Tech Brew Ride Home - BETTER SOUND! (IHP) Amazon's First Employee Shel Kaphan

Episode Date: May 22, 2023

Re-releasing this one, hopefully with improved sound quality! The earliest days of Amazon with its earliest hire: Shel Kaphan. Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Internet History Podcast. I'm your host, Brian McCullough. Shell Caffin was the very first person hired by Jeff Bezos to launch Amazon.com.

A lot of people consider Shell to be a co-founder of Amazon in all but title because he, along with Paul Davis, was largely responsible for the entire technical architecture that Amazon launched with, from the website to the back-end systems that made selling books on the internet possible. I was thrilled when Shell agreed to talk to me because he does not give a lot of interviews, and I knew that he could shed some light on the earliest Amazon details that absolutely no one else in the world could. Shell gives us the background on everything from the earliest commerce systems to the development of Amazon's famous review and recommendation systems.

This is such a fascinating, detailed look at Amazon's earliest beginnings. And I think it reminds me of all the great details we got from the Mosaic and Netscape engineering teams when we spoke to them in the earliest interviews for this project. So I know you're going to love episode 50, an interview with Amazon's employee number one, Shell Kaffin. Shell Kaffin, thanks for coming on the Internet History podcast. I'm glad to be here to talk to you. So from my research, not only are you an alumni of UC Santa Cruz, but you're also a Santa Cruz native, is that right? Well, not exactly.

I moved there to go to college in 1970, and up until the time I moved to Seattle to work on Amazon, I lived there most of the time, with the exception of a few years in first Palo Alto, which is closer to where I came from, and then my first computer industry job was in Los Angeles, so I was going

for that for a few years or so. And your degree was in mathematics? That is correct, just an undergraduate degree. So, let's start with your career with the little, the

sideline here of your time with the whole earth truck store. Oh, okay. Well, I grew up in Menlo Park. That's where we moved when I was about 10 years old, and we stayed there until I was, I moved off to college. And part of what was happening around there in the late 60s was an organization called Portola Institute, which through some high school friends of mine, I got to know the people that ran that, and one of the projects out of Portola Institute was the Whole Earth Catalog. And Portola Institute, we used to spend some time in their offices because they let our little high school rock band practice there,

and the Whole Earth Truck Store was just across the alley from there. So, you know, I got familiar with that. And also when I first saw the catalog, the first catalog, which came out in 1968, was really, you know, I got very interested in it and in a lot of the ideas that were being propounded in it and the people that were making it. And, you know, some of my friends from school

were working on typesetting for it. So in the summer between my senior year and when I went off to university, I worked there for a few months. I just walked in there one day and asked the guys who were running it if it would be okay if I worked there, and they said, sure, so I did. So that was my first exposure to the mail order world,

in particular mail order books. And I worked on various, you know, rather mundane aspects of that business, although it was a pretty fun place to be at that time, various, you know, people coming through. all the time and you know people that were running it were pretty interesting also so you you were behind the register and things like that i did that i did some subscription fulfillment i did you know some um just packing shipping and packing books and um i did some bookkeeping and

you know taking the taking the daily receipts off to the bank and that kind of nonsense it's it's it's fascinating to me how many ways and in how many stories, you know, Stuart Brand and whole earth people weave into all these different stories that we've been covering so far. So moving on to after college, moving into the 80s and the 90s, you worked at several different startups. We don't have to go too far into them, but one was Lucid, which was an artificial intelligence software. Almost. It was a, it was a, it was a, it was a, It was a LISP systems company, which in the mid-80s, that was the era of the Symbolics LIS machine and all of that.

A lot of people were in love with for various reasons. And as the commodity workstation market started, you know, coming into place, a couple of LISP systems that got built for that, and Rucid was one of them. And, yeah, so I worked for them for about a little more than three years from... 85 to, I guess, 89 sometimes. And another one that caught my eye was FROX, I guess, which was some sort of home multimedia theater type deal? Yeah, that was an attempt to make a home theater before the idea of home theater existed.

So it was grasping in the dark, and there wasn't really a market for it, and the company had trouble getting financed. So that, again, was an interesting company. The guy that ran it was named Hartman Eslinger, who was the founder of Frog Design, which is an industrial design company that's done a lot of work for Apple and other companies. Right, right.

And so then the job that you have in the early 90s prior to the story we're going to delve into was that joint venture between IBM and Apple, right? Yeah, one of them, there was a couple at the time. It was called Colla Lab. And it was basically building a platform for publishing multimedia content on. It had a scripting language and a storage layer and that kind of thing. And I was working on some of the storage layer was what I was looking on for that. So going down that list, and I know we left some other companies out and things like that.

But to be clear, you hadn't been working on anything that was internet or especially obviously web-related up until this point. Well, the web didn't really exist. Exactly, right. You know, I mean, I saw the first, I saw a mosaic, which was the first graphical web browser. One of the younger guys at Collida brought that in, I guess seen it, earlier versions of it at his school. And so that, you know, when I saw that, it was a light bulb kind of went off that, you know, I had been using the Internet for a long time even before it was the Internet. And, you know, it always struck me as one of those things where I just thought it was amazingly cool.

And I didn't really understand why everybody else didn't also think it was amazingly cool. But when I saw the, when I saw Mosaic for the first time, And it made me realize that, you know, that, of course, the reason that most people had not been using it is because it had all these very arcane interfaces that, you know, along the lines of, you know, text-based Unix systems and things like that used to give people. And so it, but it suddenly became pretty obvious that this was going to be a game changer for people being interested in using. the internet. So I knew that it was going to be a big wave of some kind. I don't think I really knew how big it was going to be, but I knew it was going to be something transformative, so I wanted to get in on it pretty soon. And I get the impression that around this time, also you're sort of on the lookout for something new to do and a new wave to catch, because

you were kicking around startup ideas at this point? Well, I left Collider. at a certain point when it sort of became clear to me that it was not too likely to succeed. There were always a lot of, you know, kind of difference of opinion between Apple and IBM about what it should do. And it just was kind of a weird hybrid that wasn't really a, you know, either company and it wasn't really a startup either. So it was kind of, you know, it was interesting work for a while, but I left there and I was definitely after having left, I was starting to think about how I could get involved in the

web, which was, you know, in its early stages. And like everybody else, I was struggling for some idea of what kind of business would make sense to do with it. And I had a friend who was also, I'm trying to remember if he had left his previous job or not, but he and I, we both lived in Santa Cruz and we had worked together a couple of times at Prox and at Xerox and we were just kicking around ideas about things that might make sense to do and that's when we started, you know, realizing that we needed to have more of a business guy involved with us and we started talking to people in our networks to see if we could

find anybody to work with. We were basically primarily technical guys. We kind of knew that we needed somebody to help define a business plan and raise money and manage marketing and all of that kind of thing for whatever it was who was going to do. So that was when we kind of started the process that ended up with us meeting Jeff. How are the introductions with Bezos made? Okay, well my friend, his name is Herb Jelanek, and he lived still in Santa Cruz, or south of there, and he had gone to grad school at Stanford.

And one of his friends from there was one of the early programmers, maybe even the first one at D.E. Shaw, where Jeff worked. Right. So Herb talked to him, and he knew that Jeff was going to be leaving to do something, because Jeff had been analyzing business. plans for D.E. Shaw, and one of the ones he had analyzed was, you know, being a bookseller. And so that guy put Herb in touch with Jeff, and Jeff came out to visit us in Santa Cruz, and we went out for breakfast and even ended up looking for looking at different office space

areas anyway around Santa Cruz and thinking about locating there once he got interested in us in us. But he did, you know, he eventually made job offers to both of us and then, well, I can't remember the exact sequence, but he also decided he didn't want to locate in California. So he had a couple of other possibilities in mind. One of them was going to be some unspecified place in Nevada and the other one was going to be Seattle. And had it been Nevada, I think the odds of me having to be Nevada, I think the odds of me having decided to make that move from Santa Cruz would have been a lot lower. As it was, it was a pretty difficult decision for me.

We were talking in the spring, and I didn't move up here in 1994. So what was it then that made you decide to jump on board? If it was a tough decision, was it you like the idea of selling books specifically? You like the idea of doing the web, or was it that Jeff seemed to be like a guy that could be successful doing something? Well, it's a combination of things. For one, one of the things that I'd been thinking about that I thought the web would be great for would be building a hypertext version of essentially books in print or the card

catalog at your library, if any of the listeners are old enough to remember those. And, you know, just to make it easier to kind of navigate the space of books. And so I'd been kind of noodling on that idea even before meeting Jeff. And so one of the reasons I wanted to do Amazon was because that was going to be, you know, a chance for me to do that I had already wanted to do. Plus, you know, it also connected with something from my past about selling books mail order. I've always been a book lover, and I really liked the idea, you know, It seemed kind of like a continuation of the Holder's catalog idea of making tools essentially

available to people wherever they might be and, you know, just making, to put it in Stewart Brands, for it's accelerating access. And then, yes, Jeff, you know, made a good impression on me. And, you know, I thought it was a good sign that to have somebody involved that already kind of understood how to operate in the financial world, and I thought that would probably make fundraising a lot easier. And he did seem like somebody, you know, he was a very young vice president at one of the most technically advanced hedge funds in New York that already had quite a reputation. So, you know, I thought, well, he's probably housed a certain gene for success in him.

So, you know, with all those reasons and probably others that I can't think of. I think I also was just ready to do something different because I really was kind of jobs, even in, you know, ostensible startups around the valley. The kind of jobs that I could get were not really something that was exciting me much anymore. And I needed to have some kind of a change. So when you do finally, after several months, make the move up to Seattle, you're still the only hire at this point, right? it's just you, Jeff, and his wife, McKenzie, and working out of their garage, right? Yeah, it was, yeah, at this point, they had just rented a house in Bellevue,

and they were going to, the first office was in the garage, which had already been converted. So it was really just a not particularly well-heated part of the house. And, you know, we didn't have anything. When I got there, it was basically just not even on a business plan on paper. it was a couple of spreadsheets and a verbal description of it. So, you know, I kind of knew what I needed to do, so I just set about doing it and started buying some computers and shopping for database software

and setting up a development environment and that kind of stuff. And how soon did Paul Davis get recruited to come on board? He got there about a month after I did. I started in late October of 94, were, and it was about a month. Were you involved at all in recruiting him and hiring him? I didn't make the, you know, a professor that Jeff knew, I guess, knew that he was on staff at UW, and he made that he introduced Paul to us, and I certainly talked to him,

and, you know, we seemed to be compatible, so I was involved in the decision, yeah. So before we delve into the systems that you set up, because that's what I want to do somewhat of a deep dive on, still around this time they're kicking around various names, and the leading contender at the time was Cadabra, and apparently you were not a fan of Cadabra at all. Well, I could have the sequence a little bit wrong here, but I recall being surprised after having moved up here that the name of it was going to be. Cadabra. And I really had a sort of thinking feeling about the whole thing after that, which I didn't really voice terribly much. But at a certain point, I guess it was the lawyer, the corporate lawyer that Jeff had hired or, you know, contracted with to work on the incorporation

paperwork and stuff. Either he or somebody he had talked to had heard it as cadaver on the phone. that, I guess, was enough to convince Jeff that maybe it wasn't the greatest name ever. And that, to me, that was also funny because I had one of my previous, in one of my previous lives, I was part of a little consultancy that never quite got off the ground that was called the Cemetery Group, which was sort of time on a mathematical concept. And people on the phone would hear that as the cemetery group. So I thought, oh, my God, this is falling. But how did you feel about the ending up with the name Amazon?

I was okay with it. You know, I've been involved in enough naming processes over time that, you know, it's very hard to find good ones. It's hard to find, it's almost impossible to find names that everyone agrees on. And, you know, after a few things got kicked around that really weren't so great, But one day I got there and just come up with Amazon.com as the name and made it pretty clear that it wasn't up for discussion. And I was cool with it.

I didn't love it, but I thought it was okay. So let's get into what you're doing at this point. As far as I can tell, you didn't have any previous experience with web development generally, but then at this point not a lot of people did, right? Yeah, and the people, most of the websites at that time were just static collections of pages. It was just HTML, so the whole thing of, you know, backing up a website with a, there was very little experience doing that. There were some hooks in the early web servers that let you run scripts, but doing anything that kept of what the user was doing, that was a little bit new. I would hesitate to claim that we were the first ones that did anything with that.

We kind of had to roll our own mechanisms for doing that. Were there packages available from the NCSA and stuff at that point? They were the ones who produced both Mosaic and the first HTTP server that we were using. So the development that you start with, you're mostly working in, C and Pearl, is that right? Mostly C. Some pearl sort of glue various things together, but mostly C.

And what about the databases? Neither Paul nor I were relational database experts, really. So I think I started shopping for databases before Paul even got there, and at that point in time, the two contenders were Oracle and Sybase. space didn't return my call, so I ended up with Oracle. And what's the main focus that you're working on first?

Is it creating the front end or getting the whole back end in terms of the catalog and that sort of stuff set up before you start worrying about like, you know, the user end of it? Mentally divide things up, the catalog was actually part of the front end because it was part of the customer-facing aspect of things. but I kind of, Paul and I kind of divided up to work. Mostly he took care of the back end, which was, you know, the warehouse functions, receiving, storing, and shipping books. And I took care of the website and the book catalog and pretty much all the directly customer-facing software.

And so the catalog is built off. of using like the books and print CD-ROMs and catalog stuff from Ingram and Baker and Taylor, right? Books in print. Okay. The only things we really had were the inventory data from Baker and Taylor and Ingram. And Baker and Taylor's inventory database

was, I'm pretty sure it was derived from the books in print data. But, you know, we had to do stock information in their data at different intervals. And so I had to come up with some kind of process to rationalize not only the availability information, but also just bibliographic information, which would be, you know, it was full of mistakes and had to be, we had to have some processes for reconciling that. I had read that a lot of the early work that you were doing was also maybe prioritizing email, which made sense to me because at this point I don't think even things like AOL and CompuServe had turned on the web yet.

So maybe you were expecting that most of the transactions would come over email. Is that right? Store in an email-based store. Paul kind of did a few things to support the possibility of that, but in opinions about things. I kind of ignored the whole email deal. We had our hands full making the website work and just never got around to that.

And I didn't think it was ever going to work anyway, so I didn't. So as you're designing the website, bandwidth has to be an issue as well, not on your end as much as on the user's end because, again, we're back in the day of 14-4 modems and things like that. Yeah, in fact, you know, we, one of the text-only browsers was one of my test cases to make sure things would work for people that were really on highly bandwidth-constrained connections.

Yeah, and, you know, a lot of it was still dial-up, yeah, at that point. So, yeah, we, you know, first off, we didn't have a lot of content that was going to soak up a lot of bandwidth, so that was okay. And, you know, even so, we tried to design things so that it would, you know, we'd get the most bang for the buck in terms of the number of the size of the pages that we were. Because it was an issue then. Took a long time. And what about the, because I haven't even asked now about the actual commerce side of the equation. So how were you originally set up to take orders and to take money from customers? People could enter their credit cards in our forms in the order process, or they could enter

just the last five digits of their credit cards and then get us the full credit card number, which we could match elsewhere. Some people would even just email their full credit card number to us as if that was somehow more secure than entering it on a form on the web. and some would call our phone number and leave it to us that way. And so we had some customers would match up full credit card numbers to the last five digits and, you know, insert them into the order information in the database that way. Then when it came time to charge people's credit cards, our first system was a batch

system that ran on a PC. we had to write out a floppy disk and walk it across the room to it and do it that way. And it was quite a while, actually, before we had enough business to justify a full-time connection to a... Well, were you also doing that to keep the numbers secure? Because that would make sense if the computer's not connected, then it can't get stolen. We did take steps to keep credit card numbers secure, but that was a different set of steps. We had a system that was called Credit Card Motel because credit cards could check in but not out. That it was connected via a serial line to our servers that ran, it all ran a kind of a one-way protocol

where you couldn't actually retrieve the credit card number, you know, that link. I guess that was Jeff's idea to do that. That's how we did it at first. I have no idea how it works now. So we basically had a system that had that kind of sensitive information on it. And from that one had the tools that let us create the floppy disk at first and later ran the connection to a credit card process. This is jumping ahead just a bit, but I was just thinking,

if you have any memory of the beta test especially, but even the first year, what percentage of people did the call in and give you the last six numbers? But what percentage were reticent to do the credit cards on the web? Gosh, I can't give you statistics at this point. You know, I think at first it was a lot of people that would just give us, you know, what we call the tail of the credit card number online, and then they would call us later. But it seemed like that phase kind of, people got over that phase relatively soon.

I mean, there were still people, I guess there probably still are people that don't want to give their credit cards online. But, you know, probably for some good reasons. But, you know, can't really give you dates or percentages, sorry. But right, right. I just wondered if it was a majority early on. But that makes sense that it would go down over time. Let's see.

A couple other detail questions. So at this point, there aren't cookies and things like that. I mean, there probably are, but how are you, what are the systems that you're coming up with to track customers around the site and things like that? Well, cookies were there fairly early, but not everybody, you know, especially at the beginning in the early adopter crowd, a lot of people have that turned off because they didn't like it. So, and anyway, before cookies came into existence, there were some techniques that, and where you basically put a session ID or something like it into the URLs that you were.

sending out as part of the page. The page is dynamically created and that you put out on the page are also dynamically created. And as part of that, you can have some information that is just relevant to the one customer. And that is, in fact,

you know, the way that even if you have multiple servers and so forth, since that information is coming in and the next request from that customer as part of the URL to say click on, that is enough to kind of create some continuity in the customer's experience and what they're doing to whatever is in the server-side databases.

So it was mostly just session IDs like that. Yeah, and Paul had brought the techniques for managing that from UW where he had learned them or had a part in picking them up. I'm not even really sure. but that type of technique was somewhat known when we got started. So I kind of just adopted that and used basically just a key that was in the URLs that would connect requests from people's web browsers, tune information, shopping basket, all that kind of stuff existed on our servers.

What about the shopping card or shopping basket? metaphor. Was that something that you guys had seen? Was that already kicking around other places before? Or was that something that you, how did you guys arrive on that metaphor? Well, I think it was just, I think it was just, I don't remember having seen it, but it was sort of like the only alternative, unless you were asking people to buy things one at the time. You had to have a way of kind of identifying something, putting it somewhere. And then when you're ready to check out, check out with everything you want to buy at once. because, you know, even from the beginning, we had a shipping rate structure that kind of

encouraged people to have multiple items in their orders. You know, it made sense. And finally, the reviews featured, does that get implemented even during the beta, or was that something that came later? It was pretty early on. That between myself and a friend of mine, Jed Harris, who I've done a lot of, I'm thinking with over the years we've never worked at the same place, but we had a lot of ideas about crowdsourcing even pretty early on

and had come up with the idea of just allowing customers to write reviews and seemed okay with Jeff, so I just did it over a weekend one time and seemed to be popular, so we kept it. And these were reviews that from day one, any customer or any person could contribute to, or was it originally designed for sort of like Amazon editorial? No, the customer reviews were designed for outside customers. Internal editorial stuff was all handled. But it's still, for all of this, getting it up and doing the beta test,

it's basically just you and Paul, is that right? And Paul. Yeah. And what was Paul like as a partner to work with on this stuff? Well, he's a delightful person. We're still friends. He's an extremely fast programmer, and so he was very productive in the year,

and I think he only stayed there for like a year and a quarter. But he basically architected the whole back end, all the warehouse stuff, and helped me with the database side of things. And, you know, we were both kind of dipping into each other's code as necessary, you know, it was a pretty open-ended collaboration. And we weren't, either of us were too proprietary about our code with respect to each other working on it.

So we both worked on everything. I mostly worked on the front end stuff. He mostly worked on the back-end stuff. And, you know, he did a lot in the time he was there. So how many months is the sort of public-slash-private beta that you guys were before you launch officially? I think it was two, three months, something like that. I got there in late October of 94.

The public opening was in July of 95, and I think it was maybe three months before that that we started letting friends and family poke at things. And it was pretty raw, but they helped us find some silly bugs and iron out a few things. The internet's tell me that John Wainwright placed the first order. Was he a friend of yours?

Yeah, he was the senior architect at Collida Labs. So he and some other later people at Amazon all used to work with him and for him back then. And he was in the beta. So, yeah, apparently he was the first person to actually pull the trigger and buy something. So once the launch happens, I know that even early on, you know, you're not getting a ton of volume early on, but how stable and how successful do you think the site was around the launch period? Well, it required, you know, constant monitoring, let's put it that way. You know, we didn't have too many outright outages, but we were still learning how to do this.

And, you know, the scale kept multiplying. So, you know, at the very beginning, we had our database and the web server all running on one machine, and one machine was a tiny little tonne workstation. So it didn't take long before we were sort of hitting the wall scaling-wise and trying to come, trying to figure out ways of redoing things to be less monolithic and be able to run partly on different machines and that sort of thing. You know, we got it, we went from nothing to a public opening in a pretty short amount of time.

So, you know, in my earlier life, everything that we did up until then would have probably been scrapped and redone several times before it ever became. became a product. But really the very first thing that held together with, you know, paper clips and chewing gum, and so we had to not only keep adding things to it and cleaning it up and re-architecting and scaling and dividing it up to run on different machines, but we had to keep adding features. So, you know, we were kind of under the gun from the very beginning once we opened to the public. Well, actually, that makes me think, too. How much are you thinking of future scaling when you're working? I mean, I don't imagine that you guys, I'm sure you're hoping that there's a day coming where you'll have millions of customers, but are you

thinking of that day when you're architecting this at the beginning? Well, I can only speak for myself on that one. I was not thinking in the millions of anything. In trading, I'm just making something work reliably. And I hadn't really had a lot, a huge amount of events previously. I did operate a bullet and board for a while, but that was pretty small scale. So most of the problems having to do with that level of success were all things that I had to learn on the fly. So with the initial launch, the actual process is fairly simple. A, a customer searches for a book, your system tells them if it's available, and if it's available, your system orders it and delivers it to you guys,

and then you guys, literally you guys, turn around and ship that to the customer, right? Sometimes there were some more manual steps involved. Like, for example, anything that was available through a distributor, reasonably well automated from the beginning, but anything that had to be directly ordered from a publisher, that was people, do we not. Publishers, especially in those days, were a little behind the curve, technologically speaking, and so you'd typically have to order things from them over the phone, and it might be weeks before they get it to you.

So in the first weeks and months, you're talking about getting on the phone and placing 10 or 20, 30 orders at a time and then having them shipped in. Yeah. Doing that side of things. But, yeah, so we had some of the early hires were, you know, they needed to work with publishers to get all that inventory. And, yeah, in the beginning of the warehouse, you know, we didn't stock anything in inventory. We would just order things as customers ordered from us. And even the way the first warehouse was organized, you know, things would go into bags on the shelves according to what customer had ordered them.

So, you know, we'd wait until all those components of the customer's order got arrived there and then it was ready to be packed and shipped. And this is a far cry from two-day delivery, right? Like the average order might take a week, two weeks to get to the customer. Yeah, it would depend if on, you know, it was all available from distributors. I mean, we could get things from distributors overnight. One of the reasons for locating in Seattle was that it was within a days drive from the West Coast book distributors. And on the other hand, if there were books from publishers, you know, we could split the order into multiple shipments. that all the pieces of it weren't going to get there until we got it, which might be

three, four, five weeks later. So how much is the front-page article in the Wall Street Journal sort of the big bang for you guys, the thing that all of a sudden really moves the needle and volume takes off? Well, that was a big challenge. I mean, the traffic to the website and the number of orders definitely shot up, starting then and it was, you know, hang on by your fingernails time. Even more, we already were sort of were in that mode, but it definitely was shifting into a higher gear. And the first year or so, the first two years, as they're, as you guys are starting to bring on new hires

and you're starting to grow this into an operation that can scale with this, with this, you know, new volume of customers, what's the, what's the type of people like that they're bringing in? Because I spoke with Jane Slade, and she has that famous Send Us Your Freaks quote. But I also get the impression that these were really smart, hardworking people that ended up being successful at Amazon. Yeah, you know, at the beginning, it was nobody really knew us from Adam. So we didn't really exactly have, you know, the first pick of people to hire. So, you know, we, I think it was people that were, you know, we weren't going to hire anybody that wasn't pretty smart, but also people that were attracted to, you know, kind of what was at that point, a pretty funky, an upbeat little startup that nobody was sure if it was going to succeed or not. That kind of is a self-selecting kind of a thing.

we ended up with, you know, I would say at that point in time, it was a bit on the bohemian side. You know, there were a lot of just kind of pretty intellectual, pretty quirky, interesting people that worked there. And that made it a lot of fun in those days. You'd already said that, at least initially, you weren't thinking in terms of millions of customers and things like that. But do you remember, was there a particular moment when you thought to yourself, oh my, this is going to be something really bigger than I had anticipated. If I'll be able to remember all of such moments, there were lots of them. You already mentioned one of the main ones, which was the Wall Street Journal, front page article.

And then, you know, the IPO was another time that was, you know, in the spring of 97, not that much later. when we got some time between those two events, when we got venture capital financing and, you know, it was pretty much our pick of who we wanted to work with at that point. All of these things, you know, tended to make it clear that it was going to be something bigger than just a startup that, you know, sort of succeeded

and provided a few people. all that I had in mind at the beginning. You know, I didn't have a lot of very grand ideas of how big of a business it might end up being. I was just really interested in having something to do with creating a business that was successful on any terms. And so many of the companies and startups I'd worked with, I was kind of thirsting for something that was going to exhibit basic signs of success. So all of the scaling and everything, in a lot of ways it caught me by surprise, but it was extremely stressful. There was also a really interesting challenge from an engineering perspective,

with kind of work that I had to do. So I was enjoying it. Well, there might be several moments of this, but I was just going to say, can you think of several moments when you thought, oh, boy, the site's going to crash, we're hanging on by the skin of our teeth here, Well, there was a number of times when it actually did crash and we were hanging on by the skin of our teeth. There was one time early on, before we had any bona fide oracle experts on board, that we had a database crash. And a couple of us that barely knew how to work the thing, had to figure out how to put it all back together.

We managed to. And there were other times when we had hardware failures and had to involve people. This was after we were a little bit bigger and a more important customer of Oracle's, but we had to involve them in helping put our database together after a hardware crash, during which time we found out that our backups hadn't been working properly. So there was a lot, a number of scary, you know, skin of the, you know, these moments there. The recommendation

engine recommendation system, when did that come in to play? Well, the first thing that happened there, I'm probably not going to be able to give you

calendar date for it. The first thing that happened was Jeff had become enamored of this collaborative filtering outfit and had licensed their software. And that

We installed that and it did not produce very good results. So some programmers that were there at the time, probably this would be sometime late 96, early 97, I'm guessing. Not really sure. They started working on what was called similarities where it would, you know, they would analyze who had bought what books and what other books. and would distill that down into some useful links. It wasn't personalized, but it did kind of show people that were interested in X were also interested in Y,

and that was a good guide for people. To the best of your knowledge, was the idea with Jeff always to move beyond books? Like, books was just sort of an initial test case? Well, to the best of my knowledge at the time, that was not the case. But I think that was just because he was being more close-mouthed about it than I thought. And he did always have the sort of odd turn of phrase that books were the first best product to sell on the Internet. and sort of implied that he was already thinking about other products that could follow. But it wasn't too long before we started diversifying into music and videos.

A while after that until we got into other types of products. Was there any sense at some point that this fun, selling books on the internet project that you had signed on for maybe a little hesitantly, had suddenly gotten to become some monster that maybe wasn't what you signed up for? Well, that's a little bit of a frog in slowly boiling water story. You know, it definitely happened, but it would be hard-pressed to name a specific date when that happened. think probably around the IPO is when things kind of, as happens in a lot of companies that go public, things, you know, kind of take a turn for the more serious and you start hiring boatloads of MBAs

and all the quirky bohemian people kind of start to get marginalized a little bit. You know, so it happened that exactly when, was there one moment? when I suddenly realized it was happening. Not really. So when you eventually do leave the company in 99, you're sort of ready to go at that point? Well, I had been ready to go for a couple of years by that point. In 97, my original role was divided up between a couple of new buyers

and I made GTO, but I didn't have access to any real. resources after that to do any new projects. So, ostensibly I was still in charge of architecture, but the demands on the development groups were such that that was kind of an afterthought for everybody. So it just became too frustrating. And I knew pretty early on after those changes happened that, I'm going to stay forever. When you do leave, is there an initial sense of exhaustion like this is a wild ride that I've just gotten off and thank God I've survived.

Yeah, it was sort of a years-long process of unwinding after that. Because, you know, that was, Amazon was not my first or only startup. It probably wasn't even, even though it was kind of a relentless march. The work itself was actually not as hard as work that I'd done in many other places, not as technical or, you know, kind of deeply complicated. So I was ready for a nice long break after that. Is there anything that you can point to, as you said, you were the veteran when you joined Amazon of several different startups,

but not a lot that, you know, were probably none that were nearly as successful as Amazon. Can you think of like one or two things that, in retrospect, made Amazon succeed where other teams you were on didn't. It was that it had a clearly defined product and a clear community of potential customer. Many of the things that I've worked on, you mentioned the FROX product that I worked on. Well, that thing was a really interesting technical problem. And there's some really smart people worked on it. And that it was not really in tune or in time with the market that existed at that time.

So, you know, it just never went anywhere. Whereas with Amazon, it was, you know, a pretty pedestrian product. I mean, you know, selling people books. Okay, that's not that exotic. You know, the way we were doing it was something a little bit different. But it was, you know, I think the thing that, you know, At least compared to my personal history, it was mainly that. It's just that there was like an obvious product and obvious people to sell it to.

I know that you left the company in 99, so I don't know what sort of access you would have. But I want to ask a lot of Amazon people that I'll be talking to, how much of a fear was there after the dot-com bust? How much did you feel maybe even personally that maybe Amazon, might not make it. It happened shortly after I left. So I don't think it was because I left. Right.

I did, you know, I did wonder what was going to happen with it. I thought, you know, I mean, every, you know, all these companies were going out of business and it wasn't, you know, I mean, I had a certain amount of faith that they, you know, were smart people that were serious about staying in business and succeeding, but I didn't fact that they were going to be able to. What do you think of Amazon when you think of the company today in terms of what it's doing and where it's at? A number of different directions.

I mean, obviously it's a very diverse, very dominant, extremely competitive, ruthless, huge company now. It's hard for me to even, sometimes I even just find it difficult to imagine that I actually had something to do with getting that thing started, anything that I ever had in mind when we started to think. Are you an Amazon customer yourself? Sometimes I'm in a bad mood about them, so I don't buy things for a while, and sometimes I'm not, so I do.

So let's end sort of with, you know, now looking back, it's almost exactly 20 years. that whole ride got started for you and for everybody with Amazon. Looking back, what's your favorite memory from, or the thing that you're most proud about the Amazon project that you worked on? Well, for me, I think it's still just making a much wider variety of books available to people in much more remote. and locations could do before. That was kind of what I wanted to do, at least the first step of what I wanted to do there,

and I think we did that pretty successfully. And personally, I'm proud of what I mentioned earlier on, that sort of building a hypertext web version of the card catalog and books in print. I think that was at a time when we had almost no content, that gave some people something to play around with and find books that were related to other books. And it was very sketchy, but it was something to do

that made it fun to use our site. And I'm also proud of the fact that we were, essentially on a very low budget and very small staff were able to build something that was pretty reliable and scaled up. You know, maybe it was messy sometimes, but we managed to keep things running and didn't have any, you know, fatal outages

and boarded the business as it needed to grow. Well, you know, also, I mean, everyone tends to give a lot of credit to Google for this world that we live in where, you know, information is always available at your fingertips and things like that. But, you know, I'm old enough to remember a day when there were certain books that you couldn't get,

certain CDs that you couldn't get, an obscure band that you'd always wanted to hear, but you would never be able to find. I think Amazon deserves a lot more credit for bringing this information to everyone's, making it accessible to everybody. Yeah, I think that that is true. Although, technically speaking, we weren't able to do anything anymore

than a bookstore or a music store could in terms of special ordering from publishers. but, you know, we did it kind of on a day-to-day basis, and it was sort of central to our model as opposed to something that just happened, you know, if you could get some time from the clerk at the store to help you out. Right, but as a 16-year-old who I checked it out, my first Amazon order was a Humphrey Bogart biography that I couldn't find even at Barnes & Noble, just to be able to do that myself as opposed to having to go to the clerk.

I mean, that was a big deal. Yeah, yeah. Well, I agree. I think that was qualitative change, and I am proud of having had some part in making that happen. Well, Shal Kaffin, thanks so much for taking the time to remember all that for us. You're welcome. Congratulations on 20 years this year. Oh, thank you.

If this is the first time you're listening to this podcast, please subscribe to us on your podcast app of choice. There's plenty more great internet history where that came from. And if you're a long-time listener, then you know what to do to help us out. Rate and review us on iTunes. Because iTunes gives credit to reviews and ratings, and the more great reviews we get, the more people will discover us. As always, there's more info on our website, www.com. The show's Twitter handle is atnethistorypod.

And my personal Twitter is at Brian MCC. Thanks for listening.