At that time software, like computer hardware, was expensive. It had to be written specifically for each kind of computer. And each time computer hardware changed, which it did regularly, the software for it pretty much had to be rewritten. Computer manufacturers provided some standard software program building blocks (for example, libraries of mathematical functions) with their machines, but most software was written specifically to solve some business’s individual problems. Some software was shared, and a few companies were selling general-purpose software, but there was very little packaged software that you could buy off the shelf.
My parents paid my tuition at Lakeside and gave me money for books, but I had to take care of my own computer-time bills. This is what drove me to the commercial side of the software business. A bunch of us, including Paul Allen, got entry-level software programming jobs. For high school students the pay was extraordinary—about $5,000 each summer, part in cash and the rest in computer time. We also worked out deals with a few companies whereby we could use computers for free if we’d locate problems in their software. One of the programs I wrote was the one that scheduled students in classes. I surreptitiously added a few instructions and found myself nearly the only guy in a class full of girls. As I said before, it was hard to tear myself away from a machine at which I could so unambiguously demonstrate success. I was hooked.
Paul knew a lot more than I did about computer hardware, the machines themselves. One summer day in 1972, when I was sixteen and Paul was nineteen, he showed me a ten-paragraph article buried on page 143 of Electronics magazine. It was announcing that a young firm named Intel had released a microprocessor chip called the 8008.
A microprocessor is a simple chip that contains the entire brain of a whole computer. Paul and I realized this first microprocessor was very limited, but he was sure that the chips would get more powerful and computers on a chip would improve very rapidly.
At the time, the computer industry had no idea of building a real computer around a microprocessor. The Electronics article, for example, described the 8008 as suitable for “any arithmetic, control, or decision-making system, such as a smart terminal” The writers didn’t see that a microprocessor could grow up to be a general-purpose computer. Microprocessors were slow and limited in the amount of information they could handle. None of the languages programmers were familiar with was available for the 8008, which made it nearly impossible to write complex programs for it. Every application had to be programmed with the few dozen simple instructions the chip could understand. The 8008 was condemned to life as a beast of burden, carrying out uncomplicated and unchanging tasks over and over. It was quite popular in elevators and calculators.
To put it another way, a simple microprocessor in an embedded application, such as an elevator’s controls, is a single instrument, a drum or a horn, in the hands of an amateur: good for basic rhythm or uncomplicated tunes. A powerful microprocessor with programming languages, however, is like an accomplished orchestra. With the right software, or sheet music, it can play anything.
Paul and I wondered what we could program the 8008 to do. He called up Intel to request a manual. We were a little surprised when they actually sent him one. We both dug into it. I had worked out a version of BASIC, which ran on the limited DEC PDP-8, and was excited at the thought of doing the same for the little Intel chip. But as I studied the 8008’s manual, I realized it was futile to try. The 8008 just wasn’t sophisticated enough, didn’t have enough transistors.
We did, however, figure out a way to use the little chip to power a machine that could analyze the information counted by traffic monitors on city streets. Many municipalities that measured traffic flow did so by stringing a rubber hose over a selected street. When a car crossed the hose, it punched a paper tape inside a metal box at the end of the hose. We saw that we could use the 8008 to process these tapes, to print out graphs and other statistics. We baptized our first company “Traf-O-Data.” At the time it sounded like poetry.
I wrote much of the software for the Traf-O-Data machine on cross-state bus trips from Seattle to Pullman, Washington, where Paul was attending college. Our prototype worked well, and we envisioned selling lots of our new machines across the country. We used it to process traffic-volume tapes for a few customers, but no one actually wanted to buy the machine, at least not from a couple of teenagers.
Despite our disappointment, we still believed our future, even if it was not to be in hardware, might have something to do with microprocessors. After I started at Harvard College in 1973, Paul somehow managed to coax his clunky old Chrysler New Yorker cross-country from Washington State and took a job in Boston, programming mini-computers at Honeywell. He drove over to Cambridge a lot so we could continue our long talks about future schemes.
In the spring of 1974, Electronics magazine announced Intel’s new 8080 chip—ten times the power of the 8008 inside the Traf-O-Data machine. The 8080 was not much larger than the 8008, but it contained 2,700 more transistors. All at once we were looking at the heart of a real computer, and the price was under $200. We attacked the manual. “DEC can’t sell any more PDP-8s now,” I told Paul. It seemed obvious to us that if a tiny chip could get so much more powerful, the end of big unwieldy machines was coming.
Computer manufacturers, however, didn’t see the microprocessor as a threat. They just couldn’t imagine a puny chip taking on a “real” computer. Not even the scientists at Intel saw its full potential. To them, the 8080 represented nothing more than an improvement in chip technology. In the short term, the computer establishment was right. The 8080 was just another slight advance. But Paul and I looked past the limits of that new chip and saw a different kind of computer that would be perfect for us, and for everyone—personal, affordable, and adaptable. It was absolutely clear to us that because the new chips were so cheap, they soon would be everywhere.
Computer hardware, which had once been scarce, would soon be readily available, and access to computers would no longer be charged for at a high hourly rate. It seemed to us people would find all kinds of new uses for computing if it was cheap. Then, software would be the key to delivering the full potential of these machines. Paul and I speculated that Japanese companies and IBM would likely produce most of the hardware. We believed we could come up with new and innovative software. And why not? The microprocessor would change the structure of the industry. Maybe there was a place for the two of us.
This kind of talk is what college is all about. You have all kinds of new experiences, and dream crazy dreams. We were young and assumed we had all the time in the world. I enrolled for another year at Harvard and kept thinking about how we could get a software company going. One plan was pretty simple. We sent letters from my dorm room to all the big computer companies, offering to write them a version of BASIC for the new Intel chip. We got no takers. By December, we were pretty discouraged. I was planning to fly home to Seattle for the holidays, and Paul was staying in Boston. On an achingly cold Massachusetts morning a few days before I left, Paul and I were hanging out at the Harvard Square newsstand, and Paul picked up the January issue of Popular Electronics. This is the moment I described at the beginning of the Foreword. This gave reality to our dreams about the future.
