Brewster Kahle started WAIS (Wide Area Information
Servers) in the late '80s while an employee of Thinking
Machines. He left in 1993 to found WAIS, Inc. WAIS
was one of the earliest forms of Internet search software.
Developed before the Web, it was in some
ways a predecessor to web search engines. Kahle sold
WAIS to AOL in 1995.
The next year, Kahle founded Alexa Internet with
Bruce Gilliat. The Alexa toolbar tracked user browsing
behavior and suggested related links using collaborative
filtering. Once captured, pages visited by
users would then be "donated" to the related nonprofit Internet Archive, to help
build a history of the Web.
Alexa was acquired by Amazon in 1999. Kahle continues to run the Internet
Archive.
Livingston: You were one of the first members of the Thinking Machines team.
What number employee were you?
Kahle: I was not one of the two founders--they were Danny Hillis and Sheryl
Handler. I was on the project team at MIT, so when we started the company,
anybody from that team that wanted to come came. There were three or four of
us. We had been working on it for a couple years before there was a company.
Livingston: Tell me about some big things back in the Thinking Machines days
that helped pave the way for WAIS.
Kahle: Thinking Machines was not my doing, but I was on the project team
beforehand and then helped start the company. What I learned out of that was:
do your homework before you are spending your own money. We did a full
couple rounds of the Connection Machine at MIT before we started a company.
It was very helpful to get your lessons learned basically on somebody
else's nickel, in a research phase.
265
Brewster Kahle
Founder,WAIS, Internet Archive,
Alexa Internet
20
CHAPTER
Another lesson that I learned out of Thinking Machines was, if you're trying
to get your company to think differently--to do something interesting--pick
your setting carefully. Thinking Machines was set in an 1800s Victorian mansion
on 100 acres of forest just outside of Boston. It was a park, basically. Working in
an environment where, if you got stuck, you'd go for a long walk is very different
than trying to do a startup and think differently if you're in Suite 201 in some
major office complex. That was a lesson that I've used every startup since.
Thinking Machines had the great fortune of starting with $8 million in the
bank, because some very rich individuals really believed in it. It was not
venture-funded, and it was founded with the idea that it was going to take years
and years and years to actually get something real done. That allowed Thinking
Machines to attract a very interesting set of people. Richard Feynman worked
there, Stephen Wolfram, Marvin Minsky. I found that I had better access to
professors in a company than I did when I was in school. So that was an interesting
way of trying to figure out, "What should the company do?"
They took a good summer to try to figure that out and, actually, through a
bunch of the first year, which is a luxury that most startups don't have. Usually
you have to work really hard to try to get your first release built. So being in an
interesting setting with brilliant people coming by, it was quite a unique
startup--very unlike the West Coast startups that I've seen.
Livingston: What were some of the big turning points early on?
Kahle: We hired a fellow from Digital Equipment Corporation--his title was
VP of Reality. The idea was to try to help a bunch of folks that had great ideas
out of MIT, but had never really produced a supercomputer before, figure out
"How do we actually do that?" I remember being invited to give a design review
of the core central processing unit of this new computer, and I really didn't
know what a design review was. It was quite embarrassing. But it was very helpful
to inject a VP of Reality. It stirred up the culture to try to get it so that we
could actually produce working machines.
There was a lot of trust in very young people in that company. People were
in their early 20s. So the basic design and building of the machines--even
though we were completely underqualified, looking back on it--was entrusted
to a very young set. But it made it fun. We were absolutely glued to the project.
We didn't really have much of a rest of a life.
Livingston: Were there any experiences where you thought, "If I start a
startup, I'm not doing this"?
Kahle: The blessing of Thinking Machines and the curse of Thinking Machines
was that it had a lot of money. If you have a lot of money, then you can be
detached from people that are going to pay you in the future.
My first startup upon leaving Thinking Machines was a bootstrap. We had
no investment at all, and I had no savings, so it was self-funded from the beginning.
That was a night and day cold bath. It was sort of like going from the
Roaring 20s, when champagne is coming from everywhere, into the depression,
where you are washing your baggies and reusing twist ties.
266 Founders at Work
Actually I really liked the discipline that came from a bootstrapped startup.
I think that everybody that goes and does a startup--even if they don't do a
major startup that way--should start a business that is having to make people
happy with them day one, through contracts, through small scale sales, whatever
it is. How low can you go? How can you build something really inexpensively?
How can you not spend money on furniture and matching carpet and
those sorts of things? The biggest thing that I probably didn't do the second
time around was have any money.
Livingston: Tell me about how you got the idea for WAIS.
Kahle: The idea of WAIS was to make network services--stuff that you take
completely for granted now--but the idea was that you could use remote
machines to answer questions. The ARPANET was just really becoming in use
by universities in the 1980s, and so at Thinking Machines we were trying to
figure out, "How would you make use of a machine that had 15 gigabytes of
disk space and would have processors that you could run at, say, a gigahertz?"
This was completely amazing. How would you possibly use that much computing
power? We said, "Well, you'd put them on the Net, and they'd be smart
machines that would answer your questions for you." So this was the idea. We'd
prototype it around Thinking Machines.
I prototyped it in my spare time. I guess there was very much an ethos that
hacking was encouraged--playing around and doing fun things, spending time
doing something that wasn't your exact job responsibility, but doing it at work,
and getting support from work. So I tried out the idea of remote publishing--of
remotely asking machines questions. It was the first Internet publishing system.
It came before Gopher and before the Web, but it was the first system that was
trying to answer questions over the Net. Yes, search was a big part of it, but you
could also click around and it had a URL system and it had a search system for
finding servers. It had all the different pieces. It was built on an open protocol.
We did it as a project of Thinking Machines, working with Apple Computer,
KPMG Peat Marwick, and Dow Jones.
So my first experience with trying to start something at that scale was actually
done within Thinking Machines--again, following that original lesson of
"learn your lessons without spending your own money." So we tried basically
building this system up within Thinking Machines. Thinking Machines wanted
to make money by selling servers, so we needed to get the rest of the system
going. We had Apple Computer to do the front end, the client piece; Dow
Jones to do the information sources; and KPMG Peat Marwick for their corporate
information and as a user base. So it was a test project. It ran about a year
and a half and was successful. Everybody loved it. Each one of the organizations
went forward to figure out how to make this all go. This was in 1989/90. So
we were all looking into the future.
Livingston: WAIS seems to have ideas that anticipated the Web.
Kahle: All these ideas were in the air. The Web came a bit later, but, as I understand,
Tim Berners-Lee was working on some of the same things, but doing
Brewster Kahle 267
them locally within CERN in Switzerland. We were doing them within a corporate
environment using supercomputers and the Internet as well.
Livingston: So Dow Jones, KPMG Peat Marwick, and Apple were all involved?
Kahle: Yes, everybody was working together. It was a project that had a project
team in each the companies, and I ran it. I moved out to the West Coast to try
and run it, because I knew I could run Thinking Machines remotely, but I
didn't think I could run Apple remotely. So I moved out to the West Coast and
had a cubicle and a project team that I worked with at Apple.
Livingston: Tell me about some of the hard technical problems that WAIS
addressed.
Kahle: One of the difficult things was just using the computer networks at the
time. This is 1989, and the Internet hadn't quite become easy in any real sense.
Trying to hook up to Dow Jones through X.25 networks and ISDN was all quite
challenging. KPMG Peat Marwick started to have an intranet then that we
could use, and fortunately they used Macintoshes. The Macintoshes were helpful
because they had TCP/IP for them, where Windows didn't. It wasn't until
Windows 95 came out 6 years later that Microsoft caught up.
What I loved about it was I got to work with four companies that were managed
very differently from each other. This was my bath of "How do companies
work?" Thinking Machines was a bottom-up company. In many ways, the ideas
and the people with power were the young engineers. They could see where
things were going better than the top level management, because everything
was so new.
Dow Jones was a top-down company. If you sold the top guy, he'd say, "OK,
we're going to do this." Then he'd issue a command, and the next-level guy
would say, "OK, sir, I'm going to do this."
Apple Computer was explained to me to be a "beanbag chair." You had to
push not only on the top, but on the bottom and the middle all at the same time
to try to get it to move. This was a time when John Sculley was running Apple,
and I don't know if it's any different now, but at that time you had to actually
keep pushing all up and down the whole chain or it just wouldn't move. You'd
push, and you'd think you were making headway, but the beanbag chair didn't
move.
KPMG Peat Marwick was a democracy. It's a partnership. Each partner
thought of themselves as in control of their piece of turf. And they were very
much so. They controlled their own revenue; they spent it; it was a democracy.
In fact, every so often they would get together and elect their upper-layer management.
If they didn't like the upper-layer management--which was just
another consulting office, like any other--they'd vote them out. And they did.
The folks that were really supporting the WAIS project at KPMG Peat Marwick
after a couple years got more or less voted out. Not because of WAIS, but the
partners said, "We want a different type of management."
268 Founders at Work
So this was still being done within Thinking Machines, but I found that
Thinking Machines wasn't going to go and build some of the pieces needed to
make this Internet publishing world work. We coined the term "Internet publishing"
and tried to say, "OK, this should happen." Apple would do their piece
and Peat Marwick would do their piece, but nobody would go and do the central
set of tools, the software needed. So I said, "OK, well, I'll do that. And start
a company to do that."
There was a decision to try and figure out where it should be. Should it be
in Boston? Should it be in Silicon Valley? Should it be someplace completely
else? So I went around to the smart people I knew and said, "Where should we
put this company? What I'm really trying to do is build an industry." Not build
a company, build an industry, so there would be all of these pieces that would
make network publishing come about. Some people thought it was a little crazy
to think about starting an industry, but it seemed like it made sense to me.
The best piece of advice that I got was from Bill Dunn, one of my mentors.
He said, "Go someplace where people don't think you're crazy." Which sounds
like a pretty simple thing to say, but it actually turned out to be a very wise piece
of advice. Boston, especially back in 1990/91, was in recession and having
trouble. California was also in recession, but in California there were dreamers.
There were people who wanted to think about new and different things and
wouldn't think we were crazy to try to build this thing.
So I decided to start the company out here and start with a contract--it was
a bootstrap--doing the information system for the Perot campaign for the presidential
campaign of 1992. They could really use an information system that
leveraged some network. Of course, they didn't know what network, but we
did. So we could build this Internet using modems and leased lines and all sorts
of things to be able to build this up. The Perot campaign collapsed, but we still
had enough money to make it through most of the first year to go and get our
products built.
One of the interesting ideas out of WAIS was its use of freeware and shareware.
This was a new idea, more or less. There had been some examples of
freeware like GNU, but there were also mixed models: Kermit, where people
basically would make source code available on the Net and sell something
related. We gave away the client version, the equivalent of a Web browser.
It was a WAIS browser and a server, so people could go and build their own
system.
During that free period, hundreds and hundreds and hundreds of servers
were set up. We got up to about 10,000 servers--all this is before Gopher and
the Web came out--based on free software. Once people got really hooked on
the free software, they wanted upgrades, or they wanted services. So we were
there as a company to sell it to them. We made the free version, and there was
a for-pay version. It was the same idea that we've seen now with Netscape.
There's a whole set of companies that also tried to give something away and sell
something else.
Brewster Kahle 269
Livingston: You were the first company to do that?
Kahle: Well, I think we may have been the first to think of the Internet as a distribution
system of software: to give something away and to sell it. I don't know
of any examples before.
Livingston: That is so many companies' business model these days; it's interesting
to think it was the first.
Kahle: I don't know if it was the first, but it was certainly early. We also found
that people--even if we sold them the software--often didn't know what to do
with it. They wanted consulting services. We started what I think became the
first web studio, or web services business. We worked with big players, whether
they were newspapers or magazines, that wanted to publish on the Net. This
allowed us to work with the big boys very early on.
We tried very hard to work with the best of class. They were always more
difficult to deal with, but they were great to work with once you got working
with them: the Wall Street Journal, Encyclopedia Britannica, Government
Printing Office. We worked with both the House of Representatives and the
Senate. So we were working with people who had insights into what it is they
really wanted. It's harder to get those customers at first, but they were terrific
because they weren't just trying to catch up. They weren't trying to be number 2.
They were number 1 in their fields, and we could learn from them.
As the Web came along and was a better underlying system, we became a
web services company, basically. We set up, I believe, the first publisher on the
Net, which was Scholastic. That was done during the Gopher era. And we put
the first advertising-based service up, which was for CMP. We put the first
subscription-based service up, which was the Wall Street Journal. So we were
trying to get publishers online, and that was what the WAIS system was.
I would stand up in conferences in 1990/91 and say, "I'm the token dot-com
in the room. I'm here to help people make money by publishing on the Net."
The idea was to try to get the Net to go commercial enough to support publishing.
Livingston: Did these big publishers at first think this was crazy?
Kahle: Usually it would take 9 to 12 months with lots of meetings before anything
would turn concrete. Eventually they wanted to do some sort of project,
and they'd throw $100,000 into a project. Since we were so inexpensive--we
were living based on furniture that didn't match; we had learned our lessons of
how to live very inexpensively--we could do things as production that they
would normally pay an Ernst & Young just to do a study.
We could deliver something that they could really learn from, and we'd
learn from as well. That kind of partnership worked well for us. But, boy, was it
tough running with so little money.
Livingston: When you broke off from Thinking Machines, were they OK with
this?
Kahle: There was a lot of consternation. There were a lot of questions of
whether Thinking Machines owned the idea or whatever. I was very careful to
270 Founders at Work
make sure that all the things that the (eventually) WAIS company needed were
actually based on the public domain of software that had been produced. It was
based on the open source software. So there weren't actually any patents or
copyrights, but there was still a bond. And there was this question of, "Should
Thinking Machines own it? How should it work?" But there was so little to own.
It wasn't like it was a VC-started company where you could value it. It was just
basically myself and one other, Harry Morris, that left to found the company.
I think there was some hand-wringing, but we couldn't figure out any way to
really do a deal. The problems that slow things down the most are these things
around intellectual property--especially when there's no money yet--and
these are the times that you could end up talking endlessly and you can't figure
it out. It's a lot easier if there's money. Then you have a mechanism of knowing
how share it. But before it, it's really, really tough.
I found it was very helpful to start a company out on the West Coast. Even
though I had moved out here, I made a partnership with a fellow named John
During--key for me, basically the cofounder of WAIS. He was a key player
because he'd been around for a long time, so he knew how to do all of this stuff.
How do you get an accounting company, the law firms, what do you spend
money on, how do you negotiate a lease--all the things that I had never learned
by being an engineer inside another company.
Livingston: How did you meet him?
Kahle: He was a consultant for Dow Jones. Also I found that there are people
that specialize in different parts of businesses. Some people just do startups
over and over again. Or they are actually in the idea stage. So a lot of people
that I saw in the late '80s and early '90s in this Internet world were the people
that had been involved in the PC revolution 10 years before. They had seen all
of that go through, and they were looking around for "what's the next thing?"
and this Internet thing started to smell kind of like it. So John During had a lot
of experience in that.
Livingston: So you said, "Let's do it"?
Kahle: Yes, and moved out to San Francisco, started the company in a Menlo
Park mansion, sort of on the Thinking Machines model. That was as far north as
I thought I could put the company and still be connected in with the Apples
and the Suns and the other technology companies.
In 1992, San Francisco wasn't the place for companies. That happened in
the mid '90s, with the whole South of Market rebuilding. That was another sort
of "learn the lesson of going someplace where people don't call you crazy." I
really needed the help of those that were in Silicon Valley, though I knew that
as this industry built up, it was going to work more with the creative people. So
it was going to transition more and more to San Francisco. When we moved
offices in 1994, we moved it into the city, so that we could work with the publishers--basically,
the people that were going to be out there on the Net, not
just building the technology, but using it for something.
Brewster Kahle 271
Livingston: So you were getting a little bit of money from clients. Did you hire
anyone?
Kahle: Yes, there was Harry Morris, the key engineer that built much of the
technology. It grew into a company that was about 30 people, 35 people by
the time it was bought by AOL.
Livingston: Tell me about some of the most hair-raising moments.
Kahle: Going broke. When you just don't have enough money to pay the bills.
We would usually live with the amount of money in the bank that would be
about 2 to 4 weeks worth of the bills. We'd never have more than that. If we got
up to 2 or 3 months, we would think of ourselves as being flush. Living a bootstrap
in a non-market--we're trying to make a market go based on making
servers in a client/server publishing environment based on the Internet, that
people didn't understand. So it was very difficult, but it was a good disciplining
time.
Times that were really interesting? I think working with really great customers.
It sort of sounds like a cliche, but to go and learn from Encyclopedia
Britannica, the New York Times, the Wall Street Journal, and try to learn how
do they view their businesses and their lives was the most fun out of the whole
experience.
Livingston: What did you learn from your customers that surprised you?
Kahle: I love working with businesspeople because they are really straightforward.
Now we have lots of lawsuits around copyrights in the music industry,
blah, blah, blah. But if you work with the actual businesspeople--not the
lawyers that work for the businesspeople, but the businesspeople--they are
very straightforward. They just want to make money. And they just want to
make more money than they are making today. They understand there are
going to be transitions and technology changes. So if you lay out a path of how
they can make more money, potentially, at the end of this than before, then they
are on board.
We got the newspapers on the Internet during that time. You take it for
granted, but all the newspapers are pretty much online now. They control their
own distribution. They have their own websites. It doesn't all funnel in through
an iTunes. The music guys, I'm not sure why they did this, but they sold their
souls. Somebody else controls not only the distribution of their product, but
they control the pricing. What do you have if somebody else controls the distribution
and the pricing of your product?
So the newspaper guys and publishers were great because we'd say, "You
want to control the distribution of your work." And they'd nod their heads, and
we'd say, "Well, there are some alternatives out there." In the early '90s there
was AOL, there was Lexus Nexus from the '80s, where they would lose the control
of the distribution of their work. We'd say, "Do you want that?" They'd say,
"No, we don't want that. We want to control the distribution of our work." So
we said, "Swing with us for a little while, while we build this Internet. Let's
build this Internet together based on open systems." So these business guys
were, in fact, wanting an open system.
272 Founders at Work
We wanted to build this up before the monopolists got to town and said,
"Oh, you don't want an open system. You want a closed system that belongs to
us, and we'll do it really well." So we worked very hard to get it to go in the early
'90s, to get an open system anchored. And it worked. By the time AOL
announced in '94 they were going to support basically the Internet protocols
and when Microsoft said in August of 1995 that they were going to support the
World Wide Web, that meant that we won. We had gotten publishing on the
Net, and at that point I could graduate and do the thing that I wanted to do.
Livingston: Do you remember things that your clients totally misunderstood
about what you were trying to do?
Kahle: I learned to try not to make too many leaps at once. Most people have a
very difficult time imagining something they can't see at least a demonstration
of. If you can get a demonstration--or, worst case, a video--it communicates an
idea better than hand-waving for hours. So get to a demo quickly.
This was difficult when the Internet hadn't been deployed. Often the executives
didn't have computers on their desks. They had secretaries that typed for
them. Remember, it wasn't that long ago that these things weren't everywhere.
And they weren't hooked up to anything. Maybe they had a modem, but that
was it. We had to demonstrate these things over modems. Trying to get clearance
so that they could dial out from a computer to the Internet to demonstrate
this system inside the CIA headquarters was actually a several-day process. So
it was difficult to go and explain very many jumps forward.
Whenever I went and said, "Really what I want to do--after we get this
publishing up and running, is build the library..." Because that's always what I
wanted to do. I just thought I had to build these supercomputers first, then
I had to get the publishing going, and, once we got that, then we could build a
library. So it was not until '96 that I got to the place where what we had
dreamed of in the late '70s as the goal--which was to build the great library--
could even start.
Livingston: That was always your goal?
Kahle: Yes, that was always the goal. We just had to do a couple things first. It
took a lot longer than I thought. We're now in 2006, and it's hard to believe how
pathetic things are. We don't even have books online yet. I don't know why the
world moves so slowly. Everybody says, "Oh, it's moving so fast." And it's like,
"No, I don't think so. It's been forever."
Livingston: Besides always running out of money, how did you find having your
own company different from Thinking Machines?
Kahle: Having your own company means that it's much harder to blame somebody
else. If you are working inside a big company, you can always blame management,
marketing, engineering, or something. But, when you are running it,
you can't, because it's all your responsibility. I found that to be quite cathartic.
The East Coast also has a little more of an aesthetic of complaining, and so I got
a little bit over that, I guess.
Brewster Kahle 273
Another, I thought, was expressed really well by Don Yannias of Encyclopedia
Britannica. He said, "Now that I'm running Encyclopedia Britannica, I have to
be Mr. Sunshine every day." Because people are looking to you, not just for the
ideas, but for the general attitude toward how to make the whole thing work.
Carrying a company is a lot of weight. You have to make sure that you keep on
the uptick--not just financially, but also make it so that it's a fun environment
and people want to work there.
Livingston: Did you have any competitors back in the WAIS days?
Kahle: There were other systems around, but one thing I tend to do is do something
that is far enough out there that nobody in their right mind would possibly
want to do it. In general, I usually take things from the "you gotta be crazy"
period to the "of course." And once it gets to "of course," then there will be
competitors, and I'm done. Because usually what I want to do is just get other
people to do it. The best way to do that is to show that it's possible.
So WAIS was all about trying to get other people to copy us. And they did,
and it worked great. And they did better at it, and flourished. Better web studios
than we were, server manufacturers and the people that made web
servers--they did much better than we did. But the idea on WAIS was to try to
guide the building of it, because WAIS wasn't the goal. Building that company
wasn't the goal. I wanted to get it so that publishing would happen on the Net,
so then I could go and actually do something.
Livingston: Publishing was happening, you sold WAIS to AOL, then what?
Kahle: Then I tried to work within AOL, and that was very difficult. For an
entrepreneur, acquisitions are very difficult to manage. That's a warning. I've
been through two acquisitions. One was WAIS; that was bought by AOL. The
next round I built two organizations at the same time. One was called Alexa
Internet (short for the Library of Alexandria), and the other was the Internet
Archive, to archive everything that was in the library. Alexa was a for-profit, and
the Internet Archive was nonprofit. I didn't make enough money to go and make
a nonprofit and fund it myself, and I didn't know how to ask for money in a nonprofit,
but I knew how to build products.
Alexa Internet was a navigation system for the Internet. Bruce Gilliat and I
started it out here in San Francisco, in a house in the middle of a park--in the
Presidio. We're in a 1500-acre park in the middle of San Francisco. We're
the second lease-holder here.
Livingston: You started both companies simultaneously? Did you have different
people running each one?
Kahle: Everybody worked at Alexa. The idea was that everything that Alexa
ever collected would be donated to the Internet Archive. Over the long term,
companies come and go. They usually don't last that long. But the great thing
that was going on with the Internet wasn't the technology. That gets replaced.
It's the information, and it's all the people. So we started collecting the World
Wide Web and making services in a commercial company, but donating all of
the materials collected to a nonprofit that was designed to last the ages. It was
274 Founders at Work
very specifically designed to think through what happens after the commercial
company is gone.
Livingston: When you first started with Alexa, did you get funding?
Kahle: I funded the first part of it with Bill Dunn. And I cofounded it with a
business-oriented fellow, Bruce Gilliat, because I'm more on the visionary side.
Building in a businessperson has been a good idea. Finding a good partner is
extremely difficult. It's as difficult as finding somebody that you want to get
married to and you'll stay married to forever. A business partner is very difficult,
and if you can find a good business partner, stick with that person.
Livingston: What makes a good business partner?
Kahle: Compatibility. Mutual respect through hard times. Maybe it's clear lines
of differentiation for who does what. But finding a good business partner is a
fantastically valuable thing to do. So the second startup, Alexa, I started with
a partner as a full cofounder and that worked out really well.
Livingston: Did you get funding?
Kahle: We got $1 million to get the first round going, and then we started talking
to venture capitalists. This is 1996; some of the companies started going
public, so there's some money around. But again, everything that we were
talking about, we couldn't communicate it in a way that made sense to them. So
we got private investment by a single individual. That was very helpful. We
grew that company to around 45 or 50 people and then sold it to Amazon.com.
Livingston: The toolbar was a brand new idea. How did you entice users to
download it?
Kahle: The idea of Alexa was to help guide you around the Net. We thought
that search engines were going to give up steam. We just didn't think that they
were going to be able to scale. I was wrong, just wrong. But the thing that we
wanted to do was help people navigate around the Net. We wanted to catalog
the Web: make it so that you knew where you were and where you might want
to go next. The concept of the company was to show you related links to every
page that you were on.
So if you're on a web page and you are looking at some car, some book, or a
website about some new computer, then you'd be able to see, "Oh, if you're on
this page, you might want to go to this page, this page, and this page." It may
not be what the owner of that website wants you to see.
Livingston: Was that what became known as collaborative filtering?
Kahle: It came to be called collaborative filtering. The way that it worked was
we collected user trails of "Where did they go?" You know, the Amazon recommendations,
"people who bought this book, bought that book." This was, "people
who went to this web page, went to these web pages." And we did it years
before those other systems. It was based on some work by Carl Feynman, when
we were talking about this at Thinking Machines. We were talking about,
"Where could this whole thing go?" and he said, "Well, there might be editors,
Brewster Kahle 275
and you might be able to discover editors," based on this idea that became collaborative
filtering.
He went off to MIT, and they started a company called Firefly that was
based on his ideas. So anyway, this idea of doing a web-scale collaborative
filtering, people who liked this web page liked these web pages, was what we
did. We did the toolbar to do this and to offer information to people as they are
moving around the Net. In return, all of these people would be giving us information
as to where they went, what trails they took, which we'd then learn
from.
Livingston: What did you learn?
Kahle: We learned what was related to other things. But in the great scale, what
I loved was watching some of these users--though we didn't know who was
who and we didn't care. In fact, that was very important because it's very private
information. Something I think we're forgetting is that some of these other systems
are collecting this information, and they do care who's who. Google's toolbar,
for instance. Going and learning where people go, you learn a lot more
than you want to know. So you have to go and delete some of them; otherwise,
it gets scary.
Livingston: Alexa deleted them?
Kahle: Yes. Alexa deleted things. Others that sort of followed in those footsteps
aren't deleting things. This is going to be a big problem. At Alexa, we started
with a code of ethics in the whole approach, because we knew that we were
gathering information that people often didn't know that we were gathering, or
they weren't really conscious of it. Throwing away a lot of information is key in
such a circumstance.
Livingston: What went wrong with Alexa?
Kahle:We couldn't get the ad model to work worth a darn. Our idea was to give
very contextual ads. We knew what web page people are on, might as well
give them an ad that made sense. But we couldn't find a way of selling ads in
such a way that it worked. We had an unbelievable number of page turns. We
had many opportunities to put ads in front of people, but we couldn't turn that
into making money.
So our underlying business idea failed. We were very successful in terms of
getting lots of people to use it. People liked it. Netscape bundled it into their
browser. Once Netscape bundled it into their browsers, getting Microsoft to
bundle it into their browsers was easily done. (We found the best way to get
Microsoft to do something was to get the competitor to do something.)
So we got lots and lots of users--millions, tens of millions of users--but we
couldn't figure out how to make the business work. At that point, when some
folks from Amazon were looking around for data mining technology, they came
and said, "Should we buy you for data mining technology?" And we said, "One,
you shouldn't buy us, and the other is that we're doing something quite different
from data mining. We have this toolbar we can get out there in front of
people and things like that."
276 Founders at Work
In talks with Jeff Bezos, the founder of Amazon.com, I said, "I tried being
acquired, and it didn't work. By AOL, a great company, but my company got
dispersed, and I don't know how to run a division; I know how to run a company."
He said, "If we're going to buy you, why don't you run it as a company?
What does that mean to you?" I said, "Well, it has a board, and I meet with the
board once a month and they give general direction and I run the place." And
he said, "OK, let's do it that way."
So we got acquired, and we ran as a separate company. The company is still
running. It's about 200 yards away from the Internet Archive, which is where I
am now. I stayed for 3 years and then moved over to build the Internet
Archive--which had nobody working here--into a real organization. Because
once we had enough materials, then we could build the library. So Alexa was
about the cataloging of the library, and the Internet Archive is trying to build
the stuff.
Livingston: This was your dream?
Kahle: Yes. One thing I learned from Marvin Minsky (one of the founders of
AI) was, "Pick a big enough project, something that's really hard, something
that over the years you can work on." I've found that that has been a great guiding
piece of wisdom. If you just set out to go and make a lot of money, then the
problem is, what happens when you make a lot of money? You're out of ideas.
So the idea of going and putting everything online is something really big
and hard. How do you make a library such that everybody has access to everything?
I remember talking to Richard Feynman, and we were looking at the
Encyclopedia Britannica in one of these 1800s rooms at Thinking Machines.
We had an Encyclopedia Britannica, and it had an index that was one volume,
then a micropedia, which was about 10 volumes, and then the next level was the
macropedia, which was about 30 volumes. We just imagined: how many more
layers of this before we have everything ever published? It turned out there
were like 5 more layers, and I said, "That couldn't be that hard. How much
information is there? It's not that much." So even in the era of Thinking
Machines, we knew what it was we were trying to build. It just takes longer than
one thinks. That was 20 some odd years ago.
Livingston: Do you think it's a good idea for those who have a big dream like
that to section it off a bit? To try to create a successful startup to get the money
to give them the freedom to pursue their dream?
Kahle: Yes. I try to make sure that every year there's some accomplishment that
you can actually point at and say, "OK, this year I'm going to do this." This year
I'm working on digitizing books. Last year I was trying to get a storage computer
to work internationally, so that we'd have copies in Europe and the Arab
world, in Egypt. We made copies so that, in case we disappeared, the information
lived. Every year, try to do something that you can point at. Otherwise, a
couple years go by, and you say, "What really happened?"
Livingston: Who were your mentors?
Brewster Kahle 277
Kahle: I've had two. Most people don't have mentors. They say, "Well, I've had
influential teachers. I've learned a lot from this person." But they don't think of
it as a mentor. A mentor is a life guide, somebody that you might work with, but
somebody who is helpful toward watching bigger issues about things that guide
your life.
Danny Hillis, who was 4 years older and whom I worked for at MIT and
Thinking Machines, has been a guide and a help ever since. The other was Bill
Dunn. I found those two men, both being very kind and smart, had the ability
to know what was going to happen--even though they had way too little information.
I'd always sort of note down their wild ideas and think, "Did they come
about?" A few years later you find out they were right. Some people are just
more right than they ever deserve to be.
Livingston: You've done startups in the East and West Coast hubs. Is one place
better than the other for startups?
Kahle: Oh, I think it's much easier to do a startup on the West Coast. There are
all the facilities and services available to you. You can put together a marketing
department out of part-time people. You can hire an accountant to just do
exactly what you need. You need a lot less infrastructure that you control to do a
startup on the West Coast than on the East Coast.
If you started with $8 million, you can buy everything you need; but if you
are starting, just you, you can do a startup out of your bedroom. In fact, a lot of
people do. In fact, most bedrooms I think are startups! The idea that you can
start on a shoestring, that you can hold a meeting in a coffeehouse and that's
OK, is perfectly legitimate on the West Coast.
Livingston: Why not in Cambridge?
Kahle: Maybe you can do that now in Cambridge; maybe it's changed. But
there's a more institutional idea that you have to be more proven. San Francisco
is full of dreamers. It's the people with the new ideas. It may be bad, they may
be inappropriate, they may fail, but I love the idea that we can do something
new and different--something that hasn't been done before, something that's
going to affect a lot of people. There's an idea that you can pull something off
here. That sort of uplifting nature to San Francisco and the Bay Area in general
really lives on. This is a city of dreamers, and that's what makes it just a
wonderful place to live and to work.
Livingston: Looking back on all of your experiences, what surprised you most?
Kahle: How long things take. To start a company and to get to a point where
one has a critical mass--you have an office, you've got your CFO, you've got all
of the infrastructure to become a viable entity. I think about 20 to 40 people is
a golden size, because you're not spending all of your time doing things that
you're not that good at. There are other people in the organization with more
specialization in different areas. It takes a couple years to really get that
debugged.
278 Founders at Work
You can grow it instantly. You can hire 40 people in an afternoon, but they
won't necessarily work together well; they won't understand what's going on. It
takes a while. So 6 months, 9 months goes by often just putting together all the
pieces of infrastructure. With Alexa, it took a year to build the company to
the extent where we could do our first real product release. WAIS was the same
way. The first product release came a year after the start. It always seems like it
should be much quicker than that.
Livingston: What can big corporations do to preserve the startuppyness of the
companies they acquire?
Kahle: My first company was bought by AOL, and what AOL wanted to do was
inject the Internet into its veins. So they went around and bought a bunch of
companies. And I'd say what they'd bought the company for, if I had been more
worldly, they actually achieved. It just wasn't what I was looking for. I had built
a little company. It made something like $3 million a year, which I thought was
pretty great. I wanted it to get to $10 million or $20 million, but that was a
rounding error for AOL--it was noise. They needed us to help on the big
issues. So I worked on strategy for the company for 12 months--to get the company
going in the Internet direction. And that's really what they wanted. It just
wasn't something that I knew how to do. I really liked running something that I
knew how to run.
When I did the startup that was bought by Amazon, I said, "Leave us on our
own. We're smart and independent enough to be able to do good work that will
inject things from the side into your other organization." The thing that Jeff
Bezos did that I thought was very smart was that he ran us through his organization
and others through ours. He used us, at least for the first few years, as a
think tank in some sense--a live and breathing example of how else they could
do things.
Alexa's major value in the first year of its being acquired by Amazon was to
take some of the lessons that we had learned of how to do things much cheaper
than they had. They had gone through an explosive growth phase, and they
were spending $100 million a year on hardware. We couldn't believe it. Here
was this little company that had been living its whole life, and we hadn't spent
$10 million.
So Jeff said, "OK, Brewster, you know how to do this stuff cheaply. What
should we do?" I said, "You should stop buying hardware. You've bought
plenty." He said, "OK, we're going to stop buying hardware." It caused enormous
pain to their organization, but it was the right thing to do. They needed to
become profitable. They learned a lot of lessons from this. They could use an
outsider that was still inside. We were not as independent as a Bain Consultants
or something like that, but we knew what we were talking about because we
had actually built stuff. We dropped the cost of their Internet connections by
90 percent just by saying that you can go and negotiate deals in this and this
way. So we paid for the acquisition of our company in the first year just by the
capital costs that they saved.
Brewster Kahle 279
AOL took our ideas and put them into their bloodstream and dispersed
us--and properly, I'd say. Amazon kept it to the side and kept it moving and
generating new ideas. Amazon spent the time; we had basically a full-day
meeting with the top execs of Amazon every month. It was just an outrageous
amount of time given to us, but that was because Jeff Bezos said "New ideas are
going to come from these guys."