How a band of UNIVAC hackers, a borrowed plotter, and five long-shot bets became one of the most influential software companies ever built — in the words of the people who were there, and the man who wrote it all down.
In the spring of 1982, someone walked into a bank in Marin County and deposited somewhere between $53,000 and $59,000 in checks — the founders still can’t agree on the exact number. That pile of money was the entire net worth of a newly incorporated software company. Three years later it went public. Within a decade, AutoCAD was the global standard for computer-aided design, running on millions of machines in 80 countries in 18 languages. John Walker, one of its founders, kept writing everything down. He called the result the Autodesk File.
Before there was Autodesk, there was ISD — Information Systems Design, a timesharing utility running on a UNIVAC 1108 out of Oakland, California. The machine was a marvel and a disaster in equal measure: no memory protection, no instruction protection, its operating system’s source code available online to any user. As Dan Drake put it in the 2020 CHM oral history: “It was hackable.”
That hackability is what assembled the founding team. A young programmer named Greg Lutz, working remotely from Long Beach, had crashed the ISD 1108 three times in a row. Drake and Keith Marcelius reviewed the dumps, identified the job, and called Lutz. The curiosity that crashed their machine was exactly what they needed. “Can we take away his return tickets?” was Drake’s verdict at the interview. They hired him on the spot.
John Walker arrived via a different route — recruited from New Jersey after a manager toured US UNIVAC installations to see who was doing interesting things with the new Exec 8 operating system. The UNIVAC user group met twice a year, swapping code and stories. Nearly every Autodesk founder traces their connection back through that network.
In 1982, I used the phrase “the game has changed” to shock people into realising that even then the stakes were rapidly rising and that to build a successful software company would require funds, commitment, professionalism, and risks far in excess of what previously characterised the personal computer business. Sometimes people forget that personal computers were already six years old when the IBM PC was introduced, and that several companies had grown to $10 million per year or more. What had been, in 1977, a game into which anybody with a bright idea and a soldering iron could jump had, by 1982, become a serious business in which millions were made and lost.
The Autodesk File, 5th Ed. — Information Letter 14Walker’s prior company, Marinchip Systems, had been building hardware around the Texas Instruments 9900 processor — a technically interesting 16-bit chip that TI had never properly promoted. By 1981 it was clear the 9900 was a dead end. Walker began making calls. The people he reached were, almost without exception, the same network that had been talking around UNIVAC machines for a decade.
The founding strategy was deliberately anti-strategic. Walker had written in the original Working Paper that 80 percent of new businesses fail, so Autodesk launched five simultaneous projects. Whichever hit, they would drop everything and concentrate. The five: Duff Kurland’s full-screen text editor (AutoScreen), a statistical library, a compiler for the Apple II, Kern Sibbald’s desktop organizer (called “Autodesk” — the product, not yet the company), and the conversion of Mike Riddle’s CAD program Interact into something that could run on an IBM PC.
The product with the company’s name on it — the organizer — never shipped. The CAD program nobody expected to dominate became everything. Walker later called this “management by lack of alternatives.” They built what was achievable, and the market picked the winner.
But what exactly was AutoCAD when it started? Walker’s development log from the Autodesk File answers this with unusual precision. It began as a program called MicroCAD, converted from Riddle’s SPL-language Interact into C. The name changed to AutoCAD on October 11, 1982 — a six-hour renaming job that required changing every module name throughout the codebase. The PAN command direction was reversed that same day because Keith Marcelius said people thought it worked backwards. Feature and naming decisions were landing in the same week.
During the summer and early fall of 1982, I was working furiously on AutoCAD-80, the CP/M-80 version of AutoCAD. At the same time, Dan Drake and Greg Lutz were working on AutoCAD-86 for the IBM (Greg) and the Victor 9000 (Dan). Because AutoCAD-80 started to work earlier, largely because it used intelligent display devices and didn’t require the extensive low-level drivers that the IBM and Victor needed, AutoCAD-80 took the lead on feature implementation through the introduction of the package at COMDEX in November of 1982. What follows are excerpts from the extensive development log that chronicled AutoCAD’s earliest formative stages. From the very start a major theme in AutoCAD development was figuring out how to make it fit in memory. The version described below had to run in a machine with a total memory of 64K bytes, of which only 52K was free for user programs.
The Autodesk File, 5th Ed. — AutoCAD-80 Development LogThe NIH problem was real but manageable. Riddle had written Interact in SPL — a language he invented — whose compiler was itself written in another language he invented. Greg Lutz called this “an extreme case of NIH syndrome.” The solution was to ignore the language entirely. Lutz and Drake split the source files alphabetically and converted every one to C without consulting Riddle, meeting in the middle. The famous “E struct” — a single instance describing every entity in a drawing — was carried forward as-is into C and remained AutoCAD’s foundational data architecture for years.
Six months after incorporating, Autodesk took a 10-by-10 booth at Comdex — the farthest-back location in the hall, right across from Lotus 1-2-3’s large well-funded booth. Several founders drove to Las Vegas with Walker in his VW Microbus. Walker had signed them up for the show after seeing a flyer with a reasonable rate and simply acted — unilaterally — without committee approval. Drake notes this “did not go over well with everyone.” He went anyway.
Two things changed everything. The first was a touch-pen input device from Sun-Flex, which Drake had integrated in a series of all-nighters three days before the show. The second was a borrowed HP flatbed pen plotter — arranged by HP’s Corporate Engineering Director Chuck House, who happened to be moderating the 2020 CHM panel 38 years later. People would stop mid-aisle and stare as the plotter zipped around picking up color pens and drawing lines. The crowd that gathered was real.
Companies were sending us their equipment saying, “Hey, put it on us so that we can show our devices aren’t toys.”
— Duff Kurland, on Comdex 1982Drake coded drivers every night of the show from the hotel room. By close of the show, AutoCAD was running in 16 other booths. Walker noted in the Autodesk File that the West Coast Computer Faire earlier that year — where the company debuted as “Marin Software Partners” before even having a company name — was the first public showing of the product. Comdex was where the product became a business.
Worth noting what AutoCAD was demonstrating at Comdex: a product that had to run in 52K of usable RAM, on machines with no floating-point chip, drawing entities that were each a single instance in Walker’s “E struct,” with a command line interface that would remain its signature for the next decade. It was primitive by any mainframe CAD standard. It was also $1,000, running on a $3,000 PC, against systems that cost $25,000 or more. The 80/20 rule before anyone called it that.
The development log Walker kept from July 1982 is one of the most remarkable documents in software history — a real-time chronicle of a product being invented, feature by feature, naming decision by naming decision, optimization by optimization, against a hard deadline and inside a 52K memory constraint. Reading it today is like watching the foundations of modern CAD laid in real time, by one person, in a machine that had less RAM than a 1980s pocket calculator.
The product was called MicroCAD when the log began. In a July 1982 entry, Walker noted that the “entity interchange format” file — the ancestor of DXF — should be an interchange standard across all MicroCAD versions regardless of internal representation. By October he had renamed it from “Entity Interchange File” (.EIF) to “Drawing Interchange File” (.DIF) with characteristic pragmatism: “There’s a hundred people who know what a drawing is for every one who knows what an ‘entity’ is.” DXF — which eventually became the universal CAD data standard, supported by hundreds of competing products for decades — started as a terminology decision in a development log entry.
I went through the whole thing and changed it to AutoCAD. This took about 6 hours because all the MCAD segment names buried in more than half of the modules had to be changed, and the modules recompiled. I sure hope this is the last time! You now start it up with ACAD, and all the file names which previously contained MCAD now use ACAD instead. Some people thought PAN worked backward, so it was changed to work backwards.
The Autodesk File, 5th Ed. — AutoCAD-80 Development LogThe product got its name on October 11, 1982 — a six-hour renaming exercise across every module in the codebase. The same entry records that PAN was reversed because Keith Marcelius said people thought it worked backwards. A rename and a UX correction, same day, six weeks before Comdex. That was the development cadence.
Walker’s dev log also shows the seeds of nearly every major AutoCAD capability, planted years before they shipped. A July 1982 entry sketches dimensioning: “one should be able to specify a default scale for the drawing, e.g., feet, angstroms, kiloparsecs.” An August entry describes a macro facility for architects who want to place dimension lines by specifying two endpoints — the direct ancestor of parametric drawing. “At some point we’re going to have to put in some kinds of macro command facility,” Walker wrote. “It might be wise to do this sooner rather than later.” AutoLISP shipped three years later. The need was visible from week one.
The “Urgent Fury” story deserves its own paragraph. By end of 1986, AutoCAD’s interface was looking dated — AutoSketch, which Autodesk had just shipped, had pull-down menus and dialogue boxes, and users were asking why the flagship product didn’t. Conventional wisdom inside the company was that retrofitting AutoCAD would be a massive, risky project touching code everywhere. Walker wasn’t sure that was true.
The moment the company shut down for the Annual Week of Rest between Christmas and New Year 1986, Walker launched into what he later called “a fury of round-the-clock programming” — integrating AutoSketch’s user interface manager into AutoCAD, extending the menu system to support pull-down menus, adding dialogue boxes. When the company reopened on January 4, 1987, it was all working. He had stuffed the documentation into every programmer’s in-box so they’d find it Monday morning. He named the project “Urgent Fury” and it became the centrepiece of AutoCAD Release 9, along with the Portable Database project he did the following month.
The command based system is the hardest for the new user to learn, but the most productive in the hands of the experienced “power user.” The event driven system, though easily mastered by the beginner, delivers little more productivity to the user who uses it constantly. An open architecture package must also consider the consequences of the adoption of a new user interface on its extensibility and the base of applications which have been built upon it. Since the history of computing has demonstrated that open architecture, extensible systems predictably supplant closed architecture, proprietary systems, any modification of the user interface of a successful, established package must be carried out in an upward compatible, responsible fashion.
The Autodesk File, 5th Ed. — “Urgent Fury”The holiday hack tradition became a recurring pattern. Walker would use the Annual Week of Rest — when the company was officially closed — to prototype capabilities that the regular development process was too cautious to tackle. Kelvin Throop did the same. It was a feature factory hidden inside a vacation. The Autodesk File catalogs them by year: AutoLISP entity access in 1985-86, pull-down menus in 1986-87, curves and sculptured surfaces in 1987-88, database recovery in 1988-89. Serious capabilities, delivered by one or two people working flat-out over Christmas.
AutoLISP deserves special attention. The IPO prospectus in 1985 described it with remarkable foresight — Walker called LISP “the first in a series of languages to be interfaced to AutoCAD, allowing users, OEMs, systems houses, and third party software developers access to the full capabilities of AutoCAD from their programs.” David Kalish, in the 2020 CHM oral history, named it the most important strategic decision in the product’s history. It created an ecosystem of 400+ third-party applications within two years of shipping — which meant that switching away from AutoCAD wasn’t just switching software, it was abandoning every customization and plugin your firm had built. Lock-in through openness, before anyone called it that.
Before AutoCAD, computer aided design was primarily done on mainframe and minicomputers, often with proprietary graphics hardware. With the introduction of the IBM PC and the many 16-bit desktop machines which followed, the basic desktop office computer reached a level of capability which allowed serious computer aided design to be done on the machine as supplied by the manufacturer. Thus, AutoCAD was introduced into an essentially vacant market: a software package for computer aided design sold separately from hardware and intended for use on existing desktop computers. AutoCAD is written in C, one of the most widely implemented and compatible computer languages for software development available today. These design principles allow Autodesk to market AutoCAD on virtually any computer system it chooses as a potential market.
The Autodesk File, 5th Ed. — Initial Public Offering prospectusWalker’s open architecture argument wasn’t marketing copy — it was the answer to the fundamental question facing a tiny company trying to displace Intergraph, Computervision, and CADAM: why should anyone trust a three-year-old startup from Marin County? Because AutoCAD ran on hardware you already owned. Because it worked with peripherals you could buy from anyone. Because your own programmers could extend it. The mainframe vendors were selling total solutions — one vendor, one architecture, take it or leave it. Autodesk was selling a standard that worked with everything. That bet paid off so completely that decades later, the DXF format Walker sketched in a July 1982 dev log entry was still the universal language CAD systems used to talk to each other.
The operating model would look unremarkable today. In 1982, it was extraordinary. The core technical team worked from home and came into the office once a week for the Thursday meeting — held at a local Chinese restaurant where they were “not quite so welcome,” Drake noted, “because we made a lot of noise.”
Duff Kurland served as the human source-control system, collecting code changes on floppy disks from every programmer, manually merging them into a working build, and redistributing the result. Walker wrote one of the first file-transfer programs — “File Tran” — so developers could exchange files by modem. Before that, they mailed floppy disks through the US Postal Service.
Greg Lutz discovered an undocumented socket on the IBM PC motherboard — the 8087 floating-point coprocessor. He wrote detection code: if the chip was present, use it; if not, fall back to software floating point. The speed difference was dramatic. No competitor matched it. Most shipped two hardwired versions; Autodesk shipped one that adapted.
The catch: Lutz had always tested with the chip installed. He hadn’t tested without it. The detection instruction would simply hang — waiting indefinitely — if the socket was empty. The very first customer, an architect, called to report that AutoCAD froze immediately on launch. His machine didn’t have the chip. AutoCAD’s most technically innovative feature had killed its first sale.
Al Green’s oral history — recorded for the Computer History Museum on June 10, 2021 (CHM accession 102792234, ref X9490.2021), interviewed by Douglas Fairbairn as part of the Software Industry Special Interest Group project — gives Autodesk’s rise an entirely different texture. Green came from accounting and finance, not programming — Ford Motor, Ford Aerospace, oil and gas, a CAD startup called Cascade Graphics. Dan Johnson of Cascade met Drake at Comdex 1983 and mentioned Green’s name. Drake called that fall with his pitch: no money, might be out of business in six months, worthless stock. Green’s wife’s response: “Tell me you didn’t say yes.”
The promotion to CEO in 1986 was similarly unplanned. Walker announced he was done with management — it was taking time he wanted to spend programming. He looked at Drake and Green. Drake said “I can program.” Walker told Green: “You’ve built the company, you might as well run it.”
Five years ago Al Green assumed the leadership of a raw-edged start-up company which had come so far so fast it had outstripped the ability of its founders to manage even day-to-day operations. Sales were running at $50 million per year, having almost doubled in the preceding twelvemonth. It was clear to everybody, especially to me, that we were careening from crisis to crisis. On November 5, 1986, I ceased to be CEO and Al took the job. The task before him was as easy to describe as it was difficult to accomplish: “Fix it.” His success has been complete and unqualified. Under his leadership, Autodesk has grown by a factor of five, into a company with divisions around the world, well equipped to sustain and accelerate its rate of growth.
The Autodesk File, 5th Ed. — “Al Green Retires”The five years between Walker leaving the presidency and Green leaving it produced the company’s greatest sustained run: zero down quarters, expansion into Japan and the Soviet Union and Switzerland, the AME solid modeling breakthrough, the first AutoCAD Expo drawing 10,000+ attendees, market share approaching 80 percent. Walker kept programming and writing Information Letters. The arrangement worked.
In April 1991, Walker circulated a 44-page internal document titled “Information Letter 14: The Final Days?” It is one of the most extraordinary documents in the history of the software industry — a founder publicly diagnosing the decay of his own company in real time, in a document he knew would escape the building.
The diagnosis was specific. AutoCAD’s development team was too small (Walker counted roughly 15 people putting content in the box for a product generating $200 million a year). Marketing budgets were being cut to protect margins. New products were being launched then immediately abandoned. The Windows transition was being treated as optional. CorelDRAW, selling for $595, was surpassing AutoCAD in the features they shared, at one-seventh the price. Walker listed the comparison in a detailed table that remains painful reading.
Autodesk possesses all the prerequisites to lead the next generation of the PC industry, yet it seems to have become stuck in the past, mired in bureaucracy, paralysed by unwarranted caution, and to have lost the edge of rapid and responsive product development and aggressive marketing on which the success of AutoCAD was founded. Not only has Autodesk failed to bring the new products it needs to the market, it is allowing AutoCAD, our flagship product and the source of essentially all our revenue, to become dangerously antiquated and undermarketed to an extent that is virtually unique for a product generating sales in excess of $200 million a year.
The Autodesk File, 5th Ed. — Information Letter 14Bill Gates called IL14 “brilliantly written and incredibly insightful” and used it as a template to write his own internal memo about Microsoft’s risks. The document circulated to Wall Street analysts, to Microsoft, and to the press within days of Walker distributing it internally. Dan Drake describes coming back to his office and finding it on his desk, then walking out through the lobby past “knots of people furiously discussing this thing,” making his way out as quickly as he could.
The IL14 aftermath produced a reorganization into business units, a CEO search, and eventually Carol Bartz. The search via Heidrick & Struggles produced three candidates. Drake vetoed the first on management philosophy grounds — that person later went to prison for accounting fraud at a major database company. He vetoed Del Yocam from Apple for lacking technical management instincts. The third was Bartz. Both Drake and Lutz say they advocated for her hire. Both later expressed regret about it. Green, in his 2021 CHM interview, is more generous: the company needed different leadership at that level, and it found her.
Bartz’s own oral history — recorded for CHM on October 13, 2021 (accession 102792289, ref X9558.2021), interviewed by Douglas Fairbairn — gives the other side of this transition in stark, unsparing terms. She describes walking in on day one not fully grasping what she was stepping into: thirteen founders, no VC money, and what she called “total chaos.” The chaos was specific. She sent a message to anyone who thought they reported to her: come to the building. Fifty people showed up. She had no idea who most of them were or what they did. Her discovery: no HR department existed at all. “If you wanted a raise, you just convinced somebody, and these poor other schmucks who didn’t convince somebody never got a raise. Ever. They didn’t know how to fire people. They didn’t know how to hire people.” The guy running service was wearing a literal sheriff’s star. This was the company that had just spent a decade with zero down quarters.
The first day she went to work, Release 11 was about to ship. A briefing: 160 new features. She kept asking the same question: “Who are they for? Are they for architects? Engineers? Civil?” The answer kept coming back: “Well, they’re just 160 new features.” That was the whole model — every two years, another hundred features, no intended user, no market segmentation. The most senior non-founder at the table sat typing, taking no responsibility for any answer. She took him for a walk at lunch and fired him on the spot. By the time she got back upstairs, people were asking where he’d gone. She said he was gone. “Well, then everybody decided they should pay attention.”
Walker was in Switzerland by then, having made a tax arrangement with one of the Cantons. Bartz describes him as “a big hand coming over from Switzerland.” The interference got bad enough that she flew to Neuchâtel — her first Concorde flight, which got diverted to Gatwick in fog and took longer than a regular jet — to tell him directly: “You got to stay out of this. You just can’t keep stirring the pot.” He reluctantly agreed. Then convinced the European VP to attempt to secede from Autodesk entirely — wanting the code but not the company. She fired that VP too.
Her customer tour produced the finding that explains everything about why the founding culture had hit its limit. Architects told her: “We don’t really like AutoCAD, but we use it — it’s really for the engineers.” Engineers told her: “I like it, I use it — but it’s really for the architects.” Nobody claimed it. The product Walker and Drake and Lutz had built by adding features they thought were interesting had no specific owner in the market. Bartz’s phrase for the old model: “Every two years they’d just come out with another hundred features. That was their whole manner of working.” Fixing that — segmenting the product, building vertical suites for architecture, engineering, civil — took years of internal fighting. It was the right diagnosis, even if the execution was imperfect.
The Windows transition was the most consequential battle of her tenure. The founders were convinced DOS was what engineers wanted. By 1992 that was still partly true — Windows wasn’t yet performing well enough — but Bartz believed it was the future and pushed the transition. The result was Release 13: a disaster, shipped too soon, bug-ridden enough that an industry outcry threatened to become a mutiny. She is characteristically direct about it in the CHM interview: “Part of that was my fault. I just kept wanting it to get out.” Release 14 was the recovery — solid, successful, and finally got Windows right. By the time she stepped down in 2006, Autodesk had grown from roughly $285 million in revenue to $1.5 billion.
Carl Bass came to Autodesk the way a lot of the best people did in the 1990s — through acquisition. He had co-founded Ithaca Software in 1981, the same year Walker was writing his Working Paper, to commercialize HOOPS, a 3D graphics system developed at Cornell. By his own account he didn’t even know who Autodesk was when he started: “We were a bunch of young folks, and Autodesk was just getting started as a company at the same time that Ithaca Software started.” Autodesk acquired Ithaca in 1993, bringing Bass in as chief architect of AutoCAD. He lasted two years the first time. Bartz apparently found his habit of roaming executive hallways in Birkenstocks dispensing unsolicited technical opinions — and usually being right — somewhat less than ideal. She fired him. Then rehired him five months later. He left again to found Buzzsaw, an online AEC collaboration platform. She bought Buzzsaw in 2001. He came back a second time. She named him CEO when she stepped down in 2006. He ran the company for eleven years.
Bass articulated Autodesk’s history in three phases, which remains the cleanest summary anyone has offered: the founders’ period led by Walker; the Carol Bartz era; and what he called a new era beginning roughly ten to twelve years before he was speaking. Walker built the standard. Bartz built the company. Bass built the portfolio — moving Autodesk from a one-product CAD vendor to a multi-platform design software company spanning 3D, BIM, simulation, generative design, and eventually the cloud. Revit, acquired in 2002 and turned into the dominant building information modeling platform worldwide, is the clearest single example. Autodesk crossed $2 billion in revenue in 2008, two years into his tenure.
On the impossibility of recreating founding-era culture at scale: “People say they want a big company to be more like a start-up, but I think that’s a little bit of a false promise. You’re not going to have a company of 10,000 or 50,000 or 100,000 people that acts like a company of 50. Having gone through that growth stage, even as you go from five to 50, some of the things that worked with five don’t even work with 50. Every attempt any company has made to do that has proven that to be true. You can simulate it, make cultures that move more quickly and kind of mimic some attributes, but they will never be the same thing, no more than a fish is going to be a gorilla.”
That’s the honest coda to the founders’ panel. Drake, Lutz, Kurland, and Kalish describe the Thursday meetings at the Chinese restaurant, Duff collecting floppy disks from everyone’s house, Walker writing File Tran so they could exchange code by modem — with genuine pride, and they should. It worked. It worked spectacularly. But Bass, who arrived a decade later and lived through every subsequent phase, understood exactly why it couldn’t last. The culture that produced AutoCAD in 52K of RAM, that crashed the UNIVAC to earn a job offer, that coded drivers from a Comdex hotel room — that culture was the founding’s irreplaceable asset and its inherent ceiling, both at once.
Bass stepped down in February 2017 and was succeeded by Andrew Anagnost, who had joined Autodesk in 1997 as a product manager — fifteen years after the founding, five years after Bartz arrived. Anagnost came from Lockheed, NASA Ames (where he worked on Mars Pathfinder simulation), and a stint at a Boston fluid-dynamics startup. He holds a PhD in aeronautical engineering from Stanford. He led Autodesk Inventor’s revenue from a standing start to over $500 million, served as chief marketing officer, and was the architect of Autodesk’s transition from perpetual desktop licenses to cloud subscription — the most operationally difficult transformation in the company’s history since the Windows port. Under Anagnost, Autodesk crossed $5 billion in annual revenue (FY2023), entered the Fortune 500, and reached a market value exceeding $40 billion. The company Walker founded with $59,000 in 1982 is now one of the most valuable software companies in the world.
The Autodesk File is unusual for a corporate history because it reveals that the founder was thinking, from the very beginning, about something much larger than a CAD company. Walker’s Information Letters trace a consistent arc: exponential growth of computing power, the coming transformation of design and manufacturing, and eventually molecular engineering and the digitization of the physical world.
At the Autodesk Technology Forum in May 1990 — with Eric Drexler sitting in the audience — Walker gave a talk tracing six industrial revolutions from stone tools to molecular engineering, positioning Autodesk at a unique inflection point. A company in the business of building digital models of real objects, at precisely the moment when that capability was about to become essential to everything.
We are living on a small blue sphere in an endless black void. Over the next twenty or thirty years, the human population is expected to double. Five billion new people are going to be sharing this world with us. To provide those people with the food, the shelter, the clothing, and all the other things we feel entitled to ourselves, we are going to have to design and manufacture, over the next several decades, as many artifacts as all of humanity have created over the last two million years. And we shall have to do that in a way that preserves this fragile home of mankind for the generations that will follow. The mission of computer aided design is to create and deliver the tools we will need to accomplish that.
The Autodesk File, 5th Ed. — AutoCAD for Windows press conferenceWalker’s bets on Xanadu, the American Information Exchange, HyperChem, and the Cyberspace project were all expressions of this larger vision. Not all of them paid off — several were arguably too early by decades. But the underlying argument — that every manufactured object would eventually be modeled in a computer, and that Autodesk should build the tools for that future — was exactly right, just on a longer timeline than he expected. The company Walker founded in 1982 is now generating over $5 billion a year in revenue from exactly the markets he described at that 1990 technology forum.
In the 2020 CHM panel, Dan Drake refers twice to “a guy at Autodesk who knows a lot, who has specialized in the history” when recommending the Computer History Museum contact the company. Both times he can’t recall the name. The editor’s notes say: [Shaan Hurley].
Bass’s three-era framework — Walker, Bartz, Bass — is useful shorthand, but it compresses something important. Each era didn’t just change the leadership; it changed what kind of company Autodesk was. Walker built a standard by accident, then documented it obsessively so it wouldn’t be forgotten. Green built the business infrastructure to sustain the standard — finance, operations, international expansion — without which none of what followed was possible. Bartz diagnosed the product’s market failure, took a brutal cultural beating from thirteen founders who didn’t want to be managed, and industrialized the company anyway. Bass expanded it from one product into a platform. Anagnost moved the entire platform to the cloud while keeping it from collapsing in the transition. That’s forty-plus years of consecutive successful reinvention, each requiring a fundamentally different kind of leader.
Walker knew the forgetting would start early. He built the Autodesk File specifically because he believed the founding story would be lost as the company grew — that the people who arrived after 1985 deserved to understand how it actually started, what the early decisions felt like from inside, and why the culture was the way it was. He compiled and annotated internal documents spanning more than a decade, added footnotes correcting his own earlier optimism, and published the whole thing freely on the internet. He kept updating it through 2017.
Now Walker is gone. Keith Marcelius is gone. The voices of the founding era exist now primarily in the CHM recordings and in the 900 pages Walker left behind. Bartz and Bass have now added their perspectives to the same archive — Bartz recorded in October 2021, Bass in October 2022, both part of the same Software Industry SIG project that produced the founders panel and Al Green’s interview. That’s five oral histories covering every CEO from Walker through Bass, all in one CHM collection, all freely available. That’s an extraordinary record for any company.
What makes these documents valuable isn’t just the technical record — though Walker’s development log from July 1982, where you can watch AutoCAD acquire its first features one commit at a time, is genuinely irreplaceable. It’s the human texture: the crashed UNIVAC that earned Lutz a job offer, the all-nighter Comdex drivers, the sheriff’s star in the service department, the Concorde flight that took longer than a regular jet. The stories that don’t survive in press releases or annual reports.
Walker had a phrase for how the company operated in its best years: “management by lack of alternatives.” They built what was achievable, with what they had, because there was nothing else to do. Bass’s response, forty years later, was equally direct: you can simulate startup culture in a big company, but a fish is never going to be a gorilla. Both are true. The $59,000 bet in a Mill Valley living room became a $40 billion company. The story of how that happened is sitting in the Computer History Museum archive, mostly unread, waiting.
That’s the project. Let’s keep going.
Primary Sources
CHM Oral History #1
“Autodesk Oral History Panel: Dan Drake, David Kalish, Duff Kurland, Greg Lutz”
Interviewed by Chuck House and Burt Grad via Zoom
Recorded October 15, 2020 · 49 pp.
CHM Reference: X9368.2021 · Accession: 102792114-05-01
© 2020 Computer History Museum
CHM Oral History #2
“Oral History of Alvar Green”
Interviewed by Douglas Fairbairn via Zoom · Belvedere, CA
Recorded June 10, 2021 · 30 pp.
CHM Reference: X9490.2021 · Accession: 102792234-05-01
Conducted by the Software Industry Special Interest Group
© 2021 Computer History Museum
CHM Oral History #3
“Oral History of Carol Bartz”
Interviewed by Douglas Fairbairn · Mountain View, CA
Recorded October 13, 2021 · 42 pp.
CHM Reference: X9558.2021 · Accession: 102792289-05-01
Conducted by the Software Industry Special Interest Group
© 2021 Computer History Museum
CHM Oral History #4
“Oral History of Carl Bass”
Recorded October 18, 2022
CHM Accession: 102792770
Conducted by the Software Industry Special Interest Group
© 2022 Computer History Museum
The Autodesk File, Fifth Edition (2017)
John Walker · fourmilab.ch/autofile
Chapters cited: AutoCAD-80 Development Log; Initial Public Offering prospectus; “Urgent Fury”; Information Letter 14 (“The Final Days?”); “Max Q”; Nanotechnology in Manufacturing (Technology Forum, May 10, 1990); AutoCAD for Windows press conference remarks (March 10, 1992). Freely available online in full.
