High-Tech Rebel

By David E. Sanger
The New York Times

September 11, 1988

It is early morning and most of the top leaders of the nation's personal computer industry are jammed into a hotel banquet hall in Naples, Fla. The annual, late-winter gathering is one that industry insiders can't afford to miss - 72 hours of closed-door deal-making, hushed corridor conversations about technologies gone amiss, and arguments over the future course of America's most visible high-tech industry. This year, marking a decade since personal computers began to revolutionize the working habits of tens of millions of people, there is also a heavy dose of self-congratulation in the air.

But at the podium, one of the heroes of that revolution, a stocky man in his late 30's, is telling the audience of engineers and millionaires something many of them don't want to hear. ''The state of the art in computing is incredibly primitive,'' Mitchell D. Kapor says as conversations halt and heads bob up from the morning papers.

''This is the auto industry, it's 1902,'' Kapor goes on. To argue, as many do, that the development of new technology must now proceed step by step to avoid further confusing users ''is to say that the development of the automobile stops before the Model T.''

Coming from anyone else, warnings that the industry's technological fire was being doused would be quickly dismissed. But Mitch Kapor cannot be ignored. He is a near-cult figure in the industry, the developer of Lotus 1-2-3, the world's most successful software package. With his quick debating style, biting wit and an easygoing manner that masks deep intensity, he is also something of an enigma; two years ago, as chairman of Lotus Development Corporation, one of the computer world's greatest success stories, he walked away from it all.

Today, Kapor is one of the harshest critics of the computer establishment, contending that it has become a prisoner of its own empire-building mentality, more interested in miniaturizing chips and refining existing machines - thus preserving its hard-won market shares - than in once again rethinking the personal computer's role.

Personal computers, Kapor maintains, are still chiefly useful only to people who crunch numbers, keep lists or compose words for a living. In fact, many people use PC's for a single task (writing memos, keeping addresses on file) and untold thousands are too intimidated to even turn them on. So far, the personal computers' broader potential - as true assistants instead of mere tools, able to cajole, reorganize, research, separate the trivial from the critical - remains unrealized.

In fact, personal computers have yet to emerge from their adolescence. They are still maddeningly literal in their execution of instructions, still awkward to deal with. ''Sooner or later, everyone that I know, except perhaps the most hard-core technical fanatic, wants to pick up the machine off the desk and throw it through the window,'' Kapor told a forum in California's Silicon Valley last fall. ''The frustration level is that great.'' Not surprisingly, many disagree vehemently. Competitors argue that Kapor's views are too black and white, that it's simplistic to see the industry divided into two camps: established companies with huge investments in current technology pitted against courageous innovators, maybe like Kapor himself, who are prepared to walk away from success in order to break new ground.

Even Kapor's friends shake their heads at his complaint that computers are still far too hard to use. ''It's a little like complaining that there's waste and fraud in the Government,'' notes Jim P. Manzi, Kapor's successor as chairman of Lotus. ''If the problem was easy, it would have been fixed a long time ago.''

Nonetheless, many believe Kapor is raising questions that the industry is ignoring at its peril. They are questions about the pace of innovation -when to charge ahead, when to let computer users catch their breath, when to fear that the Japanese would finally get the knack of new software technologies that has kept the American computer industry in the forefront of the world.

As for Kapor, the test of his ideas will come over the next three years. In a business where entrepreneurs are notoriously unsuccessful the second time around, Mitch Kapor has begun another start-up company, replete with a crack technical team and lots of cash. His goal is to develop software that will, as he says only half in jest, ''give my mother a good reason to own a computer.''

In an industry long dominated by ''wireheads,'' engineers who revel in the complexity of their inventions, Mitch Kapor has always been different.

At Yale University in the late 1960's, he dabbled in computer science but concentrated on linguistics and psychology. In his 20's, he was immersed in ''personal consciousness,'' joining the transcendental-meditation movement after college, studying Eastern philosophy, working for a while as a disk jockey. After a failed marriage, he got a master's degree in psychology and spent a brief, disillusioning time in the late 1970's as a counselor in the psychiatric unit of a hospital outside Boston, a job he once described as ''the psychic equivalent of emptying bedpans.''

Ten years ago, he quit and made a purchase that changed his life and the course of the personal computer technology. He had long been fascinated with computers and he bought one of Apple Computer's first products, the Apple II. ''There were maybe 50 Apple users in the greater Boston area,'' he recalls, and ''virtually no software to make the thing useful.''

Kapor had tinkered a little with BASIC, a simple computer programming language. He and a fellow enthusiast soon developed Tiny Troll, which, says Kapor, was inspired by an expensive, unwieldy piece of mainframe computer software at the Massachusetts Institute of Technology that kept track of statistics and figures useful chiefly to economists.

Tiny Troll led Kapor to join forces with two software pioneers, Daniel S. Bricklin and Robert M. Frankston, who had just developed the first personal computer spreadsheet program, called Visicalc. Visicalc solved one of the most vexing problems facing anyone who has ever dealt with columns of interrelated figures: Change one number or formula, and every other number on the ledger must be recalculated. Computerized spreadsheets do that instantly.

Visicalc was already a sensation when Kapor turned Tiny Troll into VisiPlot and VisiTrend, programs that plotted statistical data and analyzed trends on the Apple II. But Visicalc's founders had little interest in Kapor's next idea, a program that combined spreadsheet and graphics. So, with a separation settlement of $1.2 million in his pocket, he left to pursue his idea elsewhere.

That idea, refined with Jonathan M. Sachs, an experienced programmer who turned Kapor's vision into disciplined lines of computer code, became Lotus 1-2-3. Early on, Kapor and Sachs decided not to develop the program for Apple but for I.B.M.'s new personal computer. It was a fateful decision. Partly because it melded spreadsheets and graphics, partly because it radically improved spreadsheet technology, 1-2-3 became the biggest hit the software industry had ever seen.

Many experts credit it with making the I.B.M. PC as commonplace in the office as file cabinets. Today, five years after its first release, 1-2-3 has spawned an empire. More than three million spreadsheets have been sold; by some estimates, upward of seven million number crunchers - from small-business owners to chief executives to accountants to engineers - use the program regularly. There are versions in 65 countries. Recently, 1-2-3 became the hottest-selling software package in Japan.

Lotus, meanwhile, grew from a garage-shop operation to one of the world's biggest software houses - with 2,300 employees and headquarters on the banks of the Charles River, in Cambridge, Mass., near the M.I.T. campus. Sales in 1983, the first year, were $53 million, making the outspoken Kapor, in his signature Hawaiian shirts, an instantly identifiable industry hero alongside William H. Gates, head of Microsoft Corporation who was then 27 years old, and Steven P. Jobs, the mercurial co-founder of Apple. Last year, Lotus posted revenues of $400 million and profits of $72 million, although in recent months a host of product delays and false starts on several programs have led many to believe that Lotus has lost the magic of 1-2-3. Without question, it has lost the magic of Mitch Kapor.

In 1986, Kapor simply left his job as chairman. Unlike Jobs, who was ousted from Apple in a power struggle with John Sculley (and who is now developing a new computer at his own start-up company), no board of directors pushed Kapor out the door. He simply concluded that running a large corporation, with its Wall Street pressures and bureaucratic responsibilities, did not mesh with his creative urges.

For a while, he had tried to insulate himself from the demands of running the company, forming a ''Chairman's R and D'' group that was supposed to develop new products. But Lotus' success often got in the way. Each new innovation, each new idea, quickly became hopelessly entangled in problems of compatibility.

Users of 1-2-3 - who had already invested huge sums in training and building spreadsheets - understandably insisted that they be able to move to the next Lotus product without starting from scratch. But Kapor learned anew that true innovation and compatibility are often at odds. ''It's a millstone we all have to live with,'' said Edward J. Belove, a Lotus executive and close associate of Kapor's. Kapor soon decided, though, that he could not live with it.

''Mitch needed to start with a clean slate,'' says Jim Manzi, the buttoned-down Lotus chairman whose bottom-line management style is the polar opposite of Kapor's. (Manzi's $26 million compensation last year was the second-highest of any chief executive in the nation.) Not only did Kapor resign from his posts, he also sold off his stake in the company, once worth about $70 million.

The choice Kapor made encapsulates one of the running debates about the computer industry's future. Improving old software incrementally, tempering innovation with good business sense, often seems the safe middle course. But genius rarely comes in neat increments. The most successful new categories of software in recent years - spreadsheets, new types of databases and desktop publishing (the programs that lay out fancy newsletters and magazines and eliminate the need for expensive typesetting) - came chiefly from upstart companies with nothing to lose.

Among established companies, the financial pressure to move step by step is enormous. Apple ignored that pressure when it brought out the Macintosh; although the strategy ultimately paid off, the company nearly went broke trying. Kapor believes that the drive for compatibility has become too overwhelming, particularly at companies like I.B.M. ''It's too huge a job, and it stops too many good ideas,'' he says about I.B.M.'s connect-everything-to-everything philosophy. ''It's going to kill them.''

But the watchword of the computer industry these days is I.B.M.'s ''Listen to the customer.'' Although I.B.M. denies it, that business-school maxim usually works against radical technologies. ''Creating revolutionary change in the computer industry these days is like starting a new religion,'' says S. Jerrold Kaplan, a close friend of Kapor's and a leading software designer in Silicon Valley. ''At this point a lot of people don't really want new technology. They want to solve problems they already know about.''

The place where Mitch Kapor hopes to solve the problems that computer users don't yet know about, or cannot articulate, resembles the setting for a movie about a wacky high-tech start-up company.

Visitors to ON Technology's offices in Cambridge, about a 10-minute walk from Lotus headquarters, are greeted by a giant inflatable dinosaur in the reception area. In the back, squeezed between empty work stations awaiting new employees (so far there are 22) and a modest library, stands a brand-new Ping-Pong table. Technical debates often take place in the middle of grudge matches.

Although raucous Ping-Pong matches are in marked contrast to the hushed atmosphere that has swept over Lotus in recent years, ON is hardly a garage-shop operation. It has already attracted more than $3 million in venture capital, and somewhere between $10 million and tens of millions more will be required. Lotus began in two dingy rooms in one of Cambridge's gray, warehouse neighborhoods. ON occupies 9,000 square feet in one of the modern new towers popping up around Kendall Square.

What the 37-year-old Kapor and his colleagues are attempting to create is a set of new building blocks for the coming generation of personal computers, software that would enable personal computer users to share -across networks - audio, documents faxed from afar, drawings, words and, eventually, even video images. Most important, the system would finally release the computer from a world dominated for centuries by paper - a medium that expects a reader to absorb information in the order in which it is printed, regardless of the reader's level of knowledge.

''The computer is capable of making information appear differently to every person, but we don't exploit that power yet,'' Kapor maintains. His vision - an extension of a current technology called ''hypertext'' - is to design software that reorganizes data, providing newcomers to a topic with one level of information, sophisticates with another.

For people who deal with untidy text and ideas, Kapor envisions software that transforms computers into actual assistants, correlating related concepts, suggesting ways to help. The kernel of that effort can be seen in Agenda, a program that Kapor, Kaplan and Belove labored on for years at Lotus, and that was just released this summer after some delay.

Agenda is among the first of a new breed of programs that are, basically, spreadsheets for words and ideas. What the program does, after some initial guidance from the user, is begin to analyze text and divide it into different categories that would help a busy person stay organized.

So, if you typed in your computer: ''Call Ted S. tomorrow about our important meeting on the lawsuit,'' Agenda will let you create a host of categories: Calls to make, Ted S., important items, lawsuit. With each new note to yourself, Agenda will file the information in the right category.

By the end of the day, it will show you a list of all the people you have to call tomorrow. It will also have cross-indexed your note in an electronic file called ''Ted,'' and put another copy of it in the file on the lawsuit. Because you indicated the topic was important, it might also be put on a list of high-priority items. And, of course, the user can search later for all information having to do with Ted and a lawsuit, without having to go through notes about Ted and other subjects.

Agenda has its limitations, chiefly that it cannot create categories without a fair amount of human help. It is not the only program of its kind entering the market, but it is a remarkable example of the subtle artificial-intelligence techniques that make computers far more useful and easier to use.

Kapor likens it to a Boeing 767 - a plane that can land itself. ''It's a far, far more complex machine than any commercial aircraft before it,'' says Kapor. ''But for all its complexity, it's a lot easier to fly.''

The second remarkable thing about Agenda and programs of its ilk is that they actually have some limited understanding of the content of untidy, textual information, rather than treating it all as disparate bits of data. That is the point when ordinary personal computers begin to resemble true assistants - which is what the future of software is about.

At my old desk in the newsroom of The New York Times I sat opposite Donna Anderson, a premier researcher. Sometimes it seemed that there was not a printed fact in the world that Ms. Anderson could not extract, no matter how fuzzy my request. Oftentimes, there was no request at all. When a story broke close to deadline and chaos reigned, Ms. Anderson knew what was needed; magically, the essential background material, the key statistic, the telling quotation appeared on my desk.

The computer industry has spent years, with little success, trying to replicate people like Donna Anderson. By their nature, computers are precise and intolerant of vagueness and, unlike Ms. Anderson, they cannot translate inarticulate requests into specific searches, or know not only where to look but when to stop.

Yet that is precisely what Kapor hopes the next major generation of software will be able to do. ''It's a very different concept of what a computer should be,'' says Esther Dyson, the editor and publisher of a leading newsletter on computing, called Release 1.0, and one of the industry's most influential figures in blessing or condemning new technologies. ''So far, we have relied on computers that can store something and pull it out. What Mitch is trying to move toward is a computer that can act semi-independently on information. In a sense, it is the interface to the rest of the world - it talks to the world for you and provides preliminary responses.''

The technology to do just that is still distant (estimates vary from a few years to a decade or more), but the basic idea has been kicking around for a long time at places like M.I.T., Carnegie-Mellon University and Stanford University, the centers of artificial-intelligence research in the United States. Ultimately, scientists hope to create a code, or ''agent,'' that would represent the interests of the computer user - to know, as Ms. Anderson knows, what is important and what can be dismissed, and to operate with minimal instruction.

Suppose you need to keep track of the Middle East crisis, advances in auto design and the fate of the Boston Red Sox. Every night, when rates are low, your research agent would wend its way over the phone lines into the scores of electronic databases that are constantly updated with articles that have appeared in thousands of publications. While you are at dinner or asleep, it would pluck out stories and abstracts about the latest fighting in the Gaza Strip, rumors about Honda's new models, and a leading sportswriter's assessment of how long the Red Sox can keep their winning streak going. In the morning, the stories would pop to your screen.

There is only one problem with this happy scenario: How do you represent abstract concepts in the computer? If the article doesn't say so specifically, how do you program a machine so that it knows that a meeting between King Hussein of Jordan and Mikhail S. Gorbachev concerns the Middle East, that Honda makes cars, that a story about the Green Monster in Fenway Park probably has something to do with the Red Sox?

''This is what frustrates people the most about computers,'' says Peter B. Miller, a veteran of several computer ventures who left Lotus's advanced development laboratory and later joined Kapor at ON. ''Simple, common-sense things, where computers are incapable of taking into account the obvious, make a lot of people think that machines aren't that useful. That's the core problem we are trying to tackle here.''

Initially, the agents Kapor's team is planning would be fairly simple. They would not share Donna Anderson's sixth sense about what I wanted, nor would they learn from experience. But they would be ''trained'' to sift the important from the unimportant.

Programmed with some knowledge of the world, such software could eventually do far more than research. When electronic mail becomes widespread, a mail-sorting agent could scan each letter and place the most important ones on top. Meanwhile, another agent could act as a scheduling secretary - for instance, arranging for a car rental to be waiting after your plane lands.

Problems arise when the computer encounters something very human -(Continued on Page 90) office politics, for example. Suppose a computer user who is a late riser programs his ''scheduling agent'' with a simple rule: No appointments before 10 A.M. But if the boss's computer calls to ask for an 8 A.M. breakfast meeting, it might be impolitic for a ''no'' to be shot back. How does the computer know that the rule applies when dealing with colleagues or assistants, but not with a superior - or with colleagues likely soon to become your superior?

''We are not going to solve these problems with the first generation of products,'' says Kapor. ''But we need to build a platform that makes their solutions possible. So far, no one has done that.''

The battle in the computer industry today is over who will build that platform and how it will be constructed. Perhaps nothing symbolizes the current split in the computer industry more than the feud between Kapor and Bill Gates, who heads Microsoft. What began as a polite, academic difference of opinion over the future of the personal computer has turned, in the last several months, into a shouting match.

Gates built his fame and billion-dollar fortune as the author of MS-DOS, the operating system that I.B.M. adopted for its first personal computer. Operating systems are the core software on which everything else depends, and over time MS-DOS became too limited to handle the demands of increasingly sophisticated users. So a year and a half ago, the Gates-I.B.M. team brought out OS/2, the successor system, which allows users to run several programs simultaneously, issue instructions simply rather than in computerese - and makes it possible to develop far more complex spreadsheets, graphics programs and so on. But to many minds, including Mitch Kapor's, OS/ 2 was laden with compromises to older technology that squelched its potential.

The argument broke into the open in Gates's lakeside backyard in a posh Seattle suburb one evening in March. Gates was hosting a dinner for industry executives and engineers attending a conference. At the buffet table under a giant tent, Kapor, a guest, was ''sort of baiting Bill, complaining that Microsoft was hampered by its need to make everything compatible,'' recalls one of the many guests who gathered around the two in embarrassed silence. ''Bill replied, 'You don't understand what great things we will soon be able to do.' Neither one convinced the other, and they started shouting.'' The argument stopped for a time, and then continued through the night.

Obviously, ego as well as technology are involved in the dispute. Kapor regards Gates as an ''empire-builder, someone who wants to build the Standard Oil of computing.'' In fact, Microsoft is now the most successful software company in history. And control over OS/2 gives Gates enormous power over the future course and pace of the industry.

''If I was in Gates's position, I'd probably have to think the way he does,'' says Kapor. ''The difference is, by leaving Lotus I chose not to be in his position.'' For his part, Gates says that Kapor is a ''completely nontechnical guy who knows enough to sound technical.''

''I can't believe the things that Kapor is saying these days,'' says Gates, his voice raised in anger. ''It's nice of him to discover that computers are hard to use. When he was off in transcendental meditation, we were coming to work here every day to solve the problem. We still are.''

But beneath the ego clash, there is a real difference of view. Kapor casts the computer industry in political terms, a world of liberals and conservatives, ''irresistible forces and immovable objects.'' It is a world of people who love technology, but whose judgment is often clouded by the billions of dollars at stake.

Gates describes himself as an engineer, for whom the answers, and the art, lie in the details. ''I don't have time to be philosophical,'' he says. ''We've got a lot of concrete problems to solve for users now, and that's what I do. This isn't a question of politics.''

His position has evoked some sympathy. ''Bill's annoyance is understandable,'' says Jerry Kaplan, the former chief technologist at Lotus, who knows both men well. ''His instinct is to improve the state of computing as quickly as practical. The debate is over what is practical - and Bill, unlike Mitch, has a family of 10 million computer users to feed.''

Of course, the computer industry needs both kinds. It needs architects like Kapor to envision how people will live; it needs engineers like Gates to keep the whole infrastructure from falling apart. The trick is to keep the vision from being obscured by the demands of the market.

''Is it still possible for someone to invent an idea, a methodology that can transform computing?'' Kapor asks late one afternoon in his office, as lights begin to flick on in the M.I.T. laboratories around him. ''Yes, I still think it's possible. But it's getting harder.''

GRAPHIC: Photo of Kapor in his office in Cambridge, Mass. (Claudio Edinger/Gamma-Liaison) (pg. 60)

David E. Sanger, a correspondent in the Tokyo bureau of The New York Times, recently completed a five-year assignment covering high technology

Copyright 1988 The New York Times Company