• How Google was able to nurture Gmail within the company in 2004, and why it would be impossible today

• The PR crisis that Gmail caused

• How Paul Buchheit knew Bill Gates was wrong about the internet — as an intern in 1995

• Why it’s culturally impossible for Boeing to compete with SpaceX

• How Gmail succeeded despite not being the first mover in browser-based email

• Why Paul is optimistic about San Francisco’s future

• Paul’s view on the future of AI

-

Mike Solana: The other day you introduced me to the concept of gatekeepers and insurgents, which is such an interesting frame for understanding product — how new products are built, the environment they need to thrive, how they fail. Can you break that framework down in a little more detail?

Sure. So, first of all, it's important to step back one level and just say, really it's incentives, right? Incentives are everything. You can predict how an organization behaves over the long term just based on the incentives of the people there. But the larger part of that is not just the incentives, but that the long-term effects of those incentives form a kind of natural selection: who is successful in an organization, who gets promoted, and who leaves or gets marginalized. That's the part that gets overlooked a lot, because people are like, ‘How do we bring innovation into this organization that hasn't innovated in a very long time?’ And I think it's essentially almost impossible. I was trying to think of examples of where this has happened, and the only one that jumped to mind was Apple, where Steve Jobs came back and just fired almost everyone. The only way you can take an organization that has forgotten how to innovate, and make it innovate, is that you kind of have to fire everyone—

Paul Buchheit: Maybe a little bit with Twitter too.

Sure.

You can say you don't like them, but Twitter — or X, rather — is certainly shipping new products.

Exactly. And again, it's the same thing where he came in and fired like 80 percent of the people. It's fun to think about just like, what are the incentives of the person in their organization?

I’m a big fan of all the Elon stuff, but SpaceX in particular is really interesting because he came into a very mature space that's been around for a long time. And he's not like a rocket scientist or whatever, he’s some dot com guy, he made his money on internet stuff. And suddenly he's going to build rockets or whatever. And how is that going to compete with a company like Boeing that has this long history and, you know, a very mature organization?

But if you think about the incentives of someone inside of Boeing who works on rockets, it's basically like, don't take risks, right? Don't do anything that's going to cause embarrassment. I mean, in particular, they've been operating in a cost-plus kind of environment for a long time. Think about if you're an engineer at Boeing, you've been there for 30 years or whatever. And you get this idea like, ‘Oh, we could do this thing, instead of using rocket parts, we could use automotive parts or whatever,’ and it would dramatically cut the costs; that idea would probably never even make it to your conscious awareness, and if it did, you wouldn't say it out loud. And if you tried to raise the idea, it'd show up on your performance review next time. ‘Paul lacks the maturity to know we don't take those kinds of risks on important projects.’ And if you think about the incentives, why would you take the risk in a cost-plus environment? The gamble might cause a big embarrassment or scandal. And then if you are successful, all you've done is reduce the cost of the rocket, which means reducing your revenue because it's a cost-plus environment.

So if you think about a company or an organization that's been like that for decades, they have no capacity to innovate because all the people who think like, ‘Hey, I want to do this crazy risky thing’ have either learned to suppress that kind of thinking, or they've gotten pushed out of the organization. The kind of person that does get ahead in that organization is someone who's very good at more organizational things. Like, ‘Oh, I can think of all these risks and I'm going to grow my team to address them.’

So in a lot of these mature organizations, like modern Google, how do people get ahead? Well, you grow head count. And that's not at all what you would want in an innovative organization, like in a startup or anything like that. You want to work with the smallest team size, right? You want a small number of extremely effective people. But in a big mature organization, you want to grow your team size because that shows that you're a ‘leader.’ That's how you move up the hierarchy.

Right. ‘How many people are you managing’ versus ‘what have you done?’

Exactly.

Just before Elon took Twitter, right after he joined the board, he got in an argument over text with the former CEO, Parag Agarwal. Parag was criticizing Elon for his tweets, in like a paragraph-long text, which he said were bad for company morale, and Elon’s only response was ‘What did you get done this week?’

Yeah. That's a great example of the ‘gatekeeper versus insurgent.’ Parag has kind of grown up in that organization. His job was just to keep the wheels on. He’s just trying to keep things from blowing up. Then along comes Elon, who's throwing bombs. Right? He's one hundred percent insurgent. Like, he just shows up and sets things on fire. The first thing he did when he took over was set everything on fire. Then he says, “Alright, guys, we better move fast or the company's fucked.”

Let's break the frame down in concrete terms. First, in the context of a company, who is a gatekeeper and who is an insurgent?

If you just look at incentives in a company like Google, when you have this massively profitable — probably one of the greatest businesses that has ever existed, a goldmine — the incentives are basically ‘don't blow up the goldmine.’ It’s not unreasonable to be concerned. It's actually a rational thing, especially as you get regulators and everything else. So if you think of the incentives — let's say you're going to launch a new product and there's some risk that it’s going to cause PR blowback — a gatekeeper is someone who inserts themselves into the process and says, “Hey, wait, no, we need to do more reviews, we need to run this past four different people.” And I don't know if they still have it, but at the time I left Google, there was this thing called the launch checklist.

Oh my God.

You had to run through this checklist before every product launch. Naturally, the checklist only grew, because it was an opportunity for some ambitious person, within the bureaucracy, to insert themselves into the launch process. Now you need to get a sign off from marketing, or comms, or some other group. Do you remember widgets?

Oh, I do remember widgets, yes!

Almost 20 years ago, widgets were this super hot thing. So iGoogle [the widgets team] was a kind of growing thing. And one of the things they did was actually put their own product into the launch checklist. So you weren't allowed to launch anything at Google unless you already had your iGoogle widget. And you would see this throughout Google — the way people would move ahead is by basically getting in the way.

How does someone move up in an organization? You need to somehow gain power. Obviously we see this in government too. In California, all we have is gatekeepers, right? You can't build a bus lane or a bathroom or something like that because there's a long line of gatekeepers and the gatekeepers extract some sort of power, whether it's money or influence or something. Somehow their authority and their value derives from the fact that they're getting in the way. And the pitch is, of course, they're preventing bad things from happening, which is potentially true because if you never launched the product, then you won't have the PR blowback, right? This is part of the reason Google never launched AI — and they still wouldn't have if OpenAI hadn't launched. If chatGPT didn't happen, Google never would have launched any of the AI stuff if they possibly could have avoided it.

AI is such an interesting thing for Google too because it’s this very mature company. They've been sitting on a gold mine for many many years — search. They didn’t really have to innovate, or win any other fights. But artificial intelligence constitutes potentially the first real threat to that gold mine. It’s looking like they have to enter that space, and they have to win.

AI is this inherently unpredictable thing. Because what are the incentives of Google’s CEO at this point? You have this monopoly and you don't want it to get fucked up by regulators. I think a lot of their strategy has been around, like, ‘Don't trigger the regulators.’ And launching some kind of AI thing that tells people to eat rocks… that creates negative attention. The only reason they're able to get away with it now is that they're not the one who really pushed it out. They're just following OpenAI. So now they're kind of being forced to, but if they had actually been the first ones out, they would have taken tons of abuse. If they had actually been innovators in AI, they would have gotten tons of blowback from regulators and press and everyone else.

You think more so than OpenAI?

Yes, because OpenAI is this startup, essentially. It was just this weird little thing no one outside of the Bay Area had ever even heard of before it launched. No one had heard of OpenAI. Everyone knows who Google is. Google is like a big, scary monopoly. And so it’s not irrational that regulators would care more about what Google does than what some startup does. That’s why I think it's important to think about incentives, because usually everyone is being sort of rational from within their own frame.

It’s the same problem with communism, like: ‘Oh yeah, we're just going to have everyone do the right thing just because it's the right thing to do,’ but that's not how humans work. The reason capitalism works is we just sort of assume everyone does what's in their best interest, and that's a pretty good approximation of what actually happens. If you just assume, ‘Oh yeah, everyone's going to do the right thing just because of their communal spirit of goodness or whatever,’ all you're doing is selecting for dishonest psychopaths. The nonprofit world is the worst here because it's like 90 percent psychopaths who are good at presenting themselves as people who really care about the world, but again, because you've decoupled actually delivering value from claiming the value, the incentives just favor dishonesty.

What about the insurgents, who are they?

Insurgents are essentially people who want to disrupt the status quo. They want to create something new. Steve Jobs is a great example of the insurgent. He hated IBM, right? They were like Goliath. Think back to his commercial, 1984, when they launched the Mac. Like that's one hundred percent insurgency. Literally, an insurgent runs up and smashes the screen. Steve always had that in him. Even as Apple grew into this massive company. I think his personality was that of the insurgent who wants to disrupt things.

And obviously Elon is another extreme example of someone who thinks the status quo basically needs to be blown up.The Boeing Starliner finally launched yesterday. And of course, after so many failures, it barely made it to the space station. But if it wasn’t for SpaceX? The story would just be, ‘Well, space is very hard, experts agree…’ you know, ‘of course it just inherently costs billions of dollars to put a person in space.’ And they would put all their energy into rationalizing why a thing can't be done.

You know, experts studied rocket reusability. It wasn’t like no one ever thought to reuse a rocket. Experts studied it, and they concluded it didn't make sense. Like ULA [United Launch Alliance] had built this whole model that said it would never pay for itself. If it weren't for SpaceX, they would still just be making up all these excuses, and experts would say, ‘Space is just fundamentally hard, this can't be done. It's not that Boeing sucks, it's just that space is really hard.’

But now we've got SpaceX who’s doing this stuff all the time, and it disrupts that story because the story is just obviously not true, because SpaceX is reusing its rockets all the time.

To me, part of the joy of insurgency is just blowing shit up. The gatekeepers provide some value, but in the long term, they just become parasitic.

Well, and their population grows, right? There are never less gatekeepers. There are only ever more and more in any institution. If you're fortunate enough to work for a founder-CEO, he’ll beat them back for as long as he can, but he’ll sort of always lose, slowly, little by little. Is that your sense of it?

Yeah, absolutely. I was just thinking about this. Like, what do you do if you're an insurgent? Basically you either have to start a company, or you have to go work for an insurgent. Because if you're a gatekeeper, the insurgents are just problematic, right? You want to stop bad things from happening. In a gatekeeper organization, if they have insurgents, they put them in a separate building far away, and they call it like an innovation center or something. Containment. At one point, Google had this thing called Area 120. Every big company that sucks creates an ‘innovation center,’ where they isolate the insurgents in a corner somewhere they won't cause harm.

An insurgent is never going to be successful working for gatekeepers, because from the gatekeeper perspective, the person just lacks maturity. They're doing these things that are obviously risky. That's the nature of throwing bombs: you don't exactly know what's going to happen.

Like Elon setting Twitter on fire. The thing I love about watching the Elon stuff is like, the guy has such an incredible track record, but you go on Techmeme or whatever, when there's one of these stories and they cluster all [the reactions], especially ones from Threads and Bluesky. It's just full of people who are like, ‘Oh my God, I can't believe Elon is such a moron, how do people think he's smart?’ Right? Or — especially when he first took over Twitter — a lot of the people who'd been active early on said, “He’s such a moron because he's completely ignoring all the lessons we've learned over the last 15 years” or whatever.

From the gatekeeper’s perspective, if someone has made a mistake in the past, that means you never have to make that mistake again, right? You learn from the past, and it's like, ‘Okay, I made that mistake once and we put a process in place to stop that mistake from ever happening again.’ And what that really means is you're sort of setting yourself into cement. An insurgent essentially has to be willing to make a lot of mistakes again, to be able to say, ‘I know this is going to be a problem, but I'm going to re-encounter the problem, and maybe I'm going to solve it in a new way,’ versus a gatekeeper who's just like, ‘Well, this problem could have been avoided because we know there's a reason why tweets can't be longer than 280 characters, we've spent years discussing it.’

You can only imagine how much scar tissue is built up around preventing anything from ever changing when you’ve spent years and years preventing anything from ever changing. So inherently an insurgent just comes in and blows it all up. And it does create chaos and problems and things like that. So a good gatekeeper is going to contain that insurgent.

We have to talk about Gmail, one of the most insurgent products Google ever shipped. How did it happen? How did you manage that? I would love to hear your story.

Yeah. Depending on how far back you want to go, to me, the roots of it are actually around 1995. I was an intern at Microsoft, which was an exciting time. This is a little bit of a tangent, but they used to have a really good internship program. At the end of the summer, you got to go to a barbecue at Bill Gates's house. That was the first time I met him. At the end, all the interns are gathered around asking him questions. This was like the mid-nineties, when the internet was still controversial; it was still this nerd thing that people would be like, ‘Well, this isn't really for normal people.’ Someone was asking Bill Gates about the internet, and he's like, “Yeah, you know, it's good for nerds, or technical people like us. We love it. But normal people want a more curated experience. Normal people are going to want the newsstand.”

Microsoft was actually building a competitor to the internet when I was an intern. They had this thing called, I think, Blackbird. It was like MSN was going to be this internet competitor. Bill’s analogy was that it was going to be a more curated newsstand experience, because you could go there. It was going to be this much more contained, sanitized experience. You didn't have these weird URLs, HTTP colon, slash, slash, slash, www dot — you know? It's like, ‘What is this gobbledygook?’ A normal person's never going to get any of that stuff.

But to me, what was really cool was it’s Bill Gates saying this, and I'm thinking shit, this guy's the wealthiest man in the world, but I know more about the future than he does. Because he's obviously wrong. The internet's going to fucking blow this stuff away. That was kind of inspiring because I was 18 years old.

I had printed out some white papers from Sun — I don't know how I even came across them — for HotJava, which is now just Java. And the thing that was really cool was this idea of mobile code. So Java, I don't know how much you know about it, but it has this virtual machine that allows you to run code inside of a sandbox, which was very novel at the time. So the idea was that you could visit a website and it would download an applet and then it would actually run code inside your browser. To me, that was the coolest idea. I was reading these white papers on the airplane while I was headed back home from the internship, and I fell in love with this idea of mobile code, because it meant you could distribute computation dynamically across the internet. So I was like — I mean, first of all, I was really into the internet. But then there was this idea that code could migrate across the internet. So the subsequent summer I ended up without a job in part because of — what's the word I'm looking for — hubris or something like that? I sort of assumed they would hire me back, but when I interviewed with the IE [Internet Explorer] team, in hindsight, telling the tech lead their ideas were bad might not have been the right way to get it.

By the summer of 1996, Microsoft had a response to Java called ActiveX that just downloaded native code, and it was a total security hole. It was the dumbest idea ever. I was like, ‘You guys, you're putting a giant security hole in your web browser, but then I decided, fuck it, I'm just gonna make my own thing.

So I actually started trying to build an email client. I had this idea that you should be able to access your email anywhere, with a really high-quality interface. What was normal at the time — I was in college — if people wanted to check their email, they’d have to go back to their dorm room. Because the email would get downloaded onto your computer in your room. So, email had a location. But I was very much into the idea that internet is everywhere. Why should I have to go to my dorm room to read my email? It should just be accessible from anywhere. I wanted to make this high quality web interface that would be like a desktop client, but it would just be inside of your browser. So no matter where you were, you could just log in and it would be like a window into this computer in the sky kind of thing. I started actually trying to build that in the summer of ‘96.

I worked on it for a while, but I have attention issues. I get distracted. I went for a lot of long bike rides around Cleveland and never really finished it. But I always loved this idea of mobile code traveling across the Internet.

So fast forward to like 2001 when I'm at Google. Funny enough, there was a reorganization inside the company because Larry Page felt it was moving too slow. Google had already become slow and bureaucratic in 2001. So Larry got rid of all the engineering managers and they decided to move to this kind of free range engineer model where a bunch of engineers all reported to the VP, but each engineer would have a project. They wanted me to build an email thing — they knew I had an interest in email. So it was back to that idea. By then webmail existed, like Hotmail and Yahoo. But I don't know if you ever used that old stuff. What year were you born?

I was born in 85, I had a Yahoo account.

Yeah, your total quota for email on Yahoo was four megabytes, right? Which is like one iPhone picture today. That would be your entire email quota. Also, the interface was terrible because it was all HTML, bloated with ads. So every time you clicked on anything, the entire page had to reload. It was just a really shit experience. It wasn't something anyone would ever use in a professional capacity. It was very low status too, because it was just like a spam account, basically.

I wanted something that was a really high quality experience. And having learned about my attention issues, I knew I needed to be continually launching. So the very first version of Gmail was a thing I made in a day, basically by taking another thing I'd been working on, which was Google groups, and just loading my email into it. And I asked people, ‘What do you think?’ And people were like, ‘Well, it's okay, but it's better if it had my email instead of yours.’ I'm like, ‘Okay, feature request.’ I had that very iterative approach because that way I could constantly be launching.

In version two of Gmail, I sucked in everyone else's email. That was only possible because it was in a startup environment. Like, I wrote this thing that kind of circumvented security so it could just access everyone's emails. That’s not really possible anymore, because now Google has security mechanisms, but back then I could get into any database and just pull things out. I launched that and people were like, ‘Okay, this is nice. Now I want to be able to reply to an email.’ So we just iteratively worked on that.

One of Gmail’s key disruptions was that we built it in JavaScript, so everything could be — as much as possible — executed in the browser itself, which at the time was not well received inside of Google. There was no JavaScript inside of Google. Famously, we had one line of JavaScript in the entire product, which was when you go to the homepage, there was one snippet of JavaScript that would put the focus in the search box. That was the only JavaScript at Google. So Google had this kind of anti-JavaScript thing, that was also partially technical snobbery. One of the senior technical people said, “You can never scale anything in JavaScript, it’s basically just shit, the project will just turn to garbage.” And actually Eric Schmidt even said, “Oh yeah, my friend at some other big company tried this, it doesn't work.” There was a lot of ‘This has been done before, you just don't know why it can't work.’

And they were partially right. In older browsers, JavaScript really was terrible. But part of what made Gmail possible was, ironically, IE6 and Firefox 1.0, which were just coming up at that time, and were finally good enough that we could make Gmail work without crashing the browser. Early on, I had lots of trouble with crashing the browser because I was pushing it too hard.

Basically the idea was: create this first class experience using JavaScript, it executes in the browser, and then all it does is talk to the server using what is now called JSON. We didn't make up the name, unfortunately, but we made up the concept. The insurgent aspect of Gmail is it completely changed everyone's expectations of how a web app can even behave. Now that’s normal, but when we launched it, people were like, ‘What is this thing?’ And we gave a thousand megabytes of storage at a time when two megabytes was the norm.

A lot of our early power users were journalists. Part of what made it grow is journalists loved Gmail, because when you're working for the New York Times, they would give you something like a 30 megabyte quota for your email and of course, email administrators being sort of like gatekeepers — they're all gatekeepers, all these like sysadmins are inherently gatekeepers — they'd always like, ‘Well, you know, it's very hard to serve email, you just don't understand the technical blah, blah, blah,’ to explain why they had to give everyone a 30 megabyte quota. But we just gave everyone a thousand megabytes for free.

So how do you answer that? If you’re an email administrator and you can't give anyone more than 30 megabytes, but we give away a thousand for free, like it makes you look dumb, right? To me, that’s the one real thing about insurgency: it's just blowing up the gatekeepers. But anyway, the journalists loved it because then they could just keep all their email and search it. It was a hugely valuable thing for them.

Yeah, it's searching yourself, which seems so obvious now. But at the time it wasn’t possible. It wasn’t something people were doing, or even considering. Searching back into your old research, your old conversations, your old thoughts.

Inside of Google, even before we launched, it was a culture change. People would be in a meeting talking about something, but instead of only talking about it, someone else would just be [typing noises], and they’d pull up an old email: “It says that here.” Gmail suddenly gave you instant access to all your old records, all your old emails that had previously been buried. And so it just became a thing that was suddenly normal inside of a meeting: someone pulls up an old email. This wouldn't have happened in the pre-Gmail era, because search was a really clunky, slow thing.

How were you able to do this at Google? I’m assuming the company was a very different place than it is today — what specifically was different?

Fundamentally, it was just that it was run by Larry and Sergey, and they liked that kind of stuff. Because there was a lot of resistance, internally, to Gmail’s very existence. A lot of people were like, ‘Why are we doing this? We should just stick to our knitting: web search. This is a distraction. We shouldn't be doing it.’

There was also a fear of pissing off Microsoft. There was a lot of fear of Microsoft at the time. It was fear that if we launched something like Gmail, we’d paint a target on us as needing to be squashed. In fact, one of the big pushbacks against using JavaScript, believe it or not, was like, ‘If you use JavaScript, Microsoft is just going to break JavaScript.’ But I'm like, ‘They can't do that,’ but that was what more senior technical people inside the company gave as objections to JavaScript. My response was something like, ‘Yeah, but if they change it, I can just change the JavaScript like this, right?’ Microsoft would have had to completely break the whole web browser for me to not be able to circumvent their tactics. People like Eric and Wayne [Rosing] had been fighting against Microsoft from the Sun Microsystems days. They'd been getting squashed by Microsoft, so they had this idea of Microsoft as this unstoppable evil. But ironically, not only did Microsoft not break the browser, they pushed one last emergency patch to IE6 to improve the garbage collector to make it work better with Gmail, because we were pushing the garbage collector too hard in their browser. So that was a cool turn of events.

The Gmail interface was super weird. At one point we had this advisor kind of guy from Stanford who reviewed it, and he wrote this somewhat negative review of it, calling it a Frankenstein because it's a webpage, but it behaves like an application. I also threw in some Unix. So the keyboard shortcuts are from VI, which is a sort of obscure Unix text editor. Why is J and K up and down? It had elements of a webpage, a desktop application and a Unix command line tool.

Gmail’s conversation view was very controversial because again, it was weird and different. Conversation view had never existed. Before Gmail, people had done things like threading, but it wasn’t quite the same as when we actually put the entire thread on a single page and hid the quoted text. There was one VP who was like, ‘I just want AOL mail, AOL mail is perfect.’ I was like, ‘I can't do that, I just can't. I won't.’ Because to me, if you're not creating something new, what are you doing?

There was some pushback on the thousand megabytes [quota]. They were like ‘The competitors are like two and four [megabytes], couldn't we just do 100?’ I'm like, ‘No.’ Part of that was the idea that with Gmail, you shouldn't ever have to delete things, because part of the reason the search is so valuable is because we have everything. I wanted to change people's behavior from one of scarcity to abundance. In the old, pre-Gmail world, everyone lived in this mindset of scarcity. Quotas and things like that. You would only save an email if you thought you needed it in the future. But my thing was, Gmail should eliminate as much of the thinking as possible. So the default action in Gmail, was — we made up this new word ‘archive.’ Archive wasn't a word that anyone associated with email, but it just meant, ‘I want to get this thing out of the inbox, I'm done with it for now.’ It was very controversial when we first launched. If you go back and look, a lot of people said that Gmail didn't have a delete button, which is technically true. I hid the delete button in a pulldown menu. There was a button for archive, but there was no button for delete, because I wanted to retrain people's thought process to not be deletion-based, but archive-based. That way part of the value of Gmail is having all of those old emails, even the ones you didn't anticipate needing. Because then it turns out, for example, that I do need access to that weird email from 10 years ago because I'm getting audited by the IRS or whatever.

Again, this is a hard thing for people who weren’t around back then to wrap their heads around. But people used to delete emails. You had a limited amount of space. You had to delete, you had to choose which things to remember, and which things to forget.

Right. And so they would either delete, or a lot of people — when I went around and sat down with them, had them take me through their email process — had these very intricate folder hierarchies. I would look more closely at these, and I'm like, ‘What's inbox two?’ And they're like, ‘Oh, well, you know….’ They would create these big folder hierarchies, but then they would always fall behind and just have a bunch of shit in their inbox. Eventually they would have too much in their inbox, so they would move everything from Inbox to Inbox 2. So if they were keeping it on their desktop, like with Outlook, they would have less of a quota, because with Outlook, you could go up to two megabytes, and then when you reached two megabytes it would corrupt the archive, and you'd lose everything, which is almost hard to believe. How did Microsoft ship that product? And this was just normal. So early on in Google, I was sucking in everyone's email. So inside of Google, I had copies of everyone's email, and I would get a lot of the really hardcore Outlook users because they would reach two gigabytes and lose everything. And then I had the only copy of their email. So then they would become Gmail users.

What happens at a company following those early days of insurgency? A moment ago, you seemed to say all that matters is the founder. Is it really that simple? Once you lose the founder, does everything change in favor of the gatekeepers?

I think so, because ultimately, again, if you have an established business, the incentives point towards protecting that business. That's just natural.

You do need a certain number of gatekeepers if you have a big successful company. It’s not that they don't add any value. You can't just have people throwing bombs. Actually, one of the funny things is when Gmail launched, it was such a shit show that they added a whole bunch of new things to the launch checklist to prevent something like Gmail from ever launching again.

Do you remember what they added to the checklist?

Some of the stuff was just based around the PR disaster it created.

Wait, what? Why?

Well, the really controversial thing was the targeted advertising. When we launched, we didn't really have the product to begin with. No one had Gmail, no gmail.com email addresses even existed when we launched the product, because we were still finishing the code. So when we did finally get something up, all we really had was a landing page. It said something like, ‘Gmail is special, for three reasons: it has massive storage, powerful search, and targeted advertising.’ Essentially, it wouldn’t have annoying banner ads, it would just give you these nice little helpful links, which honestly never really turned out to be a great product, but did produce unlimited amounts of negative PR. ‘Oh my God, Google is reading your email.’ Which does admittedly sound a little sketchy.

Gmail caused a tremendous amount of legal and press blowback, and part of it was just that no one could even use the product. So all they could write about was the landing page. It could have been a much smoother launch, hypothetically. But to me, the thing that mattered was just the fact that it launched. And again, internally, there was a lot of pushback, because the product really wasn't ready from a bunch of different directions.

Aside from the PR concerns, do you remember anything else on the checklist?

I mean, the whole system was janky. I don't know what all got added. Honestly, I didn't really care. I never launched anything at Google after that, that was the end of the line. But yeah, there was definitely a lot more process put in place because it was legitimately a shit show launch, but I think that's the nature of disruption. It’s like the thing with Twitter. People said, “Elon should have been more careful” instead of firing 80 percent of the team. And everyone thought the service would collapse. They were giving their goodbyes. “Oh, farewell! Elon's such a moron. He doesn't know how software works!” Twitter had a few minor outages, but who cares? It turns out that most of this stuff wasn't that important.

Gatekeepers are one hundred percent anchored to stopping bad things from happening, and they have no concept that when you stop bad things from happening, you are inherently stopping good things from happening as well. You can't ever deliver something that's 100 percent good. If you deliver 80 percent good, that's pretty good. But if you try to go for 100% — if you try to be perfect — what you get is nothing. Innovation is inherently not clean.

This is sort of what it seems like happened with rocketry and launches and space missions in general. You had to be perfect. And that expectation froze the space.

Think about it. There's gonna be congressional hearings when something goes wrong. If a rocket blows up well, ‘We need to do an investigation and make sure this bad thing never happens again,’ and that's that gatekeeper thing of, anytime a bad thing happens, they say, “We need to put process in place.”

Process was a Larry Page thing from early on. I remember he'd joke, “Processes are things you need to kill,” referring to unix processes but also organizational processes. He understood the cost of slowing things down. Organizations like to put processes in place to stop bad things from happening, but process also inhibits good things.

Beyond tech, where do you see a lot of gatekeeping today?

I mean, everywhere. The government is inherently a gatekeeping organization. Because it only grows larger. So if you look at California, most of its problems are essentially parasitic gatekeepers. Like that high speed rail — what was it? How many billions of dollars? You probably know.

Oh, not off the top of my head, but we did write about this.

Let's say it's a hundred billion. It's something like that. Where did the hundred billion dollars go? It went somewhere, right? It wasn't into concrete. It wasn't the cement for the little 500 foot bridge or whatever it was they built. That hundred billion dollars went somewhere. Think about how many people gained power and influence by gatekeeping where that hundred billion dollars went. That's how you gain power, by getting in the way.

The problem with government is like, you can't really blow it up. The thing that makes the free market work is when an old company becomes too burdened by gatekeepers, insurgents can start a competing company and blow them up. Apple can disrupt IBM, Google can disrupt whoever. But with government, you're kind of stuck with it.

To me, the part that's most frightening right now is the AI stuff. Gatekeepers are trying to capture it. The AI bill is a huge thing, because if they lock down AI, they lock down humanity, because AI is going to be in everything we do. It will be in the tools, in our ability to compose thoughts, in our ability to communicate, embedded in all the moderation systems. If they manage to lock down AI, then by extension, all of humanity is also locked down. I think that's the most important thing in the world: that we prevent AI lockdown, and that's why open source is such an important thing.

Open source is the fundamentally insurgent thing because it's uncontrollable. And actually, going back, what are the roots of Gmail? Linux. I got into Linux in like 1993, I loved this idea that no one controls it. There is no gatekeeper because if you don't like Debian, or whatever distribution, or you don't like Linus, you think Linus is an asshole? Fine. Take the code, go make your own version of Linux.

That’s why I was always a huge fan of Linux, which is actually how I ended up at Google. The reason I went to Google was because they were a startup building on Linux and I wanted to work in Linux. And Google only exists because of Linux; in a world without open source, there's also no Google, because we couldn’t just pay $2,000 per machine for a Windows NT license. Part of what made Google work was that we built these clusters of shitty commodity machines and then built reliability at the software level, which now is so normal that no one really thinks about it. But before Google, the way you would do an internet service is that you would buy these really expensive Sun or HP servers that were high reliability, enterprise-grade, and you'd spend lots of money on the machine, and lots of money on the software. And the thing at Google was, ‘We're going to buy the cheapest performance per dollar, even if they're shit machines, and we’ll make up for it in software.’

That was all built on top of open source, so we could do whatever we wanted. Like we actually had to patch the kernel at one point to fix a problem with Gmail. We actually went into the Linux kernel.

I guess one thing — to defend the government the tiniest bit here — is the difference between Google in the early 2000s and the AI folks today. It sounds like you had people at Google actively trying to stop regulation from taking the product down. Today, we have AI people going to DC and asking for regulation. But I watched the congressional hearing with Altman and that clown Gary marcus —

It's so crazy that that guy's as influential as he is.

It's insane. It’s so crazy —

But, at least back then, congressmen from both sides of the aisle seemed reluctant to regulate. So it seems like, in this case, the industry really invited this attention.

Oh, absolutely. That’s true. But again, you have to think about the incentives. I don't think these people have bad intentions. AI is truly dangerous, I agree there is potential for harm. My belief is that centralizing the power only increases the risk of harm. It's not that AI isn't dangerous, it's that centralizing makes it more dangerous. Obviously, there are regulatory capture arguments, like why Google or OpenAI or someone might want it regulated. There are a bunch of reasons why you might ask for regulation. But my opinion is the more you give over that control, it inevitably ends up captured, sort of in the same way that social media was captured. I think about how important it is that Elon took Twitter, but what an unlikely event that was. The only reason we have some freedom of speech is there was a really crazy billionaire who's irrational and impulsive, and on a whim was like, ‘Fuck it, I'm going to buy Twitter.’ I don't think he had a really great business case for blowing $44 billion on this.

It was like, the Babylon Bee was taken down and Elon said “Not today, Satan.”

Right. I think ironically, that's historically one of the most important tweets ever, the one where the Babylon Bee named a government official ‘man of the year,’ and they got censored for that. But prior to that, social media was completely captured. And the way they captured it is really interesting. The network of control. Why were we not allowed to discuss the lab leak? Because ‘It's a racist conspiracy theory, and Disney doesn't want their advertising next to racist conspiracy theories.’ Like, how did they impose that network of control? It was a hard thing to even talk about because people were like, ’You're a crazy conspiracy theorist. These companies, they just arrived at these opinions independently.’ But obviously there's a network of control. And obviously they still want to recapture it. They hate Elon and they hate X because it's uncontrolled. The gatekeepers lost control of the narrative. They used to be able to control the narrative.

Sure, the advertisers nuked as much revenue as they could. But Elon took the company private. He still has to figure out a new model, but provided it doesn’t go out of business the only way to stop him now is some kind of congressional action, which would be extremely hard to do under the First Amendment.

I think there are a lot of levers they can still use. If Biden wins, I don't know. What they're doing with Trump is super scary. Now anyone with high surface area, anyone who gets a lot of attention, can be guilty of an accounting error or whatever, right? If that's the standard, they could put Elon in prison.

Right. At this point, I guess this election is sort of a referendum on that question. But I don't know. That is a whole big, frightening, exciting, but very different topic.

To wrap that back around to the way that they captured social media: if AI is regulated with enough of those kinds of rules in place about preventing harm or whatever — part of the way they capture things is through liability — the AI will be captured so that you're not allowed to talk about, you know, racist conspiracy theories.

Racist conspiracy theories. Right.

Anything that gets labeled a racist conspiracy. All they have to do is tag it as racist or whatever. And you're like, ‘Wait, how is the lab leak racist?’ They said it wasn't a lab, it was Chinese people eating bats. That's not racist? It doesn't even have to make sense. They just have to say it, and then it just gets repeated.

Right, that's how political movements always work. The difference is now we are adding it to the structure of society. It would be root coded into what we're able to talk about. So much of our lives are now litigated online, or I guess taking place online, that if you change the rules of the internet, it seems possible to alter people — society — in a sort of fundamental sense.

You mentioned earlier, the tech industry seems to be this sort of naturally resurrecting thing. The gatekeepers take over companies, the companies then fall, the insurgents create startups and replace them —

Well, this is the free market.

Yes —

It's the reason the market works.

Right. So now, in the context of our country: we used to have the West. You could always go West, escape established civilization, and start something new. America was almost this series of political startups again and again and again, as people expanded into less populated regions and started over.

In the context of a company, it's very obvious what happens when the gatekeepers take over. In the context of a country, it's less obvious. How do you fix a country taken by the gatekeepers?

Yeah, I don't know. This is the constant libertarian question. It’s why we ended up with concepts like seasteading. You’re talking about the loss of the frontier. I get sort of depressed about this question…

I am actually feeling sort of optimistic about the San Francisco stuff, if Garry and crew are able to turn things around. The problem with San Francisco is that what starts in San Francisco doesn't stay in San Francisco. All the crazy stuff always starts here. But if we can actually turn things around, I think that can also be a model that gets copied elsewhere. If they can actually succeed in transforming San Francisco, I think those patterns can spread, because at some point voters will be like, ‘Hey, this is insane.’ If you can demonstrate success, people are good at copying patterns.

Part of the thing that's exciting about an insurgent product is people can no longer debate with SpaceX landing rockets. People can no longer say it's not possible. You just say ‘Fucker, it's right there. I've got a video of it happening.’ Or Gmail. For years, all those email administrators said “You don't understand. It's very hard to offer 30 megabytes of storage.” Okay, now Google is doing it for free with a thousand.

So I think the thing that gives me some hope is if we are able to get a win, if we really are able to transform San Francisco — because San Francisco is world famous for being dysfunctional — into a place where people are like, ‘This place used to be just covered in shit and needles, and now it's a world class city,’ people will be able to see that a thing like that is possible, and they’ll say, “Hey, I want that for my city.” If we’re able to do that to San Francisco, maybe we can do that to California. That's my hope. I don't want to leave California. I love California, and I definitely don't want to leave America. I also don't think there's any place to escape to. I'm not one of the Bitcoiners who are like, ‘I'm just gonna go move to whatever weird country.’ I don't think you can escape America. America is too big to fail. We have to save it.

In any world in which America is gone, you won’t be safe on some little island in the middle of the ocean. We're talking about a world that would be so awful it's like, the game is lost. You lost. The battle for civilization is here. Either we fix it or it's over.

AI is the most disruptive thing we've ever done, like basically since the invention of fire. It’s like fire, and then it’s AI. It can be tremendously liberating.

There are two definitions of democracy that you see people use. One is sort of like a dictatorship of the majority: ‘Well, everyone voted for it, so it's okay, because the president was elected, so whatever they do is democratic.’ The other version that I like is more like — when we talk about democratizing knowledge or whatever, what we really mean is we just made it available to everyone. The promise of the internet was that we would democratize knowledge. And that was true. There’s never been an educational resource as good as YouTube. There are people barely surviving on a few dollars a day who have access to YouTube and they can learn how to become an electrician or whatever. We really did deliver on some of the democratization of knowledge.

If we do AI right, we democratize intelligence. Imagine if all of a sudden everyone in the world is IQ 250. That's transformative. And that’s why I think AI is the thing. AI either centralizes all the power or it decentralizes power. That's the future. It’s like, we either centralize and we're fucked, or we decentralize and actually it's really great.

— SOLANA

Interview edited for brevity/ clarity.

-