Startup: A Silicon Valley Adventure

by S. Jerrold Kaplan

  Frink, the engineer, took up hours of our time, debating our approach at every turn. He was enamored of Windows, and couldn’t imagine why it wasn’t suitable for our application. “I thought he came here to build applications, not to convince us that we’re wrong,” Robert confided during a break.

  “Look, why don’t you walk him through our logic one step at a time,” I suggested. “He’s a smart guy, he’ll catch on.”

  Robert put together a detailed list of issues, hypothesizing a product called Windows/H, for Windows with Handwriting. He did a masterly job of explaining its shortcomings. Frink took notes, but hardly seemed convinced. He spent many hours studying our documentation to understand how we addressed each of these problems.

  A few weeks later, Robert was invited to brief a larger group of people in the applications division at Microsoft headquarters, in Redmond, Washington. He was thoroughly disgusted when he returned. “All they did was beat me up about why we should be using Windows. They insinuated that if we didn’t, they might do this themselves. After my presentation, they arranged a series of meetings to convince me, including several with key members of the Windows development team.”

  “I thought they were going to keep this stuff separate from the operating systems group,” I said.

  “Based on what I saw, I doubt that very much. I don’t think they’re serious about working with us.”

  “What if they actually do start a project based on Windows?” I was asking myself as much as him.

  Robert looked frustrated with me. “It’ll suck unless they want to make major changes, just like I’ve been telling them. It’s just a plain fact that the requirements for a portable pen computer are very different than for a desktop computer, as you know.” He didn’t have to tell me this; it was the reason we were in business in the first place. But he was getting used to explaining this over and over, like a broken record.

  “Look, Robert, we don’t have time to sell uphill with them, we’ve got other fires to fight. Unless they follow up, let’s let the matter drop at this point. We can make another pass at them when we’re further along.”

  “That’s fine with me.” He scurried off to a meeting about the upcoming State Farm presentation.

  Microsoft never called to arrange a follow-up visit, and neither did we.

  As the date of our presentation to State Farm grew near, I was able to learn that IBM was scheduled to present the day after us. IBM’s strategy was to do what they knew GO could not: lavish attention on the State Farm folks. A corporate jet would whisk them off on a tour of various IBM facilities, where they would be lectured by noted IBM researchers, witness demonstrations of advanced technologies, and experience the power of the sovereign state of IBM. The point would be made loud and clear: IBM is so large and powerful that it can do anything anyone else can do.

  Our only hope was to show a distinct technological edge. The problem was that although we might be able to assemble a working unit by then, and even have some software to demonstrate, there was no realistic chance of putting the two together before the meeting. It was time to develop Plan B.

  I called a meeting with the staff. “Look, it isn’t essential that we show State Farm the software in its full glory. For one thing, we have no way to demo to a large crowd, and you can bet there’ll be one. We’ve put together a paper presentation with lots of pretty images of car parts. All we need to do is give them an idea of what it will look like on the screen.”

  Phil Ydens, the quiet engineer, spoke up. “I think Mike and I can write a special program that will display one image. But you won’t be able to do anything else with it.”

  “That ought to do it,” I said. We picked out an exploded diagram of what we speculated was a drive train, but no one knew for sure—it had lots of wheels floating around a central shaft. Then I went on to the next issue. “The critical thing we really have to have is a working hardware unit. Otherwise, it’s all acetates and yak-yak.”

  Kevin looked worried. “Everything would have to go right from now until then. Based on our experience, I wouldn’t bet on it.”

  “But you already have bet on it!” Celeste said. “That’s why I’m a nervous wreck.”

  She was referring to a challenge Kevin had accepted at the previous board meeting. She had spent the last several months working on a custom chip, known informally as a “glue” ASIC, that knits together all of the components of the system. Once her design was complete, she turned it over for fabrication, a process that would take about eight weeks. John Doerr, skeptical about our schedules, had doubted whether a custom chip this complex was likely to work the first time. If it had to be redesigned, we would be set back at least four months. In a show of confidence, and to demonstrate his faith in his team, Kevin had bet John $100 that Celeste’s GOLD chip, as she had named it, would work on the first pass.

  With less than two weeks to go until the State Farm presentation, the last of the hardware components came back from the manufacturer. We had enough parts on hand for five units, out of which we hoped to assemble two or three working ones.

  Kevin and Celeste turned the hardware lab into a room that looked like a pediatric intensive care unit, with electronic monitoring devices surrounding special clean tables. Probes and wires clung to every quadrant of the precious, fragile circuit boards struggling for life. The whole staff was on call in case someone could provide the missing piece of information or skill that might save a dying unit. Anxious engineers would hover in the doorway, waiting for news, only to be shooed away. To reduce the confusion, Celeste banned everyone from the lab, but periodically sent one of her technicians out with progress reports.

  “The CPU on unit three is booting,” she would announce.

  “Wow, a heartbeat!” yelled one of the engineers. Everyone applauded.

  Every so often the technician would arrive at my office winded from the brief sprint. “We’ve lost three of the units, but Kevin wants me to tell you that we still have enough parts for two more, if we cannibalize those,” she reported.

  After the third day, Kevin called an emergency meeting with me and Robert. He looked like hell. “OK, here’s the scoop. We have two units where everything looks fine except one thing.”

  “What’s that?” Robert asked.

  “They’re both blind—no output on the display. I can tell by probing the board that the signals are there, and I can prove that the displays are working, so the problem has to be in the Yamaha EGA controller. I’ve faxed the Yamaha engineers in Japan several times, but they keep insisting that the chip works fine.”

  I exhaled thoughtfully. “Could it just be the ones we have?”

  “Not likely. They pass all the diagnostics, and all of them behave the same way.”

  Robert looked doubtful. “How can their engineers insist that the chips work?”

  “The problem is that the chips work fine in DOS machines, but the DOS machines don’t exercise all the functionality of the chips,” said Kevin. “Yamaha doesn’t understand that we aren’t just building PC clones, though I keep telling them that.”

  This was the first of many times we were to discover just how fragile computer technologies can be. If you stray from the mainstream, there is no guarantee that accepted, time-tested components will work as expected. It’s a small miracle that computers are reliable at all: they function only as the result of endless trial and error, folklore, and superstition.

  “What should we do?” I asked.

  Kevin was ready with a battle plan. “I’m asking my staff to switch to a different schedule—Tokyo time—so we can interact directly with Yamaha during their working day, which starts around two P.M. our time and goes past midnight. Let’s set up a room here where we can sleep, if necessary.”

  “Consider it done,” I said.

  “Then we’re going in for saturation bombing—phone calls, faxes, demands that their engineers come here, whatever it takes—until they send me the schematics for their chip so I can debug i
t myself.”

  For the next several days, weary engineers labored around the clock, developing ammunition to support the hardware staff during the day and faxing volumes of test reports to Japan in the evening. The team was getting desperate as the deadline approached. They were ready to try anything. Finally Kevin got a schematic out of Yamaha, but only for part of the circuit.

  I entered the lab, violating Celeste’s ban. She was sitting there exhausted and demoralized. Kevin was calm, studying the diagrams yet again. “It’s not going to work,” Celeste said, ignoring my lapse of protocol. “We’ve tried everything. Their goddamn chip just doesn’t meet the spec. See if I ever use their parts again!”

  Kevin silently ignored us. Then he spoke. “Celeste, look at this.” He pointed to the diagram. “This control line is drawn differently from the rest.”

  “So what? They can draw lines any way they damn well please.”

  “Well, maybe it is different,” Kevin said. “Maybe we need to hold the signal high for longer than on the other lines.”

  “You’re nuts!”

  “No, really.”

  Celeste threw up her hands in resignation. “Well, I’m willing to try anything at this point.”

  I returned to my office, thinking about how the State Farm presentation would come off without a working prototype. A few minutes later, Celeste and Kevin walked in. Kevin was calm as always, but Celeste appeared ready to explode. They stood there without saying a word. Celeste looked at Kevin, but he deferred to her, gesturing for her to speak.

  “I can’t believe it. I can’t friggin’ believe it.” Kevin started laughing as she spoke. By now a small crowd had started to gather. “I simply can’t believe it.” She repeated this over and over.

  “OK, OK, so what happened?” I asked her.

  “It worked. Kevin’s crazy idea actually worked. I just can’t believe it. When I designed the GOLD chip, I had a spare flip-flop, so I wired it up to three extra pins, which I labeled EFUFD, EFUFQ, and EFUFR.”

  “What does all that stand for?”

  “Emergency fuck-up flip-flop pins D, Q, and R. I never thought we’d need it, but I used it to change the propagation delay on one of the output lines, like Kevin suggested.”

  One by one, we filed into the hardware lab to see the newborn. The people waiting their turn in line outside could hear oohs and aahs. There on the table, with its insides splayed out for everyone to see, was a pile of electronics with a picture of a drive train on its screen.

  There was only one cherished working unit, and now the team undertook the delicate task of assembling it into the slim, inch-thick black case. Robert took it to his office to practice for the State Farm presentation while Kevin and Celeste went out for a quick lunch.

  Stepping off the elevator when they returned, Celeste suddenly looked up and down the hallway, her eyes filled with fear. “Kevin, do you smell something?”

  He stopped and sniffed. There was the unmistakable odor of burning circuitry in the air. They raced down the hall to Robert’s office, where they found him staring at his desk, looking as if he had just run over a small child. In front of him was the precious prototype, scarred black with smoke, obviously ruined.

  “Software people!” Celeste said in disgust.

  “I was just writing on it when flames shot out through the connector slots,” Robert said.

  Now Celeste was truly ready to give up. But Kevin was unflappable. “Maybe it’s just a hairline short on the circuit board,” he said. “We’ve got enough parts left for one more.”

  They returned to the lab and assembled the final unit, and miraculously it worked. Like proud parents, they brought it to me swathed in soft towels. It looked exactly like the plastic mockup I had been touting for the past nine months. I pressed the on button and the screen came to life. It was pure magic—no one had ever before seen this much computing power packed into such a diminutive, four-pound frame. Someone snapped a Polaroid. I wrote on it: “6/20/89—First Working Unit.”

  Celeste demonstrated how to change the battery. “Now be careful. The battery should last about an hour, so don’t play with it until your presentation. Here’s a spare in case it drains while you’re on your trip.”

  “Not to mention that there’s no guarantee how many times it’ll come on, so I suggest you don’t wake it up unless you have to,” Kevin added. “While you’re away, we’ll try to bring up a second unit.”

  It had come right down to the wire. We were scheduled to depart for Bloomington the next day.

  “These puddle jumpers make me nervous,” Robert said as the pilot circled the airport. After landing, we loaded our gear into a taxi for the short ride to Jumer’s Chateau. Although it was only late June, it seemed that in Bloomington every summer day was the Fourth of July. I pointed out to Robert a mint-condition 1975 red Corvette—I could tell the year from the license plate. I noticed red, white, and blue bunting on a number of buildings, in anticipation of the upcoming holiday. Then a pair of shiny silver Corvettes passed us by. At the next corner, four of the eight cars waiting at the light were Corvettes. Everywhere we looked there were Corvettes of every imaginable color and style.

  “What the hell’s going on?” I asked the cab driver.

  “Huh?”

  “We’re surrounded by Corvettes!”

  “Of course. Isn’t that why you’re here?”

  Robert leaned over to me and hummed the theme from The Twilight Zone.

  I was getting worried. “Hell no. We’re here for a meeting at State Farm.”

  “Oh.” He looked at us in his rear-view mirror. “I thought you were here for the Corvette Corral.”

  The cab driver told us that each June, every Corvette owner within a thousand miles brings his car to Bloomington for this event.

  “What do they do?”

  “Just hang around, talk about their cars, drink beer, swap parts—you know.”

  I didn’t know. I owned a ten-year-old yellow Mazda station wagon, and I couldn’t imagine going to a Mazda station wagon convention.

  The next morning, before our presentation, Robert and I had breakfast at a restaurant called Bob Evans, a bright red wooden structure with white trim, designed to look like the television ideal of a midwestern barn. The inside was decorated with lanyards and bales of hay. The place reeked of cholesterol, and judging from the clientele, they had obviously been eating breakfast there for years.

  Robert bit into a wedge of toast dripping with butter. “I think I can hear my arteries hardening.” Not wanting to buck the theme, we ate eggs and bacon while reviewing our presentation one last time.

  “Do you think we should check the prototype?” he asked.

  “Let’s not risk it. Kevin warned me that it could die at any time. Let’s just leave it in the hands of God.”

  “I’m glad to see you’re coping with all this strain by turning to religion.”

  We were both used to performing under pressure, but this presentation was particularly important. We knew that if we lost the bid to IBM or Hewlett Packard, we were in trouble—back on the fundraising trail with slim hopes for success, given how hard it was to round up investors last time.

  I took the bait. “Please join me in prayer.” I cupped my hands. “Dear Lord, though we walk through the valley of dead batteries, deliver us from EGA, and give us this day our daily bid.”

  “Amen,” Robert added. The elderly couple at the next table looked on approvingly. They apparently thought we were serious.

  By the time we entered the fourth-floor conference room at State Farm, we were both wired on caffeine. Neither of us was in the habit of drinking coffee, but our presentation was scheduled to start at six A.M. San Francisco time, and we wanted to be on our toes. I laid out the overhead slides, handouts, and copies of our response to the RFP in neat piles on the table alongside the prototype, which I placed in plain view. Robert was dismayed—he was hoping for some higher drama. “Don’t worry,” I whispered to him, “they’ll thin
k it’s the plastic mockup. I’ve got a plan.”

  “That’s what Nixon said about ending the Vietnam War,” he whispered back.

  Soon the seats began filling up. I could see that just about everyone who had been at either of the two previous meetings was present. Norm Vincent entered the room last, greeted Robert and me, and introduced us. He took a seat at the head of the conference table, right next to the prototype, where the group could read his reactions.

  I began by reviewing some of the same slides I had shown in the earlier meetings, covering the potential for pen computing, GO, and our product plans. I explained that while I knew most of them had seen these slides, I wanted to be sure we were all starting with the same background information. My real intent, though, was to lull them into a false sense that they knew what was coming next. At the point in my talk where I usually pass around the plastic mockup, I reached over and picked up the live prototype, which looked exactly the same. Several people nodded knowingly.

  “Last week, Norm asked me if I’d have anything new to show today, and I told him no. Well, our engineers have been working around the clock to prove me wrong. You’ve seen this model before, but I have a surprise for you. Today, for the first time, I’d like to show you an actual, real live working pen computer.”

  The audience, which had been characteristically reserved and mannerly, snapped to attention as though someone had shaken their chairs. Everyone strained forward to get a closer look. I stood the unit upright on the table, paused dramatically, and glanced at Robert. He had the look of a defiant death-row prisoner about to give the signal for his own execution. He nodded sharply, then looked toward heaven. I pressed the on button.