the System 21 was described as three main parts,

a screen, a keyboard, and a so-called microprocessor. As an aside, the name System 21 is no doubt a reference to the fact that Viatron was offering a whole office system. Each component of that system had a model number that started with 21. Anyway, the screen and keyboard aren't much to write home about. The microprocessor is really the heart of the operation. We aren't talking about a computer-on-a-chip kind of microprocessor. Let's be very, very clear about that. I think Viatron was using this term because, well, the System 21 was built around a small

data processing unit. Micro was prefixed to emphasize the relatively small size of the system. And it definitely is small. The prototype shown in articles from around 1968 looks like a beefier home microcomputer. Not tiny by today's standards, but a vast improvement over contemporary options. It can fit on a desk no problem. In computer world, it's explained in no uncertain terms that the System 21 wasn't really a computer. I wouldn't even call this a dirty little secret.

It was being sold, first and foremost, as a replacement for punch card based office equipment. The article describes the System 21 as running a single, quote-unquote, microprogram that an operator can use for reading, editing, and searching data. The system uses a pair of cassette tapes to store information specified as punch card images. Each cassette could hold a few thousand punchcard images, and with some external hardware, it was possible to read in punchcards or punch out tape data.

Plus, you got a modem to use the System21 as a normal terminal, so that's a neat addition. So you may be asking yourself, this sounds great, but how much does it cost? This has to cost a fortune! Well, dear listener, you'd be mistaken. Keyboard, monitor, and microprocessor were all rented as separate line items. That's right, you couldn't buy anything from Viatron, at least not initially.

The three-piece starting system came in at just $39 a month. An inflation calculation shows that's roughly $311 today, so outfitting an office with System 21s wouldn't exactly be cheap, but it was a lot more affordable than the competition. So who, you may ask, was Viatron's competition? Bennett and Spiegel were really clear about this one. IBM. Specifically, Viatron was trying to replace IBM's aging punch card-based office systems. I think this should be obvious, but from the start, there was just no chance of that happening. I've seen estimates that IBM had something like 80 plus percent market share around this time. Computing, and punch cards very specifically, were an IBM business. Especially when it came to boring

back-office stuff. What's more, IBM had kept an iron grip on this sector for decades and decades. Even before computers existed, IBM was the main name in business automation. Man, that sentence really sounds boring, but it's true. IBM had captured the market with punch card tabulators and other associated machines way back in the early 20th century. Ruthless business practices, some of which were downright illegal or would soon become illegal, combined with ever-advancing technology, carved out a dominant position for Big Blue. It really was ever-forward for IBM for decades.

I'd love to say that by 1968, computers had changed the landscape of the world, that no one was using punch cards anymore, but I'm sure there were stragglers. I know there were stragglers. Punch cards would never go away in this period. Some companies were still living off punch card tabulators. Most larger offices would either be using a slightly less outdated IBM 1401 or a flashy new System 360. These mainframes were the beating heart of business operations. Payroll, accounting, business intelligence, statistics, and reports all had to flow through one of these digital beasts. But that flow wasn't always easy, and that was Viatron's first angle of attack. When it came to dealing with bulk data on an IBM mainframe, you really only had one option, and that was still the punch card.

You would have a gaggle of clerical workers sit down at their card punches, type in data, and slowly build up a stack of cards. Eventually, those could be fed into the mainframe, churned out by some software, and eventually crunched into whatever output you wanted. by some software and eventually crunched into whatever output you wanted. That would then be punched onto a new deck of cards for later use, stored on magnetic tape, or printed. No matter what, punch cards were still a core part of day-to-day business operations. It was a means of data transfer. By 1968, this was a pretty archaic technology. Punch cards were verging on 80 years old at this point. They were oddly suited for use with simple computations and similar computing devices, but the bottom line was that punch cards were starting to outlive their usefulness. Cardstock just isn't user-friendly, especially when you're dealing with a lot of data.

There's also the cost aspect. IBM rented hardware, which was already expensive enough. You'd need the computer to start with, then at least one fancy automatic card punch or key punch, plus a printer, plus whatever add-ons your particular office needed. That racked up to big monthly bills. One of their smaller systems on offer, the 1440, cost $2,600 a month in this period. These machines also burned through punch cards like fuel, and you couldn't rent those, you had to buy cards by the carton. That was Viatron's second angle of attack. Putting this all together, we get a small hole in IBM's glistening blue armor. Well, discounting the larger antitrust holes, but that's another matter.

This is what Viatron was trying to exploit. Do away with punch cards, offer a better and easier user experience, simplify things down, and do it all cheaply. Sounds simple. Maybe possible. But here's where things get into a weird split that I can't fully understand, that I'm still having some troubles with. Early coverage of the System 21 makes it sound like a fancy digital card punch. There's the implication that it could also do simple data processing. So a System 21 machine could replace an IBM card punch, maybe a key punch, a tabulator, a printer, and thousands upon thousands of punch cards for just $39 a month. It may also be able to, at least for smaller offices, do away with the need for a fully-fledged computer.

For larger offices, the terminal capabilities of the System 21 would allow it to slot in with existing hardware. This is one of those times where we have to try and pack away our 21st century brain. We know how computing shapes up, and I think that makes early System 21 seem underwhelming. But if we look at this package in context, I think Viatron was starting to tap into something big. Take just the terminal aspect for starters. The first CRT-based terminals, called Glass Teletypes, started coming out in the very late 1960s.

terminals called glass teletypes, started coming out in the very late 1960s. There were earlier similar devices, but I'm talking here fully compatible simple character-by-character terminals. The Datapoint 3300 went on sale in 1967. That was just a terminal with a glass screen. Simple. Just a year after that, Viatron announces their intent to beat that feature set and add in so much more. And it's cheap. I haven't seen this emblazoned across any advertising material, but Viatron's machine was getting pretty darn close to a personal computer. In 1968, it wasn't being touted as a computer, but, I mean, come on. It's a data processing box that a CRT sits on top of. You have a keyboard out front that you type on.

It even has cassette tape-based storage. Everything is small enough to fit on your desk. It's only operated by a single user. To my ear, that sounds a lot like the appliance computer that Steve Jobs liked to talk about. This is getting dangerously close to future technology. And to put a cherry on top, the System 21 came packed with a new buzzword. MOS, or Metal Oxide Silicon. This is the technology that currently enables us to pack millions of transistors onto a chip. It first became viable in 1962. That year, researchers at RCA were able

to fit about 16 transistors onto a chip. From there, progress was inevitable. By the late 60s, it looked like MOS chips were about to explode in complexity. There is even a word for this soon-to-be technology, LSI, or large-scale integration. Soon, thousands of transistors could be packed into a single slice of silicon. What once took multiple circuit boards full of chips could be done with a single chip. In theory, this would drive down costs of computing hardware significantly. The issue was that LSI was still a little ways off when Viatron started their marketing blitz. No computers were using LSI chips. For that matter, no semiconductor manufacturers were selling LSI chips.

It was very much an unknown quantity, and Viatron was able to spin that to their advantage. Bennett's argument in interviews and Viatron's official pitch was that they would ride the LSI wave to make their improbably cheap machine a reality. So let's round up Viatron's overall plan, shall we? Use LSI technology, which does not exist in 1968, to make a microprocessor system for every office. It will be cheap enough and powerful enough somewhat nebulously defined collection of modules that, at least in 1968, sound a lot like a direct replacement for punch card systems. But how do they get there? What's the path forward? Easy. Lots and lots of money.

For any startup, securing funding is key, especially early on. This was very true for Viatron, thanks to a little quirk of their business plan. You see, Viatron, they weren't planning to make any money for quite some time. From a 1968 article in Business Automation, quote, however, Dr. Spiegel's enthusiasm was not diminished. Quote, We expect to lose money, he admitted, until 1970 when we should start moving into the black. The reason he believes Viatron will stay in the red until that time is the system is being rented only. End quote.

This whole rent instead of buy thing meant Viatron would not get immediate returns, even when machines shipped. The tradeoff was that eventually revenue would catch up. You could also say, oh, the rental is only $39 a month. That can't be overlooked. But there is another complication here. Viatron was banking on economies of scale to help drive down manufacturing costs. By making large numbers of System 21 machines, Viatron could get cheaper bulk rates on parts. There could be more savings from automation.

Shipping could be combined to make more profitable deliveries. But that all hinged on large amounts of upfront capital to weather the lean times that were coming. The result here is that Viatron had to look like a good investment. That was the only way it could work. According to that same Business Automation article, Spiegel projected at least 9,000 System 21s would ship in 1969. That checked one big box off. Economies of scale would kick in, large orders would be placed, and investors would see growth. In 1971, the magazine Datamation ran an article titled The Rise and Fall of Viatron by David Gardner. Spoilers. Anyway, here I'm pulling heavily from that piece, since in addition to

the eventual collapse, it also covers the crucial funding phase, and perhaps the biggest driving force behind this phase was Bennett. Quote, In December of 1969, Viatron sold $25 million of convertible debitures. All told, Bennett raised about $40 million for Viatron sold $25 million of convertible debitures. All told, Bennett raised about $40 million for Viatron. End quote. A large source of this funding was backed by these convertible debitures. These are basically contracts that can be converted to stock once the company goes public. The idea here being that investors can get on below the ground floor,

getting stock in a promising company at a very low price, and then when things go public, they turn a profit. The two major hopes here being that the company goes public, and that they don't immediately fold. Investors flocked to Viatron. It seemed that early on this was from genuine interest. Their business plan was exciting. What's not to like about it? You have a group of researchers from a well-known lab striking out into the fully private sector. You have new and promising technology just on the cusp of what's possible. and you have the promise of doing battle with Big Blue itself. This was looking like disruptive technology. Spiegel mentioned in 1968 that trade show booths were swamped with interested parties. As capital flowed in, things start to take on a slightly different demeanor. Gardner's article makes it sound more like a confidence game than an actual investment

opportunity or a business. Bennett was able to get large investment firms to buy into Viatron. Convertible debitors were selling at just 50 cents a share at this point. It must have sounded like a really good price to get in at. Once a few big names were signed off, Bennett used that clout to get more investors. The other tactic employed by Viatron was the letter of intent. These were written promises from companies saying that they would order X number of System 21s once systems could be delivered. To be 100% clear, these letters were not orders, and they were not contractual. They weren't paid for. They were just, hey, we'd like a few hundred computers once your factory starts going. Investors like that kind of stuff. It shows that there will be cash flow coming soon.

Bennett was able to use these letters to get more investing in Viatron. First, millions of dollars came in. Then tens of millions. It seemed that Viatron would have all the funding it could dream of. The company was becoming, in part, an investment vehicle. This is also where we start to see mission creep. The initial business plan called for Viatron to do none of the actual manufacturing. That was going to be done by outside companies, maybe as contractors, maybe as licensees, it's not entirely clear. That got thrown out the window pretty quickly. Viatron opened a manufacturing plant in Bedford,

Massachusetts. According to Gardner, the startup also opened offices or plants in, quote, Indianapolis, Long Island, Birmingham, Japan, and Hong Kong. Now, Viatron had always had this air of grandiosity around them, but it seemed to accelerate as millions of dollars came in. In 1968, Spiegel was claiming that Viatron would be the Sears and Roebuck of computers, that the company would fill entire catalogs with new product offerings just waiting to be ordered. Viatron also called it the Ford or Chrysler or GM of computing, with allusions to reliability, accessibility, and mass production.

Gardner picks out one of Bennett's best quotes to illustrate this grandeur. You can start small, and you can stay small, and you can die small. Or you can start big, and you can grow bigger, and you can die, but you die a little bigger than you would have otherwise. Now, here's what messes me up. Viatron keeps sounding really close to something revolutionary. This is the 21st century viewpoint pulling me in a weird direction, I think. I know that computing exploded once small, cheap, desktop-sized machines entered the market. I know that mass production, economies of scale, and LSI made that possible. I also know that these pieces didn't come together until the later part of the 1970s.

And here's the part that confuses me. Viatron seems to be wrapping itself in the flag of revolution. The marketing and planning show every piece of the puzzle that would eventually spur the microcomputer boom. If Viatron lived up to their own hype, then yeah, they probably could topple IBM. They could at least put up a good fight. But more and more, I think the hype outpaced what Viatron was capable of. Bennett wanted to start big, but I don't think that was a good thing in this case.

Viatron would open multiple offices, factories, and facilities around the world. That wasn't part of the initial business plan. It was a result of trying to start off big. I think the most dangerous result, at least in retrospect, is that Viatron opened up their own semiconductor plant. Why? Easy. They needed to secure the supply chain. System 21 machines would require MOS LSI chips, a technology that was still in its infancy if it was even possible to manufacture. Manufacturing their own chips would go a long way towards keeping the supply flowing.

Remember that Viatron was projecting the shipment of thousands of System 21s every year. Possibly as many as 9,000 in 1969 alone. And I hesitate because that's a ridiculous quote. If each machine had just five custom chips in it, and that's a low estimate, we're looking at 45,000 LSI chips, all custom, all using this bleeding-edge technology that no one's mass-producing yet. The secondary effect was that Viatron was also able to use this factory to leverage other SimCos. Gardner estimates that Viatron would eventually sign off on at least $70 million worth of contracts for custom and off-the-shelf chips from third-party manufacturers. This is where the Motorola connection with Chuck Petal comes in.

Now, this part isn't explicitly backed up with sourcing, it's more of an implication. Since Viatron was manufacturing some of their own chips, at least in theory, they would be able to pull another confidence trick. A Viatron employee could show up at, let's say, Motorola to put in a request. They offer a contract for some custom LSI chips, big beasts with thousands of transistors per packing. Maybe Motorola initially turns down the order. It would be too costly to manufacture. They just weren't doing LSI at these large scales yet. The Viatron rep is able to say, oh well, we're already making these chips, we just need a second source, it's no problem for us to manufacture them, surely you can do it. Motorola is now afraid that they're being beaten to LSI and agree to the contract at possibly great cost to themselves. Viatron can then repeat this with another Simcoe,

but this time they can say, oh well, Motorola already signed a contract as a second source. That further feeds into the investing side of things, since now Viatron can point to a stack of contracts for LSI chips that haven't been made yet. You see what I'm getting at? This is an economic snowball effect going on here. Viatron is able to ratchet itself up to be a larger and larger company. At every step, it's able to throw around more weight and eventually get more funding. This could be fine. It's how a company grows and expands with success. Except this is all happening in a very short time period without actual sales or actual product to back it up. Viatron was running off hype and speculative investment alone. So would Viatron be able to

live up to the hype? Would they be able to turn all this funding and all this press into something real? Would they topple IBM? I think we can already guess the answer. It's a little hard to get a bead on when the first System 21s saw the light of day. Most of the press coverage is about the company and the systems were initially rented, so I think a big release date got lost in the shuffle. My conservative estimate is December of 1969. This guess comes from some of the marketing materials that listed a delivery schedule along with later ads calling the System 21 the, quote, computer of the 1970s. So it had to be around the 1970s, I know that, and the list says December of 69, that just

sounds reasonable. Viatron didn't take the money and run, at least not right away. They actually produced honest-to-goodness machines. So was the final product anything like the hype set it out to be? This gets annoying and complicated pretty quickly. Remember how I said the System 21 was a really wide product range, this is where that starts to matter. One 1969 brochure lists 11 possible microprocessor models of varying sizes with varying add-on features, 7 possible loadouts for displays, 3 different keyboards, 6 optional tape devices are also listed. That's all before getting into the optional I.O. choices. System 21 was intended to be a modular system, which is a great idea in practice. Mixing and matching components can really open up a world of new opportunities. Just

don't do it like this. Each loadout and feature package came with its own model number. A 2101 was your basic microprocessor. Ah, but the 2111-101 was a larger system that could support multiple tape outputs. It all gets a little confusing since, instead of talking in terms of spec, you get model numbers. The 2101 had 512 words of RAM, but the 2111 had 1024 words, but you can't tell the difference from the model numbers. They're just meaningless numbers. I can see what they were trying to do here. The idea was that machines were easily swappable when parts broke. Viatron was trying to operate on a replace-not-repair type of system. Instead of sending out a technician when a System 21 broke down, an office would grab a spare from storage,

then send in the broken machine to be replaced. All data was kept on removable tape storage, and machines with the same model number were 100% interchangeable. But once you get to the point where you have more than a few possible configurations, that can get annoying. Also not helping this is the fact that the replace-not-repair system breaks down within the first few months of operation. With that said, let's see what Viatron was actually shipping. To start with, we get the 2101 and the 2111. These are the so-called microprocessors, the desktop terminals that Viatron announced way back when the company formed. For best-case results, I'm going to be focusing us in on a fully-loaded 2111. That should give us an idea for the top-of-the-line System 21 that could fit on a desk. By the end of 1969, System 21 was no longer

a punch-card machine replacement. At least, it wasn't being advertised explicitly as that. It was a fully-fledged computer. Or something close to one. There's going to be a lot of caveats here. It was also a bit bigger than the prototype that was being shown off in 1968. That's not to say the System 21 prototype was sleek or anything. It was a nice and chunky machine, but the production model was kind of awkward looking. I'll post some photos around, but I think awkward's just the best way I can describe it. It still had the detachable keyboard and CRT, and those components looked normal. It's the microprocessor box that's a bit strange. It's stepped so that there's a shorter front section that contains two cassette drives.

Then the case gets taller in back, making a shelf for the CRT to rest on, but also ample space for a control panel. That's right, this machine had a keyboard and a weird front panel. Keep in mind that Viatron was marketing this as a computer for everyone, from technicians to C-level management. But hey, we'll get back to the weird interface in a bit. Luckily, we have plenty of technical details to go off. Viatron's brochures come complete with pretty extensive technical descriptions. In addition, a surprising number of documentation has been preserved. We're talking down to the schematic level for some devices. At this point, we're into the good part of the paper trail, so at least there's that. Reading through the operations manual for the 2111,

it's clear to see the impacts of ASOP. Sure, we aren't dealing with some cutting-edge user interface, but the System 21 terminal is built around very simplistic database control. Viatron docs describe the desktop machine in terms of data flow, which is something I can get behind. It came packed with multiple inputs and outputs. These range from keyboard to tape, printers to card punches, and a set of generic data channels. Operators would first program the terminal. Biotron's docs call it programming, but really it's more like setting up a digital form. You would lay out the format for records, call it a schema, placing fixed-length fields within a 40x20 grid. Once set up, that format was displayed on the lower half of the screen, with the actual

data displayed on the upper half. From there, you could start dealing with your numbers. To start with, you'd probably need to manually enter your dataset. That meant a lot of typing. As records were created, the System 21 would need some place to store them, and this is where the data flow part comes in. The terminal's front panel, as awkward and strange as it may be, actually seems to make a lot of sense in this very specific context.

Right in the center, below a large grid of indicator lights, was a set of dials for directing inputs and outputs. Want to save your records to tape? Just set it to output via one of the tape drives. I.O. devices were handled in a pretty savvy way. This gets to the real modularity of System 21, which is a touch I really like. Internally, the terminal was built around a central bus. The bus itself was passive, at least as near as I can tell. It was just a bunch of connected slots for cards to be dropped into. The processor lived on a few cards, a set of cards handled ROM and RAM, same for keyboard input and video output. All these necessary cards for a healthy and usable terminal were tucked into the rightmost side of this bus.

The left side of the bus board was left open for optional expansion cards. This is where you would add new I.O. devices. You could get a card to connect up to a printer, cards for multiple tape drives, or a card that would connect up to punch card equipment. The System 21 has a neat trick to support this expandability. In general, configuration options were set using dials and toggles on the terminal's strange front panel.

In a basic model, the entire left-hand side of the panel was blank, filled up with metal covers. Each I.O. card you added would replace one of these metal spacers, providing a new set of knobs and switches for the front panel. I think there are some definite pros and cons to this approach. I mean, I get it. The System 21 terminal wasn't advanced enough to have software-based configuration tools. That would take some type of fast storage for config files, software for editing settings, plus some way to extend that software when expansion cards

were added. It's a lot of complexity. Just throw on a few more buttons and you've saved a lot of work. Plus, users don't have to learn some new configuration utility this way. Just flip a labeled switch, maybe read a cheat sheet once or twice, and you're golden. The big downside, at least what seems like a downside to me, is that a fully loaded terminal has a pretty complicated looking array of inputs. A fully loaded front panel has six dials, 16 toggle switches, and a big grid of indicator lights. The most egregious part of this panel is for controlling the printing robot. And I mean, this device in general is just ridiculous.

For whatever reason, Viatron, for some reason that I will never fathom, didn't want to supply a printer or a connection to some existing computer printer. Instead, they offered the printing robot. It's a slab of linear actuators that fit on top of the keyboard of an IBM Selectric typewriter. Specifically, an IBM Selectric. specifically, an IBM Selectric. Outbound data was printed as the robot struck keys with its little actuated fingers. Sure, cool solution, but I just have so many questions. The front panel module for the robot was, well, it was complicated. Most of this control panel is devoted to a tabulation pinboard. At least, that's what the manual calls it. It's a 4x20 grid of holes, each hole corresponding to one location of the System 21's internal records. An operator places a peg in these holes to tell the printing

robot to type a tab at that location. That's it. And I just don't like it. I don't understand why the terminal couldn't automatically format for tabs. Plus, this interface is just awful. Since the panel modules are vertically oriented, the pin board is also vertical, while the CRT displays records horizontally. So you have to crane your neck to line up the holes with the letters. It's just another odd choice. There's

something viscerally wrong about the printing robot. One saving grace here is that Viatron was using MOS chips. I'm not sure if they were technically LSI, judging by the circuits, I think they were pretty complicated ICs no matter what. But hey, let's give them the benefit of the doubt and say sure, LSI chips. Their terminals came packed with these custom chips. From the looks of it, we're talking a few dozen of the little dudes. I know most of these are custom because I couldn't find any data sheets, so a lot of these chips were probably a very limited-run type of affair. Viatron also made their own general-purpose computers, something like minicomputers.

They came in two sizes, used magnetic core memory for some reason, and carry on the strangeness we've already seen in Viatron's terminals. strangeness we've already seen in Viatron's terminals. This is still pre-microprocessor stuff, so everything is custom and pretty unique at that. These machines had two separate arithmetic logic units, each with their own registers and own set of operations. Without getting too deep into the details, I just want to point out that that's not normal. These systems don't play a big role in the overall story, but I might come back to them some other time. Besides ads and manuals, I can't find information on if these general-purpose computers were ever even used outside of Viatron's labs.

So, it's an interesting side note to keep in mind for one day. That's basically all of Viatron's System 21 offerings. It sounds like a mixed bag at best. There are some interesting touches, some very compelling features, and some definite downsides. But hey, I don't think Viatron was trying to sell me in particular on these machines. The more important question is, did consumers really think that System 21 was the computer of the 1970s? And the answer is no. No, they did not. What's so great here is I actually have a fantastic contemporary source to draw from. You see, a little outfit called the CIA

looked into leasing Viatron hardware in early 1970. There's a paper trail covering the whole debacle, including some pretty systematic analysis of System 21's usefulness. Best of all, these memos and reports were declassified in 2002. I think this makes for a wonderful case study, since governmental contracts would have been a key battleground. IBM made a huge chunk of its money off federal contracts. If Viatron could sneak into that market, then solvency would only be a few signatures away. It also sounds like the System 21 was exciting to federal bureaucrats. A $39 a month system could go a long way towards helping balance budgets. But when the CIA got their hands on some hardware, they were unimpressed. The core problem came down to use cases. The CIA wanted to use a

number of System 21 terminals to replace older IBM key punch machines. The newer Viatron devices would be used to enter data, store it on tape, and then eventually send the data to an automated card punch. In this capacity, Viatron wasn't replacing IBM hardware. Punch cards would still be used for the all-important jump into the mainframes. Going over these memos was a bit of a no-duh kind of moment for me. I've been trying to drink the Viatron Kool-Aid at least a little to try and see where they're coming from. I think it's important to see where all these ideas are coming from, but this is where everything fully unraveled. Going by the raw numbers, we know that IBM hardware was in basically every office in America. Offices weren't going to just suddenly drop all

their existing hardware to replace it with Viatron. This mixed-use case was much more likely. Computations would still be done on IBM hardware. Punch cards would still mediate data transfer. System 21 terminals would just slot in for data entry. And that's the best-case scenario. So is this a disruptive technology? No. No, it is not. Best case, Viatron offers some quality-of-life improvements.

At least, that's what the CIA was hoping for. They rented three terminals and tried them out for about six months. In general, the results were pretty negative. But let's go over the upsides first. Maybe there'll be some good here. The memo points out that System 21 machines are a lot quieter than IBM key punches. So hey, that's a plus for improving the office environment. Since data is stored digitally, it's easier to fix errors. You don't have to punch new cards if you make a mistake. That can save some time and money, so another small improvement. The System 21 can also

be used to validate data. As near as I can tell, this is just making sure fields and records are the right size and right data type. And finally, these machines can store records in less physical space than punch cards. Each cassette, called a Viatape, but they're actually just normal cassettes, could hold around 400 cards worth of data. It's worth noting that in 1968, Viatron was estimating thousands of cards worth of data per tape, but whatever. That's a minor letdown. And that's also it. Those are all the upsides of the CIA lists.

None of these are very disruptive. It's just small improvements to existing technology. It's very much evolutionary instead of revolutionary. The memos make mention that Viatron's automation features may be useful, but that the CIA has yet to find a specific use for them. Workflows were already established, namely that punch cards get fed into computers. Fully leveraging the System 21's fancy data flow features could be beneficial, but it would take some major retooling. The list of downsides is a good deal longer, so I'm going to pull out the relevant issues.

The first one that I find interesting is space. Like I said, Viatron hardware would have to be in concert with existing office setups. The System 21s were designed to do that. They could connect up to typewriters to print data. They could connect to IBM card punches to punch cards. But that said, you had to have a typewriter and a card punch around to make a System 21 install useful. One of these terminals already dominated a desk. Now add third-party hardware, terminals already dominated a desk. Now add third-party hardware, and you run out of room really quickly. These machines also suffered from reliability issues. The CIA doesn't go into details on exactly what kind of problems they ran into, but they do state that System 21 terminals broke down once a week. The whole ship-in-replacements model never materialized in a

big way, so technicians had to arrive on the scene to fix machines. Factoring in the time to get a technician out there, the memos estimate that Viatron hardware was down 10% of business hours. That's just not okay. In theory, the modularity of System 21 terminals could have helped here, but it looks like that didn't work out in practice. Another big point of contention was the relative lack of flexibility. By 1970, Viatron had toned down the whole punch card aspect of their machine, but it was still there. System 21s were built around 80-column data, aka a standard IBM punch card. They were also built around the assumption that a record would fit on a single card. Both of these were bad assumptions. They were wrong assumptions.

The CIA memos point out that it was very common in their office for data to span cards. These would be punched one after the other. My guess is that this would work something like one card for a person's name and address and another for, say, banking records. Since workflows would often go from a long handwritten form to a set of cards, it made sense to punch different cards out one after the other. You couldn't really do that with the System 21. It was, as near as I can tell, one format per tape. At least the CIA wasn't able to find a convenient

way to fit their existing workflows into the System 21. Many offices also used non-standard punch cards. 80-column was kind of the default, but it was only one of many options. These memos specifically point out the use of aperture cards. These are punch cards with embedded slots for microfilm. These were the same physical size as an 80-column card, but had a number of columns taken up by film. They were also often pre-punched with a little bit of information about that film. System 21s could not deal with that. They were all 80-column all the time. They were all 80 column all the time. The final kicker was that by 1970, Viatron was already entering its death throes.

From the end of one CIA memo, quote, On the basis of the evidence summarized above, and taking into consideration also several indications that Viatron may not be able to stay in business for many more months, we decided against acquiring any Viatron equipment for long-term use. Signed, Redacted. That's a big problem, perhaps bigger than you might realize. I'll cut you in on a bit of an open secret in the computer industry. First-party support will make or break a system. Especially in this case, where System 21s broke down once a week. That meant without Viatron itself sending out technicians,

your install would only last one week. That is not a viable solution to anything. This decline and fall was not a long time coming. By this point, Viatron was only three or so years old. But there had been plenty of warning signs along the way. Take Viatron's massive expansion, for instance. As soon as funding rolled in, earlier plans went out the window. Manufacturing came in-house, which meant Viatron needed more facilities and more employees and more know-how. more facilities and more employees and more know-how.

Overhead was increasing quickly. There was also the matter of unsustainable hype. Viatron went public on the stock market in either late 1969 or early 1970. It's hard to pin down the exact date, as with a lot of Viatron events. Debitures had been sold for 50 cents prior to IPO. Sometime in 1970, Viatron stock would be valued over $60 a share. That is a huge jump. That's something like an 11,000% increase. That kind of growth is not sustainable.

I don't care how cool your terminal is, and I don't care if you're going to topple IBM in six months. From Gardner, quote, Viatron reported in February 1970 that it was producing System 21s at a rate of 600 a month. Although this was far below the firm's original target of 5,000 to 6,000 units a month, it was still a respectable number. Viatron employees were jubilant over the units that were being shipped out of their Massachusetts plant. But there is evidence that this created an illusion of actual sales when, in fact, many of the systems shipped,

perhaps even the majority, piled up in dealers' warehouses. One dealer in Texas, for example, reported that he had 350 units in his inventory in a warehouse. End quote. Investors had been drawn to Viatron reported that he had 350 units in his inventory in a warehouse. Investors had been drawn to Viatron based on that 5,000 to 6,000 number, or in some cases, a number around 9,000 units. Actual production was an order of magnitude less. Sales were, well, who even knows what sales were. The bottom line is that Viatron was unable to live up to the hype. Few System 21s were even in circulation. Those that were managed to break

down once a week. In the interim, they impressed no one. An IBM killer, this certainly was not. In July of 1970, Bennett was kicked off the board of directors. Spiegel was straight up fired. According to an article in Computer World, this ousting happened over a dinner meeting in Bennett's own home. No reason was publicly given, but I think we can draw our own conclusions here. Throughout the year, Viatron stock had been dropping. The company had switched from rent-only model to a buy-only model in an attempt to drum up cash flow, but it was already too late. They had already opened offices around the world and opened their own semiconductor manufacturing plant. There wasn't enough cash flow in the world to help them.

In 1971, Viatron declared bankruptcy. From the sound of contemporary reporting, no one was surprised. All right, that brings us to the end of this episode. What's the verdict? Was Viatron a legit attempt to unseat IBM and reshape computing? Was it all a cynical attempt to milk investors for cash? Was it, as Chuck Petal suggests, stock fraud? After spending so much time reading into this, I think Viatron falls somewhere in the middle. It's a mixed bag. I think Viatron started out as a wholehearted attempt to shake up the office environment. Look no further than Aesop. That project took a very revolutionary approach to

data management. I know, once again, that kind of sounds dull, but I can't put it any other way. If Aesop had been adapted into a product, I think Viatron could have given IBM a pretty good fight. But that was not possible. ASOP ran on highly specialized terminals connected to big, expensive mainframes. Those mainframes, by the way, were IBM hardware. Could LSI have been used to get around the issue? Maybe. But the technology was not ready in 1969, or in 1970 for that matter. If Viatron had launched in, say, 1972 or 1974, then things would have been a lot different. For this specific era, a handful of years would have made a huge difference. When you get down to it, the System 21 terminal that shipped wasn't even that good of a machine.

Here's a fun contemporary example. The Datapoint 2200. That terminal was released in 1969. It could connect up to a mainframe and function as a teletype with a fancy glass display. It was also fully programmable. Its architecture would be the basis for the Intel 8008, which eventually twisted and turned its way into the 8086. That's capability and flexibility in a cheap terminal.

Two things that the System 21 lacked. I think the problem came down to the same factors that made Viatron so exciting. Its founders wanted to move quick. They wanted to get big before getting stable. Bennett exemplified this attitude. Viatron promised too much, too soon. An LSI-based machine for $39 a month was not possible, and I think Viatron knew that. Or at least, they figured that out really quickly. They bet on technology that wasn't ready, and they lost. Instead of shutting down, they tried to change plans and keep operating. But these issues were too fundamental to be fixed. That all said, there's definitely something interesting here. Here's one last thing to ruminate on. The System 21, on paper, sounds a lot like a personal computer.

I mean, look, Viatron even calls it a microprocessor. They claimed that better integrated circuits would make the machine cheap, that the system could be churned out in bulk. The form factor, despite some rough edges, even looks vaguely familiar. Change around some words, change around the case, and it could even be a relative of the IBM PC. But the System 21 was a failure. I think this speaks to something fundamental. Call it a digital trends and forces kind of theory, or call it the electrical zeitgeist. Personal computers were coming. We were getting flashes of them before we had the finalized

language that we use today to describe them. Viatron may be a footnote, but it gives us insight into a larger trail being blazed. Thanks for listening to Advent of Computing. I'll be back in two weeks' time with another episode chronicling the history of the computer. Until then, if you like the show, there's a few ways you can help us grow. If you know someone else who'd be interested in computing's past, then why not take a minute you like the show, there's a few ways you can help us grow. If you know someone else who'd be interested in computing's past, then why not take a minute to share the show with them? You can also put reviews up on Apple iTunes and now on Spotify. If you want to be a

super fan, you can support the show directly through Advent of Computing merch or becoming a patron on Patreon. Patrons get access to episodes early, polls through the direction of the show, and bonus content. There's a new episode up on the Plus Plus operator right now, so go on over for a dollar a month, you get that, and a few other bonus episodes. And if you have any comments or suggestions for future episodes, then go ahead and shoot me a tweet. I'm at Advent of Comp on Twitter. And as always, have a great rest of your day.