Marc Andreessen on Building Netscape & the Birth of the Browser

Host (Ben Horowitz): It was this invitation to basically do anything. It was this kind of great moment where you just kind of felt like the future was like in front of you.

Guest (Marc Andreessen): It's very important to know that at this point and all the way up until really 1994, the overwhelming uniform universal expectation was the internet would never be a business.

Host (Ben Horowitz): I at the time completely lacked the skill set and the perspective to navigate, I would say, interpersonal and in particular bureaucratic situations. I had no clue how to do any of it.

Guest (Marc Andreessen): So much of computer science at that point was about optimizing scarce resources because that was all you had at that time. And they had spent decades figuring out how to do that. And we collectively decided to just break that rule.

Narrator: You've probably heard of the browser. In fact, billions of people use browsers as their gateway to the internet. And you might even be using one right now to access this very podcast recording. Now, in today's episode from The Ben and Mark Show, Mark Andreessen and Ben Horowitz share the real story behind the creation of Netscape, a web browser co-created by Mark that revolutionized the internet and, quite frankly, changed the world. As Ben notes up top, until today, this story has never been fully told, either in its entirety or accurately. In this one-on-one conversation, Mark and Ben discuss Mark's early life and how it shaped his journey into technology, the pivotal moments at the University of Illinois that led to the development of Mosaic, a renegade browser that Mark developed as an undergrad, which is widely referenced as the first widely used browser, and the fierce competition and legal battles that ensued as Netscape rose to prominence. Ben and Mark also reflect on the broader implications of Netscape's success, the importance of an open internet, and the lessons that still resonate in today's tech landscape, especially with AI. That and so much more. I hope you enjoy.

Narrator: Who will decide the future of the internet? *Read Write Own*, a book by A6NZ general partner and author Chris Dixon, takes on one of the most consequential questions of our time, which is also the key to unlocking more entrepreneurship, more creativity, and more innovation. From AI that compensates artists to social networks that reward you for participating, *Read Write Own* shares a playbook for building the next era of the internet. Learn more at readwriteown.com.

Host (Ben Horowitz): Okay, you idiot. Like this giant commercial opportunity is staring you in the face. You literally have like inbound sales leads. It's like coming out of your ears. Like, why don't you go raise venture capital and start a company?

Guest (Marc Andreessen): Yeah. And of course, the answer was because I had no idea that there was such a thing as venture capital.

Host (Ben Horowitz): Yeah, what venture capital? I literally...

Guest (Marc Andreessen): Yeah, you know what a tractor was.

Host (Ben Horowitz): Yes, exactly.

Narrator: The content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investor or potential investors in any A16Z fund. Please note that A16Z and its affiliates may maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see A16Z.com slash disclosures.

Host (Ben Horowitz): Welcome to the Mark and Ben Show. Today is a super special episode because we are going to talk about the origin of the web browser and the invention of the web browser. And we have one of the co-inventors with us right here, Mark. So it's exciting. It's also exciting because this story has never really been told, either in its entirety or accurately. And so we're going to get a chance to do that. For those of you who are so young that you're not quite sure what a web browser is anymore, it is kind of how most people experience the Internet. So you might call it the Internet. It's a thing, Chrome. Let's start at the very beginning. One of the things that is one of the largest disinformation campaigns going is this whole idea that people, entrepreneurs and people who invent things are kind of born with a silver spoon in their mouth. And almost none of the successful entrepreneurs we work with is that the case for. They all come from somewhere between like refugee and middle class backgrounds. And you certainly were not born with a silver spoon in your mouth. So tell us a little bit about where you grew up. You didn't grow up in a big city. You grew up in quite a small town. What was that like? And then what was it like growing up and how did you first encounter the internet?

Guest (Marc Andreessen): Yeah, so let me, if I could, start our session today with two disclaimers, which maybe relate to your question. Disclaimer number one is we're going to be talking about events that happened over 30 years ago. A little memory problems. So I'm going to tell the truth as I remember it. I may get things wrong or other people may have different recollections. And so I have to disclaim that. I can't swear to the factual accuracy of stuff that was that long ago, but I'll tell the story as I understand it. And then the other story is there are twists and turns along the way where I would just characterize it as at the time I was irritated at other people for things that happened. I think in the fullness of time, what I realized is that I just at the time completely lacked the skill set and the perspective to navigate, I would say, interpersonal and in particular bureaucratic situations. I had no clue how to do any of it.

Host (Ben Horowitz): Well, and so I think when you get into your background, people kind of understand.

Guest (Marc Andreessen): I sometimes describe myself that I was feral at this point in my life. And so, it's always this thing, if you could rerun prior events, like with the skills that I have today, I could have navigated it much better and a bunch of things could have turned out different. But I certainly did not have the skillset at the time. So anyway, if it sounds like I'm criticizing other people, it's actually not what I intend. It's going to be more ultimately criticizing myself in the sense of whatever happened that I didn't like at the time, I think I was at least partly, if not wholly responsible for.

Host (Ben Horowitz): You know, young men can be like that from time to time.

Guest (Marc Andreessen): Yes, I would say I was raw aggression at that point with very little. And yes, Ben, of course, remembers this because he met me shortly after the events we're about to talk about. So first of all, the modern cliche is, you know, Elon Musk's father had an emerald mine. When he showed up in the U.S. with $2,000 is the actual story.

Host (Ben Horowitz): Yeah, exactly. So there's all these kind of fake histories that are kind of retconned into people who grew up in some cases pretty tough backgrounds.

Guest (Marc Andreessen): So I grew up in the American Midwest in rural Wisconsin. And for people who haven't spent time in Wisconsin, there's basically three Wisconsins. There's the big city of Milwaukee, which is like a bit, almost like a Chicago kind of thing. That's its own world. And then there's Madison, which is like a very kind of hippie college town that was actually kind of a core of activity in the 60s. And then there's the rest of Wisconsin, which is completely disconnected from those two cities, which has nothing to do and has no interface into those two places at all. And it's the rural Midwest, which is basically a farming country.

Host (Ben Horowitz): I remember like Grant's actually, as I understand it.

Guest (Marc Andreessen): Yeah, I think that's right. If you get 10 minutes outside of either Milwaukee or Madison, you get into real, real rural territory in a hurry, right? And so there were people in the farming world in the Midwest out through the 1970s that still didn't have indoor plumbing. I have memories about houses.

Host (Ben Horowitz): Well, for a while, you didn't have gas heat, so you weren't that far out.

Guest (Marc Andreessen): There's also that, exactly. So yes, if you want to heat your house, you cut down some wood. And so look, it's up north and it's very close to Canada, so it's extremely cold. And so it's actually this amazing thing where it's frozen tundra for nine months and then three months of summer. But mostly farming, a lot of dairy farming, a lot of corn to feed the cows. So corn and cows. And then I would say light manufacturing, light tourism, a lot of Illinois people vacation up there, go hunting, fishing or whatever. So a little bit of that, but mostly agricultural. My town was sort of the sign of the outside of town forever. Population 1,309. The sign never changed. You were in it regardless of whether you had 1,308 or 1,310.

Host (Ben Horowitz): Yeah, I don't think it actually moved around that much.

Guest (Marc Andreessen): So it was probably more or less accurate the whole time. And then, of course, the running joke was that's 1,309, including the cows. So, yeah, it's a very small town environment, kind of lower middle class context and public K through 12 school. My school was very small, 25 kids in a class. So, I feel like, whatever, 300 kids in the whole school.

Host (Ben Horowitz): A lot of people don't have school choice today. You didn't have teacher choice either. Because there's, I guess, one teacher for every subject.

Guest (Marc Andreessen): No, no, exactly. And then, this is the 70s and then the 80s. And so, this not only predates the internet, but this also predates cable TV. We had no cable TV. Long distance phone calls were still a dollar a minute. So we saw actually my neighborhood growing up, actually, we had a party phone line for the entire neighborhood.

Host (Ben Horowitz): That's interesting.

Guest (Marc Andreessen): Which is actually quite an adventure. And so there's a single phone number for the neighborhood. And so when your phone rings, everybody in the neighborhood picks up the phone. This is true. This is true. And then there's an arbitration process, a verbal arbitration process for who the call is actually for. And then the expectation is that everybody else hangs up, but they don't have to.

Host (Ben Horowitz): Well, they say there are no secrets in the small town. This is one of the reasons why.

Guest (Marc Andreessen): So yeah, it's halfway between the 1930s and the 1980s or something. It just is one of those places where a lot of the country was like that at that point, which is it hadn't fully adapted to certainly the things that we all now take for granted. And then you asked why to discover the internet. So I had no idea the internet even existed until I went to college. The thing I knew that existed, and this is where I kind of got lucky in terms of when I grew up and when I came of age, is I came of age precisely at the moment when the PC happened. And so I was aware of, there were all these just amazing stories. This is when Steve Jobs was on the cover of Time magazine. And so there was this moment when the personal computer hit the popular consciousness. And it was built in the late 70s. And then it catalyzed hard in the early 80s, especially around 82. And then, our school, even though it was very small, started to get a handful of early computers. And then there were a wave of consumer computers at that time that were actually quite inexpensive. And so there were computers at that time as cheap as $200 in currency at that point.

Host (Ben Horowitz): And this was our floppy disk. So this is pre-hard drive or post-hard drive when you got your first computer? Pre-floppy disk? Tape.

Guest (Marc Andreessen): Cassette tape, exactly, yes.

Host (Ben Horowitz): Okay, okay. That's a slow loader, bro. You can power to load a game. I remember that era.

Guest (Marc Andreessen): So for people who don't remember, but before floppy disk, you would literally hook up a cassette tape player. And the thing with a cassette tape player is it didn't have any kind of seek capability. And so the way that you loaded the program was you had to fast forward by hand to the right point of the tape.

Host (Ben Horowitz): Right, and then you were always at risk. If you were going to write to the tape, you wrote a program, you wanted to write it onto the tape, you ran the risk you were going to overwrite something in the past.

Guest (Marc Andreessen): And then there was this really fundamental trade-off, if you didn't have a lot of money, is this very fundamental trade-off, which is you could buy a short cassette tape that was high quality, or a small amount of stuff, or you could buy, same price, you could buy a cassette tape that was much longer and could record a lot more, but at much lower quality. And that mattered because the lower quality tapes, frequently you would not be able to read back what you had written. And so there was a real quality quantity trade-off.

Host (Ben Horowitz): Yeah, this is a gambling exercise.

Guest (Marc Andreessen): Exactly. And then, so we didn't have the internet, but we had no exposure to the internet. We'll talk about the internet prehistory in a little bit, but we didn't have that. What was happening at that time was what were called BBSs, which was an acronym for Bulletin Board Systems. And the way to think about these is BBSs were kind of pre-social networks in a way and pre-internet. And so the way this would work is the host of the BBS would literally set up a set of modems in their house or apartment, often like 8 or 12 or 20 or something, to take incoming calls from people with remote computers. And then you, in theory, you could dial into BBSs and you just literally used a modem and you dialed into the phone number for the BBS. And then if you got on the BBS, it was really cool because you had access to early versions of like email and social networking and user profiles and bulletin boards and classified ads and downloading games and playing games and so forth. And so it's kind of pre-internet, pre-AOL kind of versions of these things. The problem that I had was, again, rural Wisconsin, long-distance phone calls were a dollar a minute. There are no BBSs in my town. And so I read about BBSs. I actually don't. To this day, I think I never actually used one because I couldn't afford it. And by the way, also, this is also predates broadband. When I first started on this stuff, this is actually pre-modem as we understand it today. So the form of the modem at the time was what's called an acoustic coupler. And so you take your old-fashioned telephone handset and you literally put it in these two rubber cups.

Host (Ben Horowitz): Yeah, I know. I remember those.

Guest (Marc Andreessen): In the two rubber cups. And then it's literally using your handheld phone as the received transmit for the audio signals. And so the acoustic coupler modems were 300 baud, 300 bits per second.

Host (Ben Horowitz): Yes, and very noisy 300 baud, yeah.

Guest (Marc Andreessen): Very noisy, exactly. So very slow. So you could kind of get a glimpse. I would say the romance of the personal computer at that point was very clear. And that's what really got me was, and basically the way that you'd bought a computer, you plugged it into your TV set in those days. And what happened was you literally got injected into the basic programming language interpreter. And what that showed up as is that it literally, the screen would say ready. And then there was a cursor. And for kind of kids my age, it was this invitation to basically do anything. And so it was this kind of great moment where you just kind of felt like the future was like in front of you. And then there were all these, this became a pop kind of culture thing. And so there are all these books and magazines that you could buy or subscribe to, hobbyist magazine for whatever kind of computer you had. And it would literally have printouts in the magazine of programs that you could actually sit and type into your computer, type into the basic prompt and make work. And so there was this incredible sense of adventure for what you could do on a computer. And then there was this additional thing out there, which was, wow, if you could afford it, you could be on BBSs and you could talk to other people. And so they kind of had that network thing from the beginning. It's just the economics at the time was not feasible to have it be a mass market thing.

Host (Ben Horowitz): Interesting, interesting. Okay, so then you get the University of Illinois. And why did you go there? Why not Madison or MIT or what have you?

Guest (Marc Andreessen): Financially, the state schools were the options. And then Illinois, it turns out, then and now is a great engineering school. And so Madison's very good, but Champaign-Urbana is one of the top engineering schools in the country. And so it was just a nice coincidence that that was kind of close enough and inexpensive enough. It also happened to have national computing, supercomputing.

Host (Ben Horowitz): So this is where I got really lucky. So I dropped into Illinois as a new student, started out in EE and then decided I did not care at all about electrical engineering and switched into computer science, which was a much better fit. But I showed up as a CS student. And basically this was my big stroke of luck was this was University of Illinois in 1989. So this was four years into two federal programs that basically created the precondition for everything that followed with the internet. And the two programs were something, I think it was called the National Supercomputing Act. And it was basically this effort to fund what were called National Supercomputing Centers at four universities, one of which was Illinois. And so the campus, the federal government just dropped in a ton of money to basically buy state-of-the-art computers, including at the time really big computers like these big crayon thinking machines computers that cost like $25 million at the time and filled up entire rooms. Those supercomputers in those days were so big that in some cases you'd actually build a building for them and you'd build the building, but you wouldn't close the roof and you would lower the computer by a crane down through the roof into the center of the building and then you would finish the roof.

Host (Ben Horowitz): The big iron.

Guest (Marc Andreessen): Exactly. So very esoteric, expensive, powerful systems. So we had those. And then the other federal program was a program to build what was called the NSF-Net. The NSF, they were being short for National Science Foundation, which is the government agency that all the money came through. And the NSFNet basically was the first internet backbone, as we understand it today. And those programs were joined because the original purpose of the NSFNet was to connect together these supercomputing centers and then to allow researchers, scientists in many other colleges, universities across the country to be able to remotely access these large centralized supercomputers. I want to take a moment here to kind of pay credit to Al Gore on this, who famously gets just like endless shit for people kind of saying that he said that he invented the internet. And so just to defend Al's honor for a moment, A, he never actually said he invented the internet. What he said is he took the lead in the Senate in creating the Internet. And what he meant by that was not that he sat down and wrote the code for it. What he meant was he was one of the real leaders, one of the main forces in the Senate to actually fund these two programs. And these two programs led directly to the Internet as we know it today. And so he and his colleagues at that time really stepped up at a pivotal moment. And I think that's actually very relevant to kind of what's happening today with AI. Like, I think that's actually the same thing that needs to happen with AI today.

Host (Ben Horowitz): Well, it's kind of the opposite of what's happening with AI in universities today, which is they're not only underfunded to do AI, but there's a push among the big tech companies to enact legislation that would essentially outlaw AI in universities by eliminating open source. So full credit to Al Gore for doing the right thing because it's clear that that wasn't obvious.

Guest (Marc Andreessen): That's right. We've been spending a lot of time in Washington lately, and I've actually been telling the story to a lot of current senators and encouraging them to basically do the same thing.

Host (Ben Horowitz): The Democrats love it because it rehabilitates Al Gore. The Republicans, I got to get them through the Gore part. But everything partisan in 2020. But they also seem to think it's a good idea. So I hope that will happen. So basically, my great luck was I showed up at Illinois in 89 and basically four years into these federal programs. And so I showed up basically just got blasted into the future in one step. Because when I showed up there, there was all the University of Illinois was actually wired. It was one of the hub nodes of the internet backbone, and the campus was getting wired for broadband. And there were computers and computer labs, it was stated there at equipment, all the way up to these giant great supercomputers that the computer science department had access to. It was just there. It was like being beamed into the future. Now, the twist on that is the assumption in those days was you would use this stuff while you're in school. And so they had started to give out email addresses to undergrads and things like that. But the assumption was that you would use these systems while you were in school, and then if you stayed and became a faculty member or something, you would use these systems for your research. But if you graduated and just went out and went into the real world, you would leave it all behind? The internet, right? There's no access to the chains, right? Maybe you could kind of describe what the internet was at that point. Because when we say internet, people imagine a lot of things, but that's not what it was then in terms of a user experience.

Guest (Marc Andreessen): That's right. And so a couple of things. One is it's very important to know that at this point and all the way up until really 1994, the overwhelming uniform universal expectation was the internet would never be a business. There would never be a business. There would never be an industry. There was never going to be money to be made. There was never going to be commerce. There was never going to be streaming video. There was never going to be stores, any of the stuff. By the way, even the idea of having like newspapers online was considered bizarre. It wasn't even going to be that. It was supposed to be like scientific research papers and experimental data and things like that. And so it was very much not viewed as like a commercial opportunity. It was viewed very much not that way. And basically, I think, if I remember it had been up until like 94, I think zero of the big tech companies of that era took it seriously.

Host (Ben Horowitz): Zero.

Guest (Marc Andreessen): And in fact, well, in fact, the opposite. They were building kind of parallel systems to the internet, often referred to as the information superhighway. You know, Bill Gates was a huge kind of champion of the information superhighway and specifically not the internet.

Host (Ben Horowitz): And he was kind of the biggest figure for sure in software and probably in the industry at the time.

Guest (Marc Andreessen): So the big computer companies wanted to build proprietary networks and they were doing that. And by the way, AOL was up and running by 94. AOL kind of got going around 89, kind of the same time I showed up at Illinois, but it really kind of hit critical mass. AOL basically was a consumer-scaled version of the BBS idea that we described. And then in 93, it famously interconnected itself with the internet. And so all of a sudden, all the AOL subscribers became internet users. And so there was that, but that didn't even happen until 93. There was certainly no vision for that, I think, in the late 80s. And there were no consumer ISPs, and then there were no normal businesses online. And in fact, the internet up until 1993, was the NSFNet, which then turned into the internet, that operated under something that the federal government dictated called the Acceptable Use Policy, the AUP. And because the NSFNet was federally funded, commercial activity on the internet was actually banned, because it was viewed as a sort of inappropriate use of federal research dollars. And so it actually would not have been legal to engage in commercial activity. And so there basically was none. And then, the big phone companies and the big media companies at that time didn't even want to build like BBSs or anything internet related. What they wanted to do was basically what we now know as streaming. They wanted to do what they called at the time interactive television. And the big killer app for internet television was called Video on Demand. And so it was the revolutionary idea at that time that instead of watching whatever was on a TV channel at that moment, you could watch whatever you want by clicking a button and then maybe you could order a pizza.

Host (Ben Horowitz): Interestingly, there was a small minor bandwidth problem with that idea, if I recall correctly.

Guest (Marc Andreessen): It was actually an idea that was ahead of its time, right? Streaming video didn't really work in the way that we understand it today until probably 15 years later.

Host (Ben Horowitz): Well, really when Netflix made the cutover from CDs to streaming, which was the late 2000s, right? I remember meeting with Reed Hastings in, I think, 2004. And Netflix was up and running and was very successful doing DVD rental by mail. And Reed's a technical genius in addition to being a business genius. And he said, I'm thinking about doing streaming. And my first reaction was, that's crazy. It'll never work. Because even in 2004, like most people on the internet, they didn't have broadband connections fast enough to do like TV quality. So video at that point was actually these little postage stamp size videos. And I was like, oh, I don't understand who's going to want to sit there and watch a little postage stamp size thing. And Reed had correctly extrapolated that the broadband wave in the 2000s was going to result in streaming working. But just to give you a sense of the delay there, even a decade later, even a decade after the founding of Netscape, development of Mosaic and founding of Netscape, in a decade later, it was still considered weird and bizarre to try to stream videos.

Guest (Marc Andreessen): So the media companies in the early 90s were not being run by technologists. And so they had a hard time, I think, mapping. There was a famous interactive television information superhighway trial in the early 90s around this time, which was to do this video on demand. So to do this thing where you'd have a remote and you could press a button and watch whatever movie you wanted. And I remember it was a trial. They were trying to figure out if there would be consumer demand for this. And so one of these companies, they wired a neighborhood to do the streaming of a video, but it wasn't digital switched, it was analog switched. And so they had a dedicated long, basically analog line wired to each house. And then in the back office, they had a bank of VCRs, videotape players. And they had a videotape player for each house. And then they had a wall of video cassettes with all the movies that were optioned. And then they had a guy on roller skates.

Host (Ben Horowitz): Who did he have roller skates to get there quick enough?

Host (Ben Horowitz): Because the user expectation was you click the button, you watch the movie. And so the guy on roller skates had to shoot over to the wall of videotapes, pull down the right tape, and then shoot down the hall to the right VCR and get the tape in the VCR and press play before the couch potato is giving up because the stream business started. Oh, my God. And so, yeah, that idea was a bit early.

Guest (Marc Andreessen): But on the University of Illinois campus, like broadband existed. Like I said, email existed. You asked how do people use the internet in those days. So sort of pre-web. It was mostly, well, there were leading apps at that point. There was an app called Telnet that you would use to basically log into another computer on the network. And that was actually very important because that's how the scientists would use the supercomputers. And actually the group I was in at Illinois actually built one of the main Telnet apps. There was an app called FTP that was a file downloading app. And so you could upload and download files. There was early email. So that worked. There was early what were called news groups, which is like basically forums, kind of early social networking.

Host (Ben Horowitz): I think at the time, those were probably the four main things that people did. And it was all scientists and computer science majors on the Internet, if I recall. I mean, there was really nobody else.

Guest (Marc Andreessen): Yeah, so this is at the time, 89. I'm going to guess there were somewhere between 500,000 to a million people total online. And yeah, it was basically the faculty, staff, and students at these four supercomputing centers. It was the remote users. And then it was like the defense contractors got wired up early. And then there were branches of the government that got wired up early. And then there was like the national labs. And then a handful of hobbyists would figure out a way to get online. And so, yeah, so basically it was, you know, first of all, it was like 100% people in the West, you know, overwhelmingly the U.S. and Europe. It was, you know, very heavily, obviously, English dominated from the very beginning. It was extremely technical, scientific oriented. Almost everybody on it had a scientific or technical degree. And it was also very, as a consequence of all that, you know, it was this incredible, you know, brilliant, you know, this is like a million of the smartest people on the planet. So the caliber of the people and the quality of the discussions was like sky high.

Host (Ben Horowitz): Oh, yeah. I remember the old news groups were unbelievable. I remember like if there was like a bug in a compiler, like you could find out about it and, you know, there would be workarounds. And like the level of expertise on those things was absolutely astounding.

Guest (Marc Andreessen): And look, many of the smartest people in the scientific and technical world at that point were in there and they would talk to you if you had something interesting to say. And so if you posted on a news group, they would respond. And so it was like this distributed community of like, you know, the smartest scientific technical minds on the planet. It was really special. And then also because there was no money, you know, there were no ads, you know, there were no scams, there was no fraud, there was no, you know, spam. You know, like I remember there was actually a scandal that there was a guy who figured out that this was like a right for, you know, basically scam, you know, kind of thing. And he started, did the first spam emails and news group messages. And it was like a big scandal at the time that somebody would actually do that because like a guy just literally started to do that. Right. But, you know, it had run for years without anybody even trying that.

Host (Ben Horowitz): Yeah, that's what a community. That's the true meaning of community, right? Like you have all people of the same culture in one place. That's scale. That's amazing.

Guest (Marc Andreessen): Yeah, so people who were on that at the time kind of always look back at that and they miss it because that has never been reconstituted. You know, look, a lot of those people in their equivalents today are like on X or they're on the other social platforms, but they're a minority population in a much larger context everywhere now, whereas at the time they were the entire population. Yeah, so that was a super magical thing. But yeah, I mean, look, the actual functional use cases were quite limited. And yeah, so yeah, that's sort of what I saw when I got there. Maybe it'd be helpful just because I think it's really relevant both to what followed, but also the AI discussion. Maybe I could go back in history now to kind of where the internet came from.

Host (Ben Horowitz): Yeah, yeah, go back to the late 60s.

Guest (Marc Andreessen): Yeah, so the idea of the internet was sort of famously this idea of what was called a packet switch network, right? So the reason the internet works is because it's this peer-to-peer system in which anybody can kind of plug a computer into the internet, and then you kind of have messages that are able to go around and route. And so, you know, famously, this is an idea that actually was developed originally by the Defense Department in the 1960s called packet switching. And in particular, there's a guy, Paul Buran, who's kind of the original, you know, kind of true founding father, you know, godfather of this idea. And actually, he's a great story on this, going back to your silver spoon thing. So he was born, he was a Jewish-Polish immigrant to the U.S. And so his parents brought him as a small child to the U in the 1920s sort of classic American immigrant success story His father opened a grocery store and then was able to make enough money to send his kid to college Baran you know was a super genius He got an engineering degree and then he actually went to your alma mater He went to UCLA for a master's degree, which he got in 1959 in computer science. By the way, his master's degree in 1959 in computer science was on character recognition.

Host (Ben Horowitz): Wow. Amazing. AI.

Guest (Marc Andreessen): Right, AI. So like he was actually, so it's like he tried to do AI first and, you know, he was early on that. Yeah, he got partway there, but he was already at that point trying to do AI. And then he went to work for the Rand Corporation, which did a lot of work for the Defense Department on military strategy topics. And this is the height of the Cold War. So this is in the early 60s, Cuban Missile Crisis, where there was a real feeling that there might be nuclear war at any moment. And one of the big concerns the Defense Department had was, in a nuclear strike, the way telecom systems worked at that point is if you took out the central office, you took out an entire telecom network. And so the fear of the Defense Department was the Soviets presumably knew where the central switching offices were for like the AT&T network. And so they would bomb those, you know, in a nuclear strike. And then basically what would happen is the U.S. would lose command and control of its nuclear weapons. And so basically it was a way, the fear was it was a way for the Soviets to do what they called a decapitation strike. They take out the central switching office from AT&T and then the U.S. can't retaliate. Literally the U.S. would not be able to fight, you know, to fire its nukes because they couldn't send the commands to actually fire the nukes. And so it was this existential kind of thing at the moment. And so this guy, Paul Buran, said, well, what if we build a network that basically is designed for packet switch instead of being circuit switch or everything's going through a central place? It's called packet switch where the packets can flow around. And if part of the network gets bombed or destroyed or taken offline or, you know, power cut or whatever, you know, the network kind of reallocates. And the significance of this idea is not just that it led to the Internet. The other significance is AT&T thought it was the craziest idea they'd ever heard.

Host (Ben Horowitz): Yeah, yeah, right. Because you need control. You need control. For control, there's no central control.

Guest (Marc Andreessen): Yeah, yeah. And I was reading, just before this, I was refreshing myself, and I read Paul Brand's obituary in the New York Times, and Vint Cerf is quoted in the obituary saying that, you know, Paul Brand was basically laughed out of AT&T when he proposed this idea, right? Because it was such a heretical idea. So anyway, so the point being is the internet was a heretical idea from the very beginning. And I thought both in that it would never work because it would be, like, entirely too slow and the packet reassembly and all that would never work, And then also that it was a dumb idea because, of course, he wanted central control. So, like, they derided him in every direction possible.

Host (Ben Horowitz): Yeah, yeah. Well, as late—so that was 1964. 20 years later— Horrible idea that would never work. Horrible idea that will never work. Exactly right.

Guest (Marc Andreessen): So that was 1964 when he proposed this. As late—20 years later, 1984, a federal judge went to break up AT&T, which at that time was the national telecom monopoly. They owned and controlled everything. And AT&T actually got the Secretary of Defense in 1984, 20 years later, to testify that if the federal judge broke up AT&T, it would permanently cripple the U.S.'s command and control capability for nuclear weapons. And so even 20 years later, people didn't believe it. It was still a heretical idea. And by the way, 84 is significant because, of course, the very next year in 85 is when the NSF net was funded. And so it was like a 20-year journey to get from that original heretical idea with resistance all the way through to ultimately get to the point where Al Gore and his colleagues figured out that actually, no, it was actually a good idea and they should actually put money in it.

Host (Ben Horowitz): So what I'd really like to know, and I think a lot of people would, is so you have this network with some supercomputers on it and a bunch of scientists and some news groups. Like, how did you get to this idea? because nobody else had the idea. I'd say, like, first of all, nobody else had the idea. And then secondly, if you had not had the idea when you had it, the internet, as we know it, probably doesn't happen in that the kind of Microsoft alternative, the Oracle alternative would have had an opportunity to gain a network effect. And so the chance of all this getting built, I mean, maybe if it was invented six months later, but four years later, definitely not. And so what happened?

Guest (Marc Andreessen): Yeah, so there were serious efforts underway. You alluded to this, but there were serious efforts at the time. So let's just itemize the efforts. And so there were three, basically, consumer online services that came out of the 80s, AOL, CompuServe, and Prodigy. AOL became famous later, but much larger. CompuServe kind of petered out at one point. Prodigy was actually an IBM joint venture, so they were actually kind of clued into that part of it early. And those systems existed in the early 90s. And those were all these proprietary, you know, sort of proprietary stovepipes, not interconnected, not open. You know, these are by definition closed systems. And for people who don't remember this, if you wanted to like put content up on AOL, you had to go, you had to pay AOL. Like you had to like, you had to get, and you had to get their permission, their approval, and they could take it down. And so it was, you know, any of these stovepipe proprietary things, you know, somebody really is in charge. And the internet is the opposite of that. And so that, you know, those systems ran like that. And then there were a set of companies that then were going to do, you're alluding to, that were going to do the leapfrog on that and kind of do the modern GUI version of that. And that was specifically Microsoft with MSN. At the time it was called MSN. And then Apple had something at the time called eWorld. And then who else was running around? A whole bunch of these other companies were running. And then the media companies were doing these proprietary video, you know, interactive TV things that would also be centralized, controlled. But yeah, in particular, I think probably at the time it was really the big ones would have been, you know, it would have been AOL getting really big. It would have been Microsoft establishing its own, you know, kind of permanent proprietary, you know, online service as an alternative to the internet. And then it would have been probably Apple, you know, with its own proprietary system, probably would have been the big three. And look, those companies would have loved for that to happen. Like all three of those companies would have been much better off had that happened. You'd get a VIG on every transaction.

Host (Ben Horowitz): Yeah, this was actually the famous, Ben just used the word VIG. So there was a famous interview with Microsoft CTO at the time that they were trying to make MSN work. Nathan. Their proprietary system. And interesting guy. And he said, he said, yeah, he said, And he actually, to his credit, he kind of said it out loud. He said, yeah, the goal of this program is to get a VIG of every online transaction. You know, VIG being the mafia term for, you know, your slice of the pie.

Guest (Marc Andreessen): Vigorous. Vigorous. Exactly. So, and you know, yeah, they would have had, you know, they would have had total control. And so if you think about the level of control that big tech has today, like it would have been like that times 10. And that really was where the industry was headed. And those companies had a big advantage at the time because one is they just had tremendous resources. Like they were funded to do this. And then the other thing is that, you know, at the time, you know, look, to just get this stuff to work was hard and they could marshal thousands of engineers and they could design everything to work in a completely integrated manner. And they had, you know, all these graphic designers to make it beautiful. And they had all these performance engineers to make it fast. And, you know, they could run big advertising campaigns and do consumers, you know, do consumer support. And I'll tell the story later of how I became the consumer support. I became the user support desk for the entire internet for about a year and a half myself. It's an amazing job. But, you know, amazing job. They would have had, you know, They would have had the ability to, you know, they had a lot of natural advantages to be able to do it in the proprietary way that they wanted. And so this was a critical period for that.

Host (Ben Horowitz): Yeah, the Internet, as we know today, didn't have to happen.

Guest (Marc Andreessen): Yeah, so basically, I think in retrospect, what just happened was a couple of things. So one is just the generation of super geniuses like Paul Buran and, you know, your friend, you know, Len Kleinrock and Vint Cerf and, you know, all these really bright guys who had created the Internet as, you know, kind of the network side of it. And they were networking people and their natural world was not user interfaces and consumer services and content and media and gaming and all these application level things. The way these systems are built is they just assume somebody else is going to do all that. And that hadn't really happened yet. And so, yeah, part of it was people, they hadn't tried. Part of it was it wasn't a business. And so there was no business motivation to try to do it. And then also part of it was there was another coincidental thing that happened, which was the arrival of the graphical PC. And so the arrival specifically of the Macintosh and then followed by that, the arrival of Windows version 3 and the first graphical PCs. And so before about 1992 or so, you weren't going to have a graphical user interface to anything because you didn't have graphical user interfaces at all. And so there was also a moment in time thing that happened there. So there was all that. Yeah, and so then my part of the story was I had a bunch of jobs in college and was getting my computer science degree. And then I ended up working for what's called NCSA, which was the National Center for Supercomputing Applications.

Host (Ben Horowitz): And were you a freshman at the time or a sophomore when you got that job?

Guest (Marc Andreessen): No, so I actually didn't start working there until I was a junior. So when I was a freshman, I worked in a physics lab called the Materials Research Lab there, which was actually a great entry point for me because that was one of the main labs using the big supercomputers. And so I kind of got plugged into that world very early. And then my sophomore year, I spent nine months actually as a co-op student working for IBM in Austin. And I worked on the workstations at the time, which was, again, very, very helpful for this because these were kind of the leading edge user interface. You know, so I worked on graphics. And they had TCPIP, right? The workstations, yeah. They were built in. And then IBM in Austin at the time, you know, this is when IBM was on top of the world. It was, you know, by far the most important tech company still at that point. and we had this 6,000-person division in Austin working on these Unix workstations at the time, graphical workstations. And again, they had the resources. Everybody there was also wired. Everybody was on the internet. These workstations were designed to be used on the internet.

Host (Ben Horowitz): Right, right. And I was- At the time did not have any internet capability. You know, there was no, in the early 90s, there was no TCP IP. There was none of it.

Guest (Marc Andreessen): Yeah, well, computers, you recall this. Yeah, computers, PCs and Macs up until 93, 94 didn't even come with TCPIP built into the computer, built into the operating system. You actually had to buy what was called a TCPIP stack as a separate thing. Winsock. Winsock and these things. And so, yeah, but the Unix workstations of that era, the problem with the Unix workstations is they cost like $50,000. And so these were not consumer products. But if you had one, you had internet networking built into it. And then if you were at IBM like I was or at Illinois like I was, you were also on the internet. And so I kind of got to see, I kind of saw the bricks being put in place. for what followed. And then I knew how it all worked because my job was to do engineering on all these things. And then basically there was this moment in basically, I think it was 92, 91, 92. There was this moment where there were basically three online efforts to kind of do front ends, new kinds of sort of user interfaces, interaction models for the internet at that point. And so, and it was famously, it was three of them. There was one called Gopher. There was one called Waze. And then there was the World Wide Web, Dembrenners-Lee. And so, and these were in the very beginning, And these were actually in a real bake-off with each other. There were a lot of people who had different opinions about which of these were going to win. Gopher was actually based on the user interfaces of the BBSs from the 80s. And so it was a menuing system. And so you could go down all these different menus and download content and so forth. And then Waze, W-A-I-S, I think it was called Wide Area Internet Search. Waze was like a pre-Google search engine. And so it was sort of the idea that you would type in search keywords and get back results. And then this guy, Tim Berners-Lee in Switzerland, English guy working at CERN in Switzerland, had this kind of really, at the time, radical idea, which was to take an old idea called hypertext and bring it onto the internet and basically be able to have documents that can link to other documents. And this is an idea that goes back to the 50s. A guy named Doug Engelbart and another guy, Ted Nelson, had kind of conceived of this idea decades earlier, but the computer systems in the 50s, 60s were not quite capable of doing it yet. And so Tim basically said he was sitting at CERN, kind of doing related kinds of work. He was working in kind of support of the physicists at CERN in the same way that I was working in support of the physicists at Illinois. So kind of a similar universe. And he had this idea basically, okay, we're going to put hypertext on the internet in the form of the web. But I would describe this at the time as all three of these were like very, I would say, nascent experimental efforts. And then really critically, all three were text-based. And again, the assumption here was the internet, basically, the assumption was the internet's slow, which for most people at that time it was. And so the assumption was the internet's slow, your computer's slow, your network connection is slow. And so Gopher was literally text-based menus. Waze was type in a text keyword and then get back text results. And then the web was hypertext, and there's a text document, and then it contains links to other documents. But just as an example, the web, when it was originally conceived, didn't have images. Because the idea that the network would have the capacity to be able to do images was just still too much of a reach. And then basically people started, you know, a bunch of people online started to play with this. And then, you know, and then I was, we talked about what I was doing at the time, but I was working on a semi-related project and then figured out that, you know, with a bunch of my colleagues there at the time that there was a, you know, we had ultimately this idea that led to Mosaic.

Host (Ben Horowitz): So what was your other job?

Guest (Marc Andreessen): Yeah, so what I got hired to do at NCSA, what I got hired to do is I got hired into a group called the Software Development Group, which was the group that was basically supposed to make, it was again funded by the National Science Foundation, supposed to make tools, specifically open source tools that would make it easy for scientists to use the NSF net and to use the supercomputers. Right. And so those tools I mentioned earlier, this thing called Telnet. So we made this, we made NCSA Telnet, which was the main way you would log into remote computers. We had, I think at the time, I think we had an FTP client server. We had, you know, so we had like a variety of these kinds of tools that you would use for, you know, for these things. By the way, we also were doing early work in VR at the time. Because there was this whole focus on what was at the time called scientific visualization. And this is sort of what later became special effects in movies. But this was actually pre-Jurassic Park and pre-Terminator 2. And so then the idea was to, you know, these supercomputers would do these black hole simulations or weather simulations or something. And then you could actually use these graphical workstations to actually render movies. And you could actually show scientific results in visual form. And so the group did a lot of that. Actually, those guys, actually, a lot of those guys actually went on and actually ultimately then created the computer graphics industry in both the computer industry and then also in film and television. So that was also a thing that was happening then. And then actually the VR idea was actually already present at the time. And so there were attempts to do VR. And our sister campus at University of Illinois, Chicago actually had something called The Cave, which was an alternate vision of VR. So the main VR idea, of course, was a headset strapped to your face, which is what people have today. at Chicago, they had the idea of the cave, which was, no, you're actually in a physical space and you have giant monitors around you, right? And so you're actually in a-

Host (Ben Horowitz): Like the sphere in Las Vegas.

Guest (Marc Andreessen): Like the sphere in Las Vegas, like the sphere in Las Vegas. And also like the way that a lot of these new movies and TV shows are filmed now in something they call the volume, which are sound stages that are literally made up of walls and ceilings that are giant display panels showing, you know, graphic rendered scenes and imagery. And so I'm in the race, that's become the most practical idea.

Host (Ben Horowitz): Yeah, yeah, yeah, exactly. That works really well.

Guest (Marc Andreessen): So the first TV show that used that technique was called the Star Wars show, The Mandalorian, which is a huge success. And if you watch The Mandalorian, it's actually really, they did a great job on it. If you watch it, it seems like they're outside all the time and they're not. Most of that series was filmed when it seems like they're outside. That was filmed in a very small soundstage in an environment with giant LCD displays on all sides. And then with all the software, it's the control software they have where they could literally shift the perspective of the background scenes to match the motions of the actors. And they'd have overhead lights so they can replicate the sun or sunset. And so actually, so anyway, so that cave idea in 1992 actually is now state-of-the-art in Hollywood.

Host (Ben Horowitz): Yeah. It was too early. We didn't have the quality of the screens back then or quality of the graphics. So it wasn't ready yet, but now it is ready. So it's another one of these back-to-the-future things.

Guest (Marc Andreessen): Yeah, so anyway, so that's what the group at Illinois was doing. And then the specific project I was on was a project called Collage. And the idea for Collage today, you would describe it as sort of a forerunner to Zoom or to maybe Skype or something like that. And it did have the idea of doing audio and maybe hopefully doing video conferencing. But the main thing was like a shared whiteboard and then shared documents. And so specifically, the idea was real-time collaboration. And so you and I are scientists in two different locations. We're writing a paper together. We want to be able to look at the paper at the same time and be able to make edits such that we can see each other's edits. like Google Docs, like the way Google Docs works today. Or to have a whiteboard where you could, you know, you could look at, you know, imagery or you could look at, you know, draw diagrams and you could share them and work on them together and annotate them like you had a shared whiteboard. And, you know, I think that was like a good, I mean, that was a good idea. And like all of that stuff has happened and it's very important today. It was just, I think it was in retrospect, it was just before its time, which is just like the internet.

Host (Ben Horowitz): What's that? Out of order and what you should build first, for sure. Out of order.

Guest (Marc Andreessen): Yeah, well, there were two problems, which is one is it was just hard to get it to work because you needed a certain speed of network performance and graphics capability on the computers to make it work. And it was kind of choppy. And then the other problem was to get a real-time collaborative system to work, people have to be online at the same time. And so you and I, we're collaborating, we have to be online at the same time. And we might be in totally different time zones. So it's hard to get critical mass for a network effect if you have to be online at the same time. Whereas things like email worked regardless of when you're online, so-called asynchronous. and so yeah collage was a good idea it just wasn't quite clicking and then it just kind of became clear to me and a few of the other folks that I worked with there at the time that it's just like okay something else is going to pop here and it's going to pop hard and it's going to be something involving like this you know the web go for waste like this confluence of basically new thinking of user interfaces that's going to take place and then by the way it's basically two big things the two big things that sort of we insisted on the two big leaps was one is we're just going to assume that everybody has a graphical computer We're just going to assume that everybody has like Windows or a Mac or a Unix workstation. We're not going to support not DOS, right? And all systems up until that point, you know, including, by the way, the early, you know, the first web browser, the first web browser was a text-based browser. You know, the Tim Berners-Lee browser, it was a browser that actually was a text-based browser that ran on the NextCube, of which there were maybe 5,000 in the world at that point. Because that was like one of the most graphical machines ever. But he had this problem. He had this problem. He had the problem. So first of all, the problem of just like, there was no graphical content. There was no graphical, you know, you just like, just the assumption was that it was going to be text-based. And then the other problem was we just assumed that the internet was going to be fast. And that was, again, a heretical assumption. That was a heretical assumption at the time, because at the time the internet was really slow. And most people were on very slow connections. And so, you know, the experience a lot of people had the first time they used Mosaic, even on a broadband connection, the first time they had it, is you would literally watch the page load line by line. And then you would watch the images load line by line, right?

Guest (Marc Andreessen): And so, but it was a heretical idea at that point to say, no, we're just going to assume everybody's on a graphical interface and we're going to assume that everybody's on a fast broadband connection and we're just not going to compromise. We're going to build the correct user interface for that new world on those two fronts. And we're not going to compromise to try to be backward compatible with the old text-based UIs or with the old narrowband connections. And in a way that you were a university with no company, no need to sell anything in the beginning. So it was like, fair enough.

Host (Ben Horowitz): Yeah, and my computer, because I worked at IBM and then I worked at NCSA and we had all this money for the government at that point, my computer was an SGI, Silicon Graphics at the time, leading Unix computer company at the time. You know, amazing company, but they made these workstations and the workstations cost $50,000 in 1992. It was like a hundred and some thousand today. But my computer was one of those and on a fast connection. And so I'm just like, look, I'm just going to build it for that. And then my colleagues built it, versions for Windows and Mac. but like we're just not going to compromise for the old hardware that everybody else is on. We're just going to assume that in the future everybody gets something like this.

Guest (Marc Andreessen): Yeah. And then, yeah, to your point, like we were running on federal research money so we had no commercial incentive. We had no reason to go for, you know, large numbers of users or, you know, try to, you know, make money or whatever at the time. So we just basically, again, and the heresy, we just designed for the future. And then there was a little bit, I had a little bit of a glimmer. I wouldn't say I was confident on this but I had a little bit of a glimmer at the time that was like, look, if we design for, if we design for broadband, if it's a compelling enough user interface, it will actually cause broadband to happen.

Host (Ben Horowitz): Is that what's aggressive? Right. This is my favorite philosopher, Nick Land, has this term called hyperstition, which is the idea of sort of willing an idea into existence just by proposing it. It's sort of like pulling the future forward.

Guest (Marc Andreessen): And the idea basically was if people could just see what was possible with a modern Unix workstation on a modern broadband network with what we then built, if they could just see that with Mosaic, they would be like, wow, I need that. And then they would price it and they would be like, oh my God, I can't afford that. But then they would say, well, I need a version of that that I can afford. And then that would be a motivation for the phone companies to start to offer broadband and for the PCs to start to get built in internet connectivity and for people to upgrade from DOS to Windows and all these other things that followed. And so, as I said, I wasn't confident about that, but I had a glimmer of it because I was like, look, if you could get through this knothole and if you could get the world to the other side where everybody has a GUI and everybody has broadband, then all of a sudden, it's just very clear that you have all of these incredibly compelling things that you can do that are impossible otherwise. And so it was kind of a hard shove in that direction.

And then the other, I would say, big breakthrough, or I would say not breakthrough, the other really important conceptual rule that we had at the time, which was sort of consistent with the internet philosophy of the time, was it had to be an open platform. And specifically, it had to be, whatever it was, it had to be where anybody could create servers and anybody could create content. And so it had to be very easy to do that. And so, and you remember in those days, it was sort of famous that you could implement a web server in four lines of Perl script, right? To do whatever you wanted. And so, and you could create a webpage just by writing HTML by hand. And then by the way, people did, and there were scaling issues with those four lines of web servers. That's how a lot of the big internet companies started out that way.

Host (Ben Horowitz): Yes. And then that's one of the reasons why the sites always crashed is exactly, but, but the point was to optimize for the quality of experience and then optimize for the openness and the creativity that would follow. And again, there was a leap there, right? And you remember, we got a lot of criticism at the time, which was, wow, a lot of computer scientists at the time were like, wow, these guys are building the most inefficient computer systems that's ever been built. This thing is incredibly inefficient. It's unoptimized. It's wasting network bandwidth. It's doing all these things. Where's the ASN1 encoding? Like, this is strings. You guys are crazy. You're wasteful.

Guest (Marc Andreessen): Yeah, we were doing wasteful.

Host (Ben Horowitz): Yeah, yeah.

Guest (Marc Andreessen): Yeah, big environmental arguments, you're wasting bandwidth, you're burning power, you're causing pollution.

Host (Ben Horowitz): Yeah, these are all text-based protocols. So one of the design principles was all protocols have to be text-based. There were no binary protocols. Text-based protocols are much less efficient, much, much slower. But the enormous advantage is you can program a text-based protocol by writing text. And you can read it by reading text. Whereas if it's in a binary format, you're always dealing with an intermediary system and it's just harder to develop for and harder to understand.

Guest (Marc Andreessen): Source. There was a little counter to the belief of the computer science world at the time. I mean, everybody agreed on ASM1 and coding. You have to do that.

Host (Ben Horowitz): Yeah, any CS professor of that era who looked at this said they're doing it wrong, 100%. They said they're absolutely doing it wrong because it's not optimized. I mean, so much of computer science at that point was about optimizing scarce resources because that was all you had at that time. And they had spent decades figuring out how to do that. And we just decided, we collectively decided to just break that rule. And again, it was not to break the rule just to break it. It was because what was on the other side of breaking that rule was openness and creativity and empowerment and anybody can do anything. And then-

Guest (Marc Andreessen): Inclusivity, right? You didn't have to be a computer scientist, networking expert to build a web server.

Host (Ben Horowitz): Yeah, that's right. And the experience people had, you know, the killer kind of version of this that ended up working really well was this idea of view source.

Guest (Marc Andreessen): Yeah. So there was this feature built, we built into Mosaic and it was built into browsers, which was, I forget exactly when it popped up, but view source. The idea basically was you're looking at a web page, you're looking at the rendered version of a web page, but you could click on view source and it would show you the HTML source code for the page. And so anytime you wanted to see how a web page had been created to accomplish something that you wanted to do, you could just look at it and then it made it easy to learn and replicate. And again, that was like, that was not, that was like, there was no, there was no view source for like network protocols before that. That was like, that was a new idea.

Host (Ben Horowitz): That created so many web design jobs. It was crazy. like, which could never have come about. It'd be, you know, like that little or a seemingly small thing was a massive thing.

Guest (Marc Andreessen): Yeah, no, look, I meet people, I met people, I met somebody just the other day who, you know, literally it's like that. It's like, you know, they first got, you know, they first got access to the browser and, you know, from their, you know, high school or whatever and college. And then they literally, they literally did the view source thing and they're like, oh, I can write HTML. And then they got a job as a web designer.

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): And then that paid their way through, you know, whatever to, you know, get to the career going or start their company. And yeah, so that was, yeah,

Host (Ben Horowitz): Yeah, economic empowerment, yeah, inclusivity, yeah, maximal, I would say maximal inclusivity, you know, and look, you know, technical whizzes could do more, but like you did not have to be a technical whizz to get started and get going. And then it gave you a very powerful motivation to learn more and a very easy way to learn more.

Guest (Marc Andreessen): Yeah, and so that worked out really well. And then basically the Mosaic idea basically was, okay, pull all this stuff together. So build basically the unified visual interface. And Mosaic out of the gate actually supported all three of the systems that I described. So we supported out of the gate with the web in its sort of text form at the time. And then Gopher and then Waste. We actually also supported FTP. We supported, actually had native support for internet news groups. And so it was sort of a single graphical user interface to rule them all. So we had support for all this. And then the web obviously is the one that took off. And then the other part of it for Mosaic was to then make the web graphical. And so to transition it from a text-based, kind of text prompt DOS kind of situation to be full graphical web pages. And that was, of course, that was then the thing that really kind of just got lit on fire.

Host (Ben Horowitz): Yeah, you famously invented the image tag, if I recall.

Guest (Marc Andreessen): So there was a big dispute. There was a big dispute early on. So there was opposition early on within the internet community, and I won't name names, but within the set of people who were into this kind of thing working on it, there was actually a lot of controversy around the idea of adding images. And there was a big argument. There are actually multiple arguments to not add images. And by the way, to not images means not make it graphical, not bring it into the GUI world. And one argument was just efficiency. Again, network optimization, use of resources. you know not you know and by the way you know sort of an equality argument you know not everybody has a graphical workstation right

Host (Ben Horowitz): Yeah right.

Guest (Marc Andreessen): It would be unfair to them if there's web pages that they can't view so that was part of the argument you know there was certainly a speed performance you know waste argument to it and then there was also there was also a cultural argument and this was around the time that the internet was starting to really open up and you know that that kind of nirvana i was mentioning where everybody is like you know cs degree holder is starting to become a consumer thing early on. And there was a lot of anxiety around that. And so there was an argument at the time that content of the internet should remain only scientific and technical, right? And if you add features and capabilities like images and graphics, then you are encouraging the creation of sort of mass market content, right? And if you have mass market content, that's going to draw more of the wrong kinds of users.

Host (Ben Horowitz): A valid argument.

Guest (Marc Andreessen): Which it turns out that argument was correct. But, well, yes. Well, it was, it was, it was correct. The people who made that argument were correct based on their own presuppositions.

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): I was, I was on the other side of that argument and I was on, I was on the other side of each, each of those arguments. But specifically on that argument is I, I just always thought everybody should be able to use this. I was very much on the side of, this is amazing. Everybody on the planet should be on the internet. Everybody on the internet should be on the web. Everybody on the web should be graphical. Like there should, yes, there should be content all over the internet that's graphical. There should be, you know, all kinds of pictures and movies, animations and streaming and games and like, yes, you should have all this and everybody should be on it and we should maximize this. And that made, and again, heretical idea, that was, there were a lot of people at the time who were very important at the time who were very anxious about that. And then, and basically we just, so there was a big fight argument around that and we weren't making progress on it. And then I just did a, cause I controlled, you know, I controlled Mosaic at that point. So I just did a fait accompli and I just declared it and I created the image tag.

Host (Ben Horowitz): Yeah, and people put up images.

Guest (Marc Andreessen): And you put up images.

Host (Ben Horowitz): Yeah, so I won the, what's the, I won the de facto argument, which is through sheer authoritarian action.

Guest (Marc Andreessen): Yeah, you were king of the internet. And then, well, tell us about like, okay, how did it take off? When did the press recognize it? And then how did you become, go from king of the internet to customer support for the entire internet?

Host (Ben Horowitz): It's the same thing. It's the same thing. Also the main blogger, maybe the first blogger, depending on how you score it. So it was also the sort of, I was the front page for a while. So, yeah, so we started basically, a group of us at NCSA basically kind of went rogue in 1991 and, you know, started kind of working on this idea on nights and weekends. And in particular, my partner at the time, Eric Bina and myself, you know, we were the first two to kind of work on this. And just a full kind of acknowledgement here, Eric and I co-wrote the first version of Mosaic, which was for Unix workstations. and then we had other colleagues who are very famous in the history who developed the Windows version and developed the Mac version. And so we did develop for all three of those platforms. But the first version was the Unix version. Eric and I built it. I always credit it as I did the front end and then Eric did all the hard work.

Guest (Marc Andreessen): Yeah, yeah, he was a great programmer, I guess, yeah.

Host (Ben Horowitz): He was fantastic, absolutely outstanding programmer, is an outstanding programmer. And so I did the UI and Eric did the rendering engine. And so Eric built, the rendering engine is the core of it. Like the rendering engine is the thing that actually renders the page and has all the user interface elements and makes the links work and displays the images and all this stuff. And that was definitely the harder half of it from a programming standpoint. So I give Eric at least half the credit, if not more, for that. And then my role was the front end. So it was kind of everything around the rendering engine. And so it was the rest of the UI. And then I did the networking protocols and everything, user preferences, and all the caching and all the things to kind of make the rendering engine work. So it was the two of us. And it was really like the core work was sort of a crash renegade project kind of off books. And for me, it was like off books in two ways. It was not what I was supposed to be doing at work, but it was also I was doing this instead of going to class. This is when I almost got kicked out of college also.

Guest (Marc Andreessen): Yeah, so we sort of a crash course over the course of, I guess, the fourth quarter of 92. And then, you know, we kind of worked really hard over the holiday break of 92 to kind of get it working. And then I forget the exact sequence, but we put out the first kind of acceptable version, which I think was the 0.9 version. like around Christmas or a little bit after 92. And then I think got to quote 1.0 and kind of the spring of 93. And yeah, and basically it went vertical basically out of the gate with the 0.9 version. So it was basically a, yeah. So it's sort of a year of preparatory work in 92. And then it was sort of 93 was the vertical takeoff. And 93 was a very important year for me because it was my senior year in college also. And so, and I was off a semester. I was off cycle by a semester for reasons. And so anyway, so January to December in 93, that was my senior year in college. And so this was my chance to really do that. And then, because I just assumed I was going to graduate and leave and get a job at the end of the year. So I had about a 12-month run there where the thing really took off. And then, yeah, look, it was the tiger. Oh, and then we had other colleagues who did the, what was the, not the first web server, but the first kind of widely used, I would say, robust, scalable web server. And again, that also, again, gave us a lot of ability to move quickly because we actually controlled for that period. We controlled both the client and the server. And so we can move very fast.

Host (Ben Horowitz): Yeah, so the two of them came out and then people started to figure this out and it started to get widely used among existing internet users. And then it was an immediate reason for people to get online at home. And it was really the first reason for a lot of people to try to get online at home. And so it also, I think, helped catalyze the boom in what were called at the time consumer ISPs and for people to upgrade their PCs to be graphical and then have network, you know, have the network stack. And so 93 was like this upward, you know, the straight kind of upward hurricane. But again, it was in this context of we're working for a research institute funded by the federal government. And so, you know, we have, you know, we have no money. We have no revenue. We have no business model. You know. We've got a product out there that's taken off. It's taken off like crazy. And then we put it out as open source, but under what's called a hybrid license. We put it under a hybrid license that says it's all free for academic and individual use and nonprofit use. But if you want to use it for commercial applications, you have to come talk to us. And then I had the mailbox for the incoming commercial queries. And I remember when it hit like 400 of basic companies, you know, like general counsels and procurement officers at big companies saying, you know, we want to deploy this throughout our company. You know, who do we pay? And we had no way to take the money. We didn't even have like a price sheet. We didn't have any of this. And then very critically, we didn't have any support, right? So we didn't have any customer support resources. And so we had the support email address, and I also had that email box. And so in my spare time between coding sessions, I would literally just answer questions. But it was literally, it was supposed to be tech support for Mosaic, but it turned into tech support for the entire internet. So I helped a lot of people.

Guest (Marc Andreessen): You know the difference between that was the internet for everybody, and a lot of ways still is.

Host (Ben Horowitz): Exactly. And so it really started to take, it then sort of became a formal project. It kind of got embraced and became a real project and got more resources. And then, but we were, you know, we were kind of just dying from the overhead and, you know, we needed more servers and we needed more people and the whole thing. And so we wrote, I remember we wrote a proposal to, for a grant, we wanted an incremental grant from the National Science Foundation. And it was to staff a customer support desk so that we could support, so we could hire like, you know, whatever, four people.

Guest (Marc Andreessen): Customer support.

Host (Ben Horowitz): Yeah, yeah. So it was actually my first trip to Washington, D.C. as we issued this grant. We sent it to Dan. And the National Science Foundation people, to their credit, they were fascinated by this whole thing and they were glad that it was working. But, you know, we sent in this grant. It was the only place we had any sense of where to get money. And we sent in this grant. Literally, it came back denied. You know, the National Science Foundation is not in the business of, you know, funding customer support.

Guest (Marc Andreessen): One of the sort of fun twists here is, okay, you idiot. Like, this giant commercial opportunity is staring you in the face. like you've, you, you, you literally have like inbound sales leads, like coming out of your ears. Like, why don't you go raise venture capital and start a company?

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): And of course the answer was because I had no idea that there was such a thing as venture capital.

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): What venture capital? I literally, yeah.

Host (Ben Horowitz): You know what a tractor was. Yes, exactly.

Guest (Marc Andreessen): You know, I had no conception whatsoever for, you know, for, I had no, just no clue at the time that it was actually a tractable thing that you could, you could, you could do that. And that was like something that like, you know, really, you know, rich, famous, fancy people did, or, you know, I don't know, some people got lucky or people in Silicon Valley or something, but like people in Illinois certainly were not doing that. And so, you know, and there was no venture capital in Champaign or Banda at that point. And then there was exactly one software startup at that point called Spyglass, which is a name that will come up later in the story.

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): But it was not doing well. And so it was like, it was more of a cautionary tale. And so we basically just, yeah, we had the tiger by the tail and then we just kind of held on, held on for dear life kind of through that year.

Host (Ben Horowitz): Incredible. And then when did the kind of media and the press skip wind of it?

Guest (Marc Andreessen): So, this is a funny story. So, there was a Wall Street Journal reporter named Jared Sandberg.

Host (Ben Horowitz): Yeah.

Guest (Marc Andreessen): Well, yeah, yeah. I remember Jared Sandberg. He was a funny guy. He was a great guy. He was a great guy. You know, this is 30 years ago. So, we're using the was here for people who are still in perfectly good health.

Host (Ben Horowitz): Yeah, yeah, yeah.

Guest (Marc Andreessen): He was a great character. Yeah, yeah. And it was and it was a great writer, great reporter. And so, he had sort of the tech beat at the Wall Street Journal at the time. And he figured this out early and called me up, the blue, probably in, I don't know, late 92, early 93, like super early. And he had the story ready to go. And he was intensely frustrated because he could not get his editors to run it because it was not an important story.

Host (Ben Horowitz): Yeah, of course not. What are you talking about? It's just a bunch of idiot kids with bad grades. Can't do writing software.

Guest (Marc Andreessen): Yeah, yeah, yeah. On this internet thing that nobody cares about, that's never going to be a thing, that's about to get squashed by MSN and AOL and interactive television. And I'll tell you, again, talk about the heresy aspect of this. Like this is, you know, this is why I'm so distrustful of experts. Like this is my origin story for why I don't trust anybody with credentials or anything anymore. So I remember another one. Like, so I was working on, I remember working, it was, it must've been, it was like December of 92 or January 93. So I'm like working around the clock in my little office at NCSA. And, you know, this is at Urbana-Champaign is also, you know, it's a little further south where I grew up, but it's still frozen tundra most of the year. And the wind comes whipping over the Illinois plains and everything is frozen. And, you know, you slip and fall every 10 feet, you know, in the middle of winter. and just this kind of crazy thing. And I'd be working in the middle of the night. None of the restaurants are open and I'd get hungry. And so I'd walk down to the one convenience store that was open 24 hours and I'd buy my whatever, my hot dog or my cookies and something to drink. And I remember walking in one night and there was this new magazine on the newsstand called Wired. And it was Wired issue number one. And I was like, oh, that's interesting. It's a magazine about, it appears to be a magazine that's about things that I'm interested in, which was a novel concept at the time. And so I bought it and I took it back to my office and I read it cover to cover and they did not have the word internet in it once. And I was like, okay. I was like, okay, I guess I don't count. We don't count. This whole thing doesn't count. These are the experts. They have a magazine. They're pros.

Host (Ben Horowitz): Yeah, these are the pros.

Guest (Marc Andreessen): And what we're working on is clearly not important enough to merit the magazine. And so I was just like, okay, I guess we're just going to keep working on our thing. It's sort of like this constant message from the media, which is like, this is not important, this is not important, this is not important. So anyway, so Jared had this whole story, he was ready to go, and he could not get his editors to run it. And then later in 93, John Markoff at the New York Times figured this out, and he was...

Host (Ben Horowitz): Oh, a good tech reporter.

Guest (Marc Andreessen): I mean, John Markoff's a smart guy, yeah.

Host (Ben Horowitz): Legendary, yeah, legendary, legendary, you know, tech reporter going back, you know, quite a bit, you know, still very active, but at the time he was like a veteran and was very well respected in the industry and wrote a bunch of good books and so forth.

Guest (Marc Andreessen): And so he wrote a story for the New York Times. So the good news is he wrote a story. The bad news is it featured my boss and my boss's boss. Who had nothing to do with the project? Who had you work on something entirely else? My boss at the time, to his credit, you know, he was aware and he, you know, he kind of, you know, he didn't vigorously oppose us. And then he supported us. And then he ultimately sort of adopted as a project. And so he kind of eased his way into it. But he ultimately was very supportive. And then, but my boss's boss was the director of NCSA at the time. And look, this is one of those things where I like I owe these guys a tremendous amount because they created this environment that I was able to do my work in. And I wouldn't be here today if not for them. But, you know, you know, I kind of I kind of a little bit of an iffy relationship with my boss to start with. And then I had never met my boss's boss. Right. And there was no reason for me to meet my boss's boss because I'm like an undergraduate like staff member. Like, you know, he's a big like a huge he's like a huge important researcher, astrophysicist and, you know, directing this huge supercomputing center. So he had no reason to meet me. But, you know, the story shows of the New York Times, it's smiling photos of the two of them and not, you know, Eric and me. And I'm like, oh, OK, I see how this works.

Host (Ben Horowitz): And how did you feel at the time? Because I know how you would feel now if that had happened. Like, how did you feel then?

Guest (Marc Andreessen): It was just a little bit. I don't know. It was a little bit. It was a little bit annoying, but it was also a little bit of like, look, it's the New York Times. Like, I don't know. You know, I don't I guess they write about important people. Right. And these are important people because they're running this thing. and I'm not an important person because I'm only writing code.

Host (Ben Horowitz): Yeah, yeah.

Guest (Marc Andreessen): And, you know, it's like the project is important enough where they can write about the project, but like the people actually writing the code are not important enough to talk about. And so it's like, it's kind of the same reaction I had to Wired Magazine, which is like, it's a little bit annoying, but I guess it means I just need to go back to work because there's nothing else to be, it's not like I can call the, it would never occur to me to call the editor of the New York Times and complain. You can't screen her out of the game. Nor would he have taken my call, you know? And so he certainly didn't know who I was because it certainly wasn't in the story. so so it was you know it was just like it was like whatever but it's like this weird you know it's like the plato shadows on a cave wall thing just like okay there's this there's all these things that people believe right including up to the internet is like doomed and it's never going to be a thing and it's going to get swamped and and all these things and then we're just like okay no we just have this stuff in front of us and it just plainly works and we believe in it and i guess we're just going to keep working on it and maybe people will figure out maybe they won't anyway so jerek sandberg calls me up i remember the day that story broke in the new york times jerek Sandberg calls me up and he's just absolutely livid. And he told me that he, he, he got the morning paper and, you know, this is like six months later or something after he, you know, he, he would have had the scoop. He would have had the first story on basically on all of this stuff. And he, he literally told me the story. He said, he charged into his office that morning and slapped the New York times down at the desk in one of these dramatic moments and said, you know, see, I told you so, you know, John Markoff front page of the business section. I told you, this was actually a story. And his boss was like, Oh yeah, I guess, I guess you were right.

Host (Ben Horowitz): Oh my God. That's so crazy.

Guest (Marc Andreessen): so that's how again the heresy like that's how heretical it was it was like actually it was actually hard to get it in print and i think the only reason it showed up in the new times is john markoff was such a legend he was just a legendary he was further along in his career at that point so 93 was a phenomenal you know phenomenal year because it was the takeoff year for the it was a takeoff year for the web for the browser for for all for all this stuff for mosaic yeah and so it was and then you know the other part of it was just the ping pong effect was very interesting which was, it basically was, you know, it started out with like a few people with browsers and then a few people with web servers putting up, you know, these little individual pieces of content. But then you got in this kind of feedback loop back and forth, kind of ping pong thing where basically every time there was a new compelling piece of content put online, there was a reason for people to start using the browser. And then every time more people started using the browser, there was an additional incentive to put more content online, right? And so it was like the writers, you needed writers and readers. And so more writers meant more readers, more readers meant more writers. And so, and this is what our partner, Andrew Chen, refers to as the cold start problem, right?

Guest (Marc Andreessen): Is by default, if you have a situation like that where you're going to have a network, you know, at scale, it's going to be great. But to actually get it to scale is really hard. And most things that need to have that kind of network effect never get past the cold start problem. They just strangle early because there's not enough writers, which means there's not enough readers. There's not enough readers, there's not enough writers.

Host (Ben Horowitz): But this one completely was a winner for you.

Guest (Marc Andreessen): That's right. And so what happened from the very beginning was people all over the world started creating web servers and started putting up content. They either used NCSA Mosaic, the NCSA web server, which actually later became Apache, so it's still kind of in use today in derivations. Or they literally could write their own web servers for the reasons we described earlier. And so there was new content arriving online. Just to give people a sense of this: in early '93, it was like one new website a day. And by one new website, I don't mean a new eBay a day. I mean a new web page.

Host (Ben Horowitz): I remember it was a big day when the first restaurant menu came online. It was an Indian restaurant in some second-tier city in England. Somebody just put the menu online.

Guest (Marc Andreessen): And then I remember the first webcam. The first streaming video, first webcam was a coffee pot. A guy had rigged up an early webcam with a coffee pot down the hall in some computer science department so he could see when the coffee pot was empty and go refill it. He put the coffee pot online. It was the first webcam. We all sat for like a week and just watched the coffee pot.

Host (Ben Horowitz): Why put it on the LAN when you can put it on the internet?

Guest (Marc Andreessen): Exactly. You could see it from home. So we all watched the communal coffee pot. One of the features built into Mosaic was what we called the "What's New" page. That was something I had at the time, sort of one of the first blogs, maybe the first one. Every day it was, "Here are the new webpages for that day." It started out being one new page a day. It was a big deal when it started to be two a day, then four a day, then six, then eight. By the end of the year, I couldn't keep up anymore.

Host (Ben Horowitz): Then we had the "What's Cool" page. The editorial function.

Guest (Marc Andreessen): Exactly. Getting to the Windows and Mac versions was a key moment. Those guys did a great job, and that really opened things up. This was leading up to AOL formally interconnecting into the internet in September of '93. It started to become clear that they were going to interconnect to the internet and bring their users out. Then it seemed inevitable they would build in a web browser, so you started to get momentum from that. This also coincided with the NSF handing off the NSFNet to the commercial telecom companies. All this activity from Mosaic was crushing the backbone bandwidth. The NSF wasn't in the business of providing commercial backbone services, so they did a handoff to the three big telcos. The acceptable use policy was revoked, and commercial use of the internet became legal.

Immediately after that happened, there was a really pivotal moment. There was a big computer company at the time called DEC. They had a research lab in Palo Alto called the DEC Western Research Lab. There was a guy there, I think it was Brian Reed, a computer science guy who was into this stuff. There was a cult science fiction bookstore, a tiny hole-in-the-wall on El Camino Real in Mountain View called Future Fantasy Books. It had local clientele but also a lot of Japanese and German tourists who would buy science fiction novels they couldn't get back home. This guy at DEC went and talked to the very hippie, ponytail, old-school owner and said, "Let's put your bookstore on the internet." The guy had no idea what any of those words meant. Brian described creating a website, putting a catalog online, and letting all those foreign buyers see the catalog on the web and click to buy, which was a new idea. The guy said, "That's great. The problem is I don't actually own a computer." Brian asked what he had, and he said, "I do have a fax machine." Brian said, "Let me get back to you." The DEC guys literally created the Future Fantasy website, figured out how to digitize his inventory, and created the first e-commerce site for books. Original Amazon. They set up a fax gateway where you would order on the web, it would fire off a fax to the guy's store, and he would ship you the book.

Host (Ben Horowitz): And then his business doubled overnight.

Guest (Marc Andreessen): It was the best thing that ever happened. Then of course Jeff started Amazon and destroyed him. But there was a year there where he had the best year of his life, shipping books all over the world. That may have been the first formal e-commerce thing.

Host (Ben Horowitz): Wow. So then you graduate. What do you do?

Guest (Marc Andreessen): I was an undergrad staff member getting paid $6.25 an hour, capped at 30 hours a week. It was fine. I was having a great time. I assumed I'd graduate with a computer science degree and go get a job. It was a schizophrenic experience because all this stuff was happening all day long, this tremendous cascade of incoming activity. But there was no money in it, no funding, no venture, no startups, no business. The media was telling me it was stupid. I was halfway between seeing the future in front of me and the rest of the world not taking it seriously. Maybe I was just smoking my own exhaust, and maybe it would all get crushed next year by MSN. The assumption was you left this stuff behind when you graduated. I talked to the NCSA guys about staying there. They offered me a job, but it would have been a staff programming job staying in Urbana. I wanted to get to a coast. I was ambivalent as to which coast, but I definitely wanted to get to one.

I had control of the "About" page for the web browser. I added a line saying, "One of the primary authors of this browser is graduating and is available to be hired. Please send job offers to this mailbox." I got about a dozen job offers, some on the East Coast, some on the West Coast. I got an offer from a little software company out in California. I also got an offer from Sun at the time, from a unit called First Person, which was creating what became Java later with James Gosling. I almost went there, but they had a "phantom stock option program."

Host (Ben Horowitz): That doesn't sound good. If you're applying the word "phantom" to your stock option program, that's not a good sign.

Guest (Marc Andreessen): They had a classic problem. They had a software group they wanted to incentivize, but they were a hardware company with small stock options. They wanted this thing to be a separate research project but didn't want to spin it off. They were creating a shadow stock option program. I didn't understand it. It sounded like a scam. So I turned that down and went to a little software company called EIT. I moved out to California in January of '94.

I went to work for EIT, a little contract research organization with very smart CS people in Palo Alto doing work for the government and companies. To their credit, they were on the internet idea early and wanted to create internet software products, which was a very new idea. They made me an offer, flew me out, and I moved. I thought, "Okay, I'll work on some internet stuff and see if there's a way to make money on internet software." I kept having this out-of-body experience. The internet started to get serious media coverage. Books started to show up. O'Reilly Publishing became famous. At the peak in '94, '95, you'd have walls of books about the internet in bookstores: guides to the internet, how to use it, how to write web pages. There would often be a floppy disk in the back with the TCP/IP stack or Mosaic. People started to figure this out. The press started to take it seriously. There started to be more interesting content. The thing was going. But I had left Mosaic behind, so I didn't have the email addresses anymore. I knew how much commercial demand there was, but it was still this schizophrenic thing. I had never heard of venture capital, so I didn't really know you could start a company.

Then I got a call from Jim Clark, the co-founder of our company, Netscape. He was a legendary figure. He had been a co-founder of Silicon Graphics, one of the leading tech companies of the era. He got sideways with the CEO and decided to leave and start his second company. But he had a non-solicit agreement with SGI, so he couldn't hire the amazing team he had put together there. He was sniffing around for talent. A guy who worked for him, Bill Foss, who later joined Netscape, mentioned to Jim, "One of the guys who made this Mosaic browser just moved to Silicon Valley. The 'About' page says he's available. Maybe you should go talk to him." Jim called me up and we had breakfast. It was a traumatic experience because I wasn't eating breakfast in those days. I was on programmer hours. I had to be up at seven in the morning to meet Jim at El Fornaio in Palo Alto at 7 a.m. on a Sunday. I had to recalibrate my entire sleeping schedule. I was still bleary-eyed when I got there, but they had good coffee.

Jim and I decided to start a company. But it was still this weird thing. Jim knew all about the internet and the browser, and I was still watching everything. But the overwhelming assumption in the industry was that this was not a serious thing. It was a momentary thing that would go away. The big companies would take over. The internet was not going to be a commercial medium. Our first two business plans were not this. Our first plan was to do interactive software for interactive television, to build software to replace the guy on the roller skates. Jim's company had been one of the main companies building those systems. We modeled out how interactive TV would cost and work and concluded it wasn't going to work. It was too expensive and the technology wasn't ready. Plan number two: Jim had a good relationship with the CEO of Nintendo because SGI had done a deal to do the first 3D graphics chip for a game console, the Nintendo 64. He went to visit Yamauchi-san, who ran Nintendo, and struck a handshake deal to build the online service for Nintendo gaming machines. To build what today you'd call Xbox Live or the PlayStation Network, but in 1994. That was also too early. It would have been all dial-up with 14 kilobit modems and low latency. Nintendo had actually had an online service on earlier devices in Japan with early e-commerce and food delivery in the '80s. It didn't quite take. We modeled the whole thing out with modem banks and figured out it couldn't quite work.

We took a walk, feeling discouraged. We had these two ideas that didn't pan out. Was there still anything to do together? I said on the walk, "This internet thing keeps going. Think about what we just experienced. The press and all the experts were telling us it was either going to be the big companies doing interactive television or home video game-like services. Those were going to replace the internet when this stuff got serious for consumers."

Host (Ben Horowitz): When daddy comes home, you little kids can go play in your room.

Guest (Marc Andreessen): Exactly. If you look at the magazines from that era, the way people got their news was reading one of the three big news magazines or newspapers. That's where all the experts showed up. The magazine covers were all these big company CEOs pouring scorn on the internet, declaring it a joke and a toy. Their proprietary thing was going to be so much better. We had this moment where we proved to ourselves that interactive TV wasn't going to work and you couldn't build this on video game boxes. By process of elimination, it had to be the internet. It was the only thing that works. Yes, it has every issue people complain about: slow, inefficient, insecure, no business on it, hard to get online. There were all these reasons to believe it wouldn't work, but we knocked out all the other ideas and said, "This thing has to be the thing. It's the only thing that works." Jim is a total whiz on these things. We sat down and said, "What if we did this incredibly heretical idea?" As late as April '94, it was still very heretical: "How about we build a software company to make internet software?" That seemed like a risky, crazy idea. In retrospect, it was the most obvious idea of all time. But that's the true story of how we got to that idea. We ended up building the commercial version of everything we had built at Illinois. We got there the hard way. It was not the obvious idea.

Host (Ben Horowitz): Counterprogramming conventional wisdom and media advice still works now. It's amazing.

Guest (Marc Andreessen): I live in this constant state of out-of-body amazement. These people show up on TV or in the papers with all these credentials, degrees, initials after their names, incredible resumes, publication credits, government grants, Harvard and MIT. And they just say shit. Maybe they're right, but if they're wrong, and they're wrong a lot of the time, what consequences do they bear? None at all. Zero. They're back on tomorrow with some new line of bullshit. There's this great book by Phil Tetlock called *Expert Political Judgment*. He did a comprehensive study of political predictions from experts in the media. The conclusion is that the average well-credentialed expert on any topic involving politics or global affairs is somewhat less than random likely to be correct. The credentialed experts score about 40%, and a monkey flinging shit at a dartboard is 50%. He points out there are no repercussions for being wrong. No career damage, no economic damage. If they were being epistemically honest, when a talking head is on TV, there would be a scoreboard showing their last 20 predictions with red or green for right or wrong. That scoreboard never appears. He wrote that book 20 years ago. Everybody read it and said, "Yeah, that's right." Then everybody completely ignored it and kept doing things the same way.

I had this experience where I still feel dumb. I read all this stuff. I believed it all the time. It caused me to be insecure about the thing I was actually doing that I saw was actually working. I knew it was working, and I knew why it would keep working. Even still, it was this wall of doubt and skepticism eating away at me.

Host (Ben Horowitz): The good news is we've used that to our advantage many times since then. I remember when we made the Coinbase investment. Bitcoin was some number of hundreds of dollars. Everybody, economists, everybody was writing that it was a complete scam and total bullshit and never be worth anything. Here it is at $63,000, $64,000. All you had to do was listen to the experts and do the opposite.

Guest (Marc Andreessen): It's amazing. But it's hard. I have 30 years of evidence that this is the case, and even still, it's a real effort of will to remember they don't actually know what they're talking about. They don't have any predictive capability. They're in a system with absolutely terrible incentives.

Host (Ben Horowitz): If somebody's super dismissive about something, that's a great thing to study. It's almost surely not a dead zero. It may not work, but there's no way it's as bad as they're saying. If somebody says something is a scam or a Ponzi scheme or bullshit, that's almost always worth looking into. So you started Netscape with Jim Clark. It was actually called Mosaic at the time, right?

Guest (Marc Andreessen): Mosaic, yeah. I had graduated. Our other colleagues at Illinois were either students or staff members working under federal research funding on a specifically non-commercial, open-source project. The university didn't have any commercial value, at least according to what everybody thought at the time. We start Netscape. Jim and I flew back out to Urbana in the middle of an incredible snowstorm and hired all but one of the original Mosaic team members to join us at Netscape. Most of them moved out to California and were core members of what followed. We booted up the company and named it Mosaic. That was the first issue that got us in trouble. The reason we named it Mosaic was not because we planned to literally offer Mosaic as a product. We very specifically decided we were going to leave the source code behind. You wrote it when you were skipping class. Really?

Guest (Marc Andreessen): Yes. We knew it was not commercial grade. We knew it had issues. Performance issues. When you create something like this, you know all the issues. We knew we needed to start from scratch and build the correct commercial product. We knew what we needed to do to do that. We very specifically did not bring the source code with us.

But all the HTML and HTTP standards were open standards, free for anyone on the internet to use. So we thought, okay, we have no copyright issues here. We're not taking the code. On the trademark side, there's a long history in Silicon Valley of companies named after the projects that spawned them, often from a university setting.

Two famous examples: Sun Microsystems, which was named after the Stanford University Network project. And Oracle was the code name of a project Larry Ellison had done for the government in the 1970s. I thought at the time it was a time-honored tradition to use the name as an homage. Illinois didn't have a commercial interest in this, so we'd just do it.

In retrospect, that was the crack. We introduced a crack in the armor from the very beginning.

As I was leaving for California, the other people at Illinois started to figure out there might be money in it. The founders of Spyglass approached the University of Illinois administration without my knowledge. They struck a deal to license the Mosaic software code we had written for commercial sales. They started offering a commercial product called Spyglass Mosaic, totally within the university's rights.

Then we announced Netscape. We were the team that had written all that code. The press started taking it seriously, we became famous, and Jim Clark was a legend. Spyglass got worried we were going to snuff them out, so they enlisted the University of Illinois administration to try to kill us.

They didn't sue us because they didn't have a good claim. Instead, whenever we were competing with Spyglass for a sale, the University of Illinois administrators would call the customer and tell them they were going to sue us.

Host (Ben Horowitz): Good Lord.

Guest (Marc Andreessen): That's the dirtiest thing you can do to a startup. Super nefarious. Any big company is already worried about doing business with a startup. If the startup's about to get sued, why take the risk? Our whole sales pipeline froze up. We were running on venture capital, and money was getting tight. VC wasn't what it is today. We didn't have that long of a runway. We needed revenue. This became a big problem.

We got everybody together and talked about it. We decided, and I'm very proud of this decision, to preemptively sue the University of Illinois. We sued them for torturous interference. It's not great law, doesn't often get enforced, but at least it's on the books. We sued Illinois.

Furious negotiation followed. We offered them $4 million worth of stock when the company was worth maybe $20 million. For those of you who are youngsters, it eventually sold for $10 billion. So you can do the math. It would have been a lot of money. They turned it down. Instead, they demanded cash. We didn't have cash. But at that point, it was starting to work. We raised money from Kleiner Perkins, had other investor interest, sales were starting to come in. So we paid them the cash and settled.

Host (Ben Horowitz): That decision on their part cost them at least a billion dollars in direct stock, plus all the downstream philanthropy from you, from Jim Clark, from all the other founding engineers. They lost a few buildings. They lost a campus. Probably $3 billion in 1990s dollars.

Guest (Marc Andreessen): Now, I gave the disclaimer up front. In an alternate universe with a different skill set, I would have dealt with this differently. But as I actually existed on Earth-1, I never met the University of Illinois administrators. I didn't know the president of the university. I was a random undergrad.

Host (Ben Horowitz): You could have called him.

Guest (Marc Andreessen): That's true. But the calculus they made, harassing their own students and siding with somebody who wrote them a check, is still a little on the evil side. I thought it was really bad. I was really upset for a long time. To be totally honest, I'm still upset. Every subsequent administration has attempted to reach out and repair the bridge. I have not returned the calls.

Host (Ben Horowitz): You're a grudge holder. You and my wife are the biggest grudge holders I know.

Guest (Marc Andreessen): I greatly value my grudges. They're very important to me.

The broader point you brought up is really key. There are a small number of universities in the world that really understand this. Stanford's great genius over the last 50 years is understanding that it's actually really good when this kind of thing happens. If your students or faculty go off and do something new and are successful, the money you'll get back in philanthropy is orders of magnitude higher than whatever technology licensing fee you could extract.

Host (Ben Horowitz): It goes back to the original Mosaic story. It seems like you're giving a lot up by being open, but you're actually opening the whole world to you. It's a great metaphor for life. If you live in abundance, you will get abundance. If you live in scarcity, you'll screw yourself. It's good that those universities are abundant. It's good that you were abundant with the Internet. That's how we live in the world we are in today.

I often think about the alternate universe where you don't write Mosaic, it doesn't work, you don't start Netscape. Most systems would be proprietary. Smartphones, everybody pays a tax to download their app from the app store because Apple owns it. Google is an extremely powerful company because if you want to search something, that's where you have to go. It's such an amazing anomaly that the Internet happened, that anybody could join, put up a website, build a great business. Including Google, including some of the big tech today, all created because of the openness of the Internet. Thank you for that. And thank you for the conversation. What a good story.

Guest (Marc Andreessen): Oh, I'm not done yet.

Host (Ben Horowitz): Oh, we got one more. Never mind. Pause.

Guest (Marc Andreessen): I got more. I got to bring it. I got a great climax to the whole thing.

Host (Ben Horowitz): Okay, okay. Climax.

Guest (Marc Andreessen): I'm glad you just went through what you just went through. I totally agree. This has a lot to do with the debate raging around AI right now. The big companies in those days had every reason. A lot of the story the big companies told about the internet early on was that it's unsafe. And it was. But it took a long time for e-commerce to take off because people were worried their credit card would get stolen. You had to get over that hump. There was constant fear of cybercrime, spam, abuse.

The big companies always had reasons why they needed total control. The government needed to protect them. They needed regulations. The world needed to be a world of proprietary systems, not open systems like the internet. A lot of the way the world works today is proprietary systems. The banking system is not open. If the bank decides to debank you, they debank you. Most of the economy is these big companies with total control.

I think it's a major miracle when you're able to get one of these open systems to work. Years later, you're just like, oh my God, I can't believe we almost had the much worse world where the big companies ran everything. But the pattern keeps repeating, specifically again today with AI. Ironically, some of the companies lobbying hardest for regulatory capture and government cartel status for AI are companies that exist today because the internet was open. They are engaged in a particularly advanced form of hypocrisy and mental gymnastics.

Host (Ben Horowitz): I think the founders are gone. It's been taken over by the other people.

Guest (Marc Andreessen): The other people. Although I still blame the founders.

Anyway, the climax of the story. I just can't resist because I think it's so amazing given what happened, what followed. We settled with Illinois. We got underway. We were shipping products, getting revenue, starting to work. This is around the time you joined us. But we were competing with a bunch of other companies. People figured out the internet was going to be a thing. Spyglass was still out there. They were selling Spyglass Mosaic, our own code, against us.

Host (Ben Horowitz): Yeah. Right?

Guest (Marc Andreessen): It's a little maddening. A little frustrating. They had the legal right to do it, but it's a little frustrating. It was fine. We were competing with them. Price war going on. Then Spyglass gets this call from Microsoft. Microsoft calls Spyglass and says, we want to license Spyglass Mosaic so we can build it into Windows. Spyglass says, great. How much per copy? Microsoft says, don't worry about it. We're going to pay you a flat fee. We're going to sell it as an add-on to Windows. You'll still have Spyglass Mosaic. You can sell on other operating systems.

They struck the deal. I think it was a million dollar one-time payment. Spyglass thought they had struck gold. Massive endorsement. Big blow to us that we didn't get the Microsoft deal. Then Microsoft had the press conference where they announced Internet Explorer, their browser, which was Spyglass Mosaic. Our code relabeled. There's a famous moment where the Microsoft guys are on stage, they do the "one more thing" part. "Oh, by the way, we're going to make it free." There's a famous moment of the Spyglass CEO with his head in his hands in the front row, realizing his business had just ended. He sold out his entire business for a million dollars. That was the end of Spyglass.

Host (Ben Horowitz): All's well that ends well. Happy ending. Not that I'm competitive and not that I don't hold grudges. That's hilarious.

Thank you all for listening to the Mark and Ben Show. We won't tell the Netscape story, which is also a good story, but maybe if you reply in the comments, you want to hear it, we'll consider that one too. Thank you again. We enjoyed it. We hope you did too.