A conversation with Shel Kaphan, Amazon’s first employee.
Employee #1 is a series of interviews focused on sharing the often untold stories of early employees at tech companies.
Shel Kaphan was the first employee at Amazon. He is currently pursuing personal interests and still living in Seattle.
Discussed: Getting Online, Prior Startups, Vetting Jeff Bezos, Moving to Seattle, Early Versions of Amazon, Finding Traction Through Netscape, Building the Company, Changing Roles, Life After Amazon.
Craig : What were you doing before Amazon?
Shel : I worked as a programmer starting in 1975. My first real programming job was with an MIT spinoff called Information International, Inc. (also known as Triple-I) that was in Los Angeles, Culver City actually. I went there for a summer job in 1975. I was in college at the time, for the second time around, having dropped out once already, but stayed at Triple-I for three years. In 1978 I decided I should finish my undergraduate degree so I went back to school at UC Santa Cruz for a while and stayed in Northern California until I moved up to Seattle.
Craig : Ok, and how did you end up going to Seattle?
Shel : In early 1994, I had been working at a company called Kaleida Labs, which was a joint venture of Apple and IBM. I left there in the spring of 1994. One of the younger guys who had recently come out of university showed me Mosaic, which was brand new back then. The first time I got onto the the ARPANET, the precursor to the Internet as we know it, was 1969 or 70. I always had the feeling that there was something incredibly cool about it, but for some reason very few people seemed to think it was all that exciting. I couldn’t really imagine where it was all going to go eventually.
When I saw Mosaic, something clicked and I knew the Internet was finally going to open up to a much wider audience. Having missed a couple of earlier waves of technology that didn’t seem that interesting to me at the time, I thought this wave was one that was really going to be interesting and I wanted to do something with it. I didn’t know exactly what, but I knew I wanted to be involved. There were already a few new web-related businesses hiring some of the hotshots that I knew at the time. Netscape had been founded and some people I knew were already working there. It didn’t seem like a good fit for me, if I even could have gotten a job there. I wanted to do a startup and be more involved in the early phases of a business.
I was kicking around ideas with a friend of mine in Santa Cruz, who I had worked with at Frox and then Xerox, Herb Jellinek, trying to figure out some business idea that would be attractive to us that we could work on together–one that we thought might possibly work. We started talking to various people in our networks. Herb had gone to grad school at Stanford and one of his friends from there had gone to work at a hedge fund in New York, which happened to be the same one where Jeff Bezos worked at the time. He connected us with Jeff because he knew that Jeff was going to leave to start a web-related business that he had analyzed for this hedge fund. For whatever reason, that company didn’t want to pursue it but Jeff did.
Herb talked to Jeff, and then Jeff flew out to meet us in Santa Cruz. We had breakfast together and Jeff told us about his idea to start an online bookstore. We were even talking about possibly locating it in Santa Cruz. This was in spring of ‘94. Jeff went back home to New York and started thinking about where he wanted to locate. We were looking at office space in Santa Cruz but as he learned more about mail-order business he eventually decided it made more sense to be in a smaller population state or one that didn’t charge sales tax. He narrowed it down to someplace in Nevada or Seattle. I pretty much knew that I wasn’t going to be moving to Nevada. When he finally decided on Seattle, it took him all summer to convince me to move because I had lived in Santa Cruz for nearly 20 years and I liked it there.
Eventually, I decided that there was enough that I wanted to do on the project that it made sense to move. Herb, who had only recently moved to Santa Cruz, decided to stay. At first I was a little bit tentative about it–I kept my house in Santa Cruz and I only moved the minimum amount of stuff I needed to live.
At the time I thought, “Okay, I’m going to be building this website to run a bookstore and I haven’t done that before but it doesn’t sound so hard. When I’m done with that I’m not sure what I’ll do.” At that point there was no idea of doing anything but a bookstore. I thought maybe I would be able to go back to Santa Cruz and monitor it from there. I was pretty wrong about how the business would develop and how ambitious Jeff was. I didn’t know him at the time. We had just met.
Craig : How did you vet each other? Was he technical? Had you worked with many entrepreneurs before?
Shel : I’d worked in a few different startups. We both gave each other references to check out. But I think my choice was mostly based on the sense I had that he was fairly likely to make something work. I had been in a number of startups where there was an absence of people with sufficient business intelligence who understood how to fundraise, market something, and make business plans that weren’t based on hopelessly complicated technology that was super interesting but that nobody was ever going to pay for. I liked the idea that it was a very straightforward sounding business. I liked that I could explain to people who the customers were likely to be, what they were going to be paying for, and how the company could pay its own way.
I liked the guy when I met him. He has a very engaging personality. I was excited about the project so there were other reasons too besides the people involved. At the time there was this confluence of networking, hypertext, graphics and all this stuff that was coming into play with the web for the first time.
I wanted to take Books In Print or something like it and make a hypertext version. I’d been thinking about that before I even met Jeff. I wasn’t thinking about it in the context of selling books, but I was thinking, “Man, I hate going to the library and ruffling through those card catalogues and trying to find that thing that I’m looking for.” Nobody alive probably remembers that anymore. You actually had to go through drawers full of index cards to find books you were interested in. Then you’d walk around the library and browse the shelves to see if maybe the thing you’re interested in would be nearby.
I thought solving that problem was a perfect application for hypertext. There were a couple other online bookstores popping up around that time, too. Nobody seemed to grasp that issue particularly well. So, I thought, “Okay, well, I really want to build this and this project is a chance for me to do it.” Plus, I had worked just after high school in a mail order operation that sold books and other things, so I felt like I was going back to my roots, and that also felt good.
Craig : What was the first thing you guys had to build to get started?
Shel : Well, there was basically nothing except for a little library from NCSA of primitive things that you could use to build slightly interactive websites. I started by building up machines and getting a database system and putting together a little development environment. There weren’t any cloud services or anything like that at the time. It was all build your own and run it yourself. There was precious little in the way of tools or development environments for web stuff or libraries to build things out of. It was all cobbled together. At that point the web was a very static thing–mostly just a collection of pages.
There were hooks in the HTTP servers for running scripts, which is what we were depending on because all of our pages had to be dynamically generated. There was really nothing that existed to build a stateful application. In other words, where you are serving different things to different people and you have to keep track of that person’s progress, as they’re adding things to a shopping basket and going through an ordering process and all that. We pretty much had to figure that out and learn how to do it.
Craig : How did you troubleshoot? Today I use Stack Overflow constantly. What would you do when you ran into a bug that you couldn’t figure out?
Shel : Stay up late.
Craig : [Laughter] Fair enough.
Shel : I don’t recall that we had a lot of help from outside parties. At one point we decided to switch from Sun Microsystems to Digital servers. I was more familiar with Sun’s machines at that time so when we got the Digital machines and had some performance issues at first, Jeff found a professor at UW who could help with kernel tuning, which I was not terribly familiar with.
Everything else was debugging as usual. Technically speaking, Amazon was really pretty straightforward to build in the beginning. Although believe me, we had our share of bugs. But they were mostly relatively straightforward.
Craig : Were you also doing the software for inventory management?
Shel : A month after I got there we hired a guy named Paul Davis who had been in the Computer Science staff at UW, and who was a really great hacker. He and I worked together quite well. I was mostly focused on the website and customer-facing stuff. He was primarily focusing on shipping, receiving, inventory, charging credit cards and all that kind of thing. But we both had our hands on all of the code. He only stayed for a year and a quarter or so. When he left, it was just me for several months before I could hire anybody else. I was doing all of it at that point in time. I remember I worked seven days a week for 3 months straight, and they weren’t 8-hour days. Then we started hiring a few people who could take over specific aspects of the code.
Craig : Do you remember how many orders you were getting per month at the time?
Shel : I don’t remember numbers but it was minuscule by today’s standards. That said, our business was doubling quarterly for about six quarters or more in a row. In the beginning before we got the Digital servers, we ran the whole business on a couple of small Sun desktop machines. That was everything. We didn’t have much of a budget for hardware. We were trying to get absolutely everything we could out of a small number of tiny machines. We were always a little bit behind the curve on adding more hardware as we needed it.
Craig : Did that ever come around to bite you guys?
Shel : Yeah, for sure. We were always up against the limit of what the hardware could do. For one reason or another, sorting out architectural issues to scale more gracefully was something I could never convince Jeff to allocate resources to do. There were always too many customer-facing features that needed to be developed.
There were times when some piece of hardware would crap out and corrupt a database and, of course, some of the backups hadn’t been working. But somehow we survived.
Craig : Were you guys running any kind of analytics at that point?
Shel : No. Not in the beginning. I think it was maybe spring or summer of 1997 when the first people came in that were starting to work on that. In the beginning we were saving our server logs thinking they’d be really interesting to analyze, but not right now.
Craig : When we have a ton of extra money and time.
Shel : Yeah, maybe when we have a thousand extra programmers or something.
Craig : Exactly. So had you been working with Jeff on the product? How was the ship being steered?
Shel : During that time we were just a bookstore, so it didn’t seem to me there was a lot of steering to be done. That said, I don’t know all the things that Jeff might have been doing that I wasn’t aware of at the time. If I think back, I can’t even clearly picture what it was that he was actually doing. He wasn’t working on any of the technical stuff. We never even had a written business plan that I know of.
Craig : [Laughter] Was he sourcing the books?
Shel : He wasn’t doing that. Well, maybe he was at the very, very beginning before we hired people to interact with publishers. At the very beginning we were mostly just working with distributors. But we wanted to have a large catalogue, so we also had to work directly with publishers who weren’t represented by the distributors. That was what allowed us to claim a million titles, which was a big deal back then.
Craig : How were people discovering you guys?
Shel : Well, Google wasn’t around at the time. There are probably other opinions about this but, first off, there weren’t that many websites that were interesting back then. Second, as it turned out, there was a couple who worked at Netscape at the time–Eric and Susan Benson. I had worked with Eric in three different places by then. We had a “friends and family” soft launch in the spring of ‘95. They were among the people that were trying it out. We later hired both of them at Amazon. Susan was working on Netscape’s website in an editorial capacity. I only learned this last year, but when we opened up to the public it was she who put Amazon on their “What’s New” and “What’s Cool” pages. She put us on those lists. Then because the name started with an A, it was above the fold so lots of people saw it. That was, in my opinion, a super important connection for us. It might have happened without the personal connection, but who knows, maybe not.
Craig : That’s wild. Were you shipping to all 50 states in the beginning?
Shel : Yeah, and several foreign countries as well. We had a lot of international orders from fairly early on.
Craig : At what point did you realize, “Maybe I shouldn’t keep my place in Santa Cruz. I might be at Amazon for a while.”
Shel : I think that happened two years in or so. It started to be annoying to have to manage the house remotely, even though it was a friend of mine who was renting it. I was still responsible for it. At that point I also thought, if I do move back there, I’m probably not going to live in that house. It was not a particularly great house. I decided it would make my life a little simpler to get rid of it.
At that time we started extending into different product areas, too. We started having sites in a couple of other countries. Germany and England were the first two. And we started to make acquisitions. I think the acquisitions were probably post-IPO, which was spring of ‘97.
Craig : Whoa. I didn’t realize it was that fast. Three years?
Shel : Yeah, about two and a half.
Craig : What was your role around the time of the IPO?
Shel : My role was always primarily technical with some technical management. At that time I was VP of Development and was responsible for writing the software and keeping the systems up and running. One of the people who were hired to replace me in my original role came in only a month or maybe two before the IPO. The next guy came in September of that year. Early on, the small team of technical folks had been working for me, and I was mainly doing a lot of programming myself, but also system administration, network configuration, and so forth. I was putting disk drives in enclosures and running Ethernet cabling around the building and that kind of thing.
Craig : Not quite the glamorous startup life some people imagine.
Shel : Well, yeah. Those days were still before everything that’s happened with glorifying startups. If you were going to do a startup business, there wasn’t a huge expectation that it was going to be glamorous in any particular kind of way. You were going to work really hard and maybe it was going to work, though probably not.
Craig : Are there any vestiges of your work at Amazon?
Shel : Up until recently, maybe a few years ago, I could have said yes. But I don’t really see anything that looks like it now. It seems highly unlikely that anything I did actually still exists over there anymore.
Craig : What about even design patterns? Like, this is how a shopping cart works.
Shel : I don’t remember learning that from some other website. Though it seems like, as soon as you start thinking about that problem, it’s self-evident what it has to be like. You don’t want to make people go through a transaction every time they pick something. You have to let people go through the site and pick things that they’re interested in, put them somewhere so they can come back and get it all shipped at some point. I remember doing a lot of framing of that process on the website. Putting in text that said you could always take this out of your shopping basket later if you change your mind. So people wouldn’t feel they were overly committed.
Back in those days nobody was used to being a customer of online businesses. You had to be careful to make people feel comfortable and let them know their actions were reversible. Even though I was mainly doing technical things, the appearance of the website and a lot of the text on it were my doing in the early days. I was careful about making things acceptable to what I understood the culture of internet users to be in those days.
Craig : That’s so wild. You were designing for a completely different level of knowledge. You can assume so much more today.
Shel : Yes. Absolutely. Back in those days, it was the very beginning of doing commerce on the internet. There was a whole debate around allowing commercial activity on the internet. At least that’s how I recall it. Many people online were like, “We’re not so sure about commerce on this thing. You better not overdo it. You better be tasteful.”
There were these cases where people doing overtly commercial things were chastised by the community at large. There were some lawyers advertising their services for getting Green Cards and they were doing it by massive spamming activities. It created a huge fuss back then. It’s laughable now because so much stuff like that happens all the time.
And when cookies started being a feature in web browsers a lot of people were really concerned about their privacy, so they would turn them off. So we had to figure out how to make things work without that. Some people were running text-only browsers back then. Plus they were on dial-up. Sending pictures was not a good thing for those people.
Craig : Wait, you could use the early Amazon as a text-only site?
Shel : Yeah, we tested it. It always seemed important for us to make things continue to work, even for people that didn’t have a high bandwidth connection or the latest and greatest computers and all that.
Craig : How long did you end up staying in Amazon?
Shel : I was there for five years.
Craig : At the time you left, was it still a bookstore? What was it like?
Shel : Well, they had branched out into several other product areas. There weren’t any digital products yet. Ebooks hadn’t happened yet. They hadn’t developed any hardware products yet. Their computing services were not public yet. It was still pretty much a retail business. Although it was definitely branching out into other product areas and other countries.
Also around the time of an IPO and afterwards, the kind of people that are attracted to go to a company changes a lot. There were boatloads of MBAs and people like that coming. It was already a big company from my perspective.
Craig : So you’ve observed all these technological shifts. Do you have thoughts about how people should evaluate technologies when they’re about to start building something?
Shel : At this point, I don’t know. It’s a huge subject. For myself, when I look at technology these days, I see that it’s either doing something to connect people or it’s doing something that isolates people. I tend to make value judgments based on that kind of consideration about what is worth working on.
You walk down the streets, you have to weave around people standing there in random orientations in the middle of the sidewalk looking at their cellphones. Then you see people speaking robotically so that their speech recognizer can understand them. Now they are running around in mobs in parks with their phones in front of them trying to catch imaginary animals. I don’t necessarily see all that as a positive development.
I think technology has a role to play but I don’t see it being exploited very carefully in that way. But this is the kind of economy that we live in. And it’s very, very addictive. Even people who complain about it are still subject to it.
Craig : For sure. Where do you see Amazon falling on that spectrum now?
Shel : I think that a lot of what they do is more on the isolating people side. Everything caters to convenience so much that you don’t even have to get out of bed to take care of your day-to-day business. To me, that’s a step too far. Of all the major online businesses, I don’t really think that they extend in that way much beyond what we did very early on by allowing for customer reviews.
Craig : Had you and Jeff stayed close?
Shel : Not really. When he replaced me in my original job and I was moved into the CTO slot, I was nominally in charge of architecture, but in fact that just meant rubber stamping projects that were 95% complete by the time I saw them. That was all after having told me that my job was mine as long as I wanted it. And I didn’t have resources other than myself to work on anything I was interested in either. So I would say we were not really on particularly friendly terms at that point.
Craig : In retrospect, how do you feel about how things unfolded with Jeff and Amazon?
Shel : He’s obviously a super brilliant businessman. If I had any inkling about what kind of a company Amazon would turn into, both in terms of how successful they are and some of their business practices, I probably would have been a little bit more careful about my own relationship to it at the beginning. I also might have decided it wasn’t what I wanted to do.
One thing that the Amazon experience taught me is try to imagine what a project or company would be like if it was more successful than you could ever possibly imagine. It’s very unlikely but it’s possible. You have to think about what the environment will be like if that happens, and how the people involved in it might change. When I was joining Jeff to form Amazon in the beginning, I didn’t even allow myself to go there. I’d worked for a lot of startups so it almost felt like a jinx to think too much about what might happen if it really succeeded in a big way. That was my mentality. I was like, I hope this makes it and is a moderate success. Maybe it even generates enough cash to let us retire at some point. You don’t really want to think about massive success beyond what you can imagine. Then, if it is successful, you have to start thinking, what’s my role in enabling this? Is that something I really want to be doing?
I would say, of all the jobs I had, and I had quite a few between when I started programming and when I left Amazon, the first couple of years at Amazon really were a high point for me. I really, really liked that. A lot of that wasn’t so much because of the technical side of it, it was because of having worked in lots of small businesses that didn’t go anywhere or were the wrong thing at the wrong time or something like that. Being a part of something from the very beginning that engages people and has an astonishing growth curve–being part of making that actually work–was hugely satisfying, and I still look back on those first couple of years as a really exciting and great time in my life.