Brian Armstrong is the CEO of Coinbase. Please support this podcast by checking out our sponsors:
- Audible: https://audible.com/lex
- Skiff: https://skiff.org/lex
- BiOptimizers: http://www.magbreakthrough.com/lex to get 10% off
- Fundrise: https://fundrise.com/lex
- Athletic Greens: https://athleticgreens.com/lex and use code LEX to get 1 month of fish oil

EPISODE LINKS:
Brian's Twitter: https://twitter.com/brian_armstrong
Brian's Instagram: https://www.instagram.com/brian_armstrong
Coinbase's Website: https://www.coinbase.com
ResearchHub: https://www.researchhub.com
NewLimit: https://www.newlimit.com/

PODCAST INFO:
Podcast website: https://lexfridman.com/podcast
Apple Podcasts: https://apple.co/2lwqZIr
Spotify: https://spoti.fi/2nEwCF8
RSS: https://lexfridman.com/feed/podcast/
Full episodes playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4
Clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

OUTLINE:
0:00 - Introduction
1:49 - Programming
7:47 - Coinbase
38:35 - Startup advice
56:26 - Economic freedom
1:05:36 - Coinbase technology
1:17:03 - Listing on Coinbase
1:26:17 - Regulations
1:40:55 - Crypto dominance
1:45:12 - Employee activism
2:01:30 - Leadership
2:04:09 - ResearchHub
2:13:47 - NewLimit
2:18:15 - Advice for young people
2:22:25 - Meaning of life

SOCIAL:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Reddit: https://reddit.com/r/lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

The following is a conversation with Brian Armstrong, co-founder and CEO of Coinbase, the largest cryptocurrency exchange platform with 98 million users in 100 countries, listing Bitcoin, Ethereum, Cardano, and over 100 popular cryptocurrencies.

I recorded this conversation with Brian before this week's SEC probe into whether some of the crypto listings are securities and thus need to be regulated as such.

As always with conversations that involve cryptocurrency, I try to make it timeless so that the price soaring high or crashing down low doesn't distract from the fundamental technological, economic, social, and philosophical ideas underlying this new form of money, energy, and information.

Our world runs on money, the exchange and store of value.

And cryptocurrency seeks to build the next chapter of how money works and what it can do.

Coinbase and Brine are trying to do this by working together with regulators and governments, which is a long and difficult road.

Bureaucracies resist change, for better and for worse.

The latest SEC probe is a good representation of this.

It is a serious attempt to limit fraud, but 1 that also runs the risk of limiting innovation and limiting financial freedom of individuals.

This is a complicated mess, and I applaud everyone involved for trying to work through it.

I hope in the end, the interest of the individual wins.

Decentralization, after all, is a hedge against the corrupting nature of centralized power.

To support it, please check out our sponsors in the description.

And Now, dear friends, here's Brian Armstrong.

Let's start with the fact that you're a programmer.

What was the first program you've ever written?

The first memory I have of programming was probably in middle school.

And I remember it was recess, and they had this time period where you could read books and the other kids were reading comic books and stuff.

And for some reason I had gotten into this idea that I wanted to get into computers and I was playing with computers at home.

And so I got this book, I think from the library, it was called like, How to Learn Java in 30 Days.

So I was reading this book at like, you know, the recess, and I didn't understand anything.

And I remember I went home and I tried to get this thing working.

And you know, if you've ever written a Java program, the first lines are like, public static void main string args or whatever.

And it's just like, it's so foreign, and it's so difficult to get started.

I was like, I don't understand anything that's happening in this book.

So the first thing I wrote was probably just like a Hello World app in Java, but it was so, I felt like I was so confused about what was actually happening that I later learned a bit of PHP and PHP was like more fun for me because it was like, oh, just print out what you want.

It didn't have all this complexity around it.

So then I got more into PHP, started building like some simple websites, I think learned some HTML.

So I think that was my introduction to programming, at least the very beginning part.

Yeah, you know, Java has a lot of, out of all the Hello Worlds you can possibly write, Java is the 1 where, I think it's the longest.

Which is quite interesting, because Java is often, at least for a long time, was used as the primary programming language to teach people how to program, or at least about object-oriented programming.

I think most universities have now switched, in high school switched to Python.

It lowers the, it makes it less scary, there's like less of a hurdle.

I love PHP and I feel like it's a dirty secret I have to keep private to myself.

Like it's somebody I'm seeing on the side or something like that because it's just not a respected programming language because I think there's so many ways you can write poor code with PHP, which is why it's not respected.

Yeah, it's a scripting language more so, although of course Facebook built like a huge stack on top of it and valuable company, but I still love Ruby to this day.

Python's great too, but I just love the idea behind Ruby that it's like, let's make it easier for the human, harder for the computer, and make it a joy to be expressive and all these things.

So I was never the best computer scientist, but I was a good hacker.

I could rapidly prototype products and using languages like Ruby.

Do a lot of computer science programs still use like Lisp and Scheme and things like that?

No, no, they do for, that's like, that's if you're hardcore.

If you're legit, you're gonna do some of the functional languages.

I think there's a few others that popped up, but Lisp is a distant memory for a lot of people.

That's like somebody that's like, you go to the library, you dust off the book.

I think if you're starting, I mean there's courses about languages themselves, like programming languages.

Lisp might be 1 of those, you know how there's languages that nobody uses anymore, like ancient languages?

You might have to go to school in that same way for programming languages.

Back in the day, we used to use parentheses.

I, of course, still use Emacs as the editor for most things that I do.

And Emacs is, you know, a lot of the customization you can do is in Lisp.

And that's the language probably when I first really fell in love with programming is Lisp.

Because for a long time throughout the earlier history of artificial intelligence, Lisp was the primary language, but it still had a life in the 90s and the aughts where some people would use it.

It was such a beautiful, functional language, but it just somehow didn't pick up.

That said, I should say, sort of push back, I feel like it's still true that most of the web runs on PHP, most of the back end is still PHP.

So if you look at, you know, it's like the stuff that people don't talk about.

It's like what runs most systems in the world, what runs most backend, what runs most frontend?

The Stack Exchange surveys show JavaScript's the most popular language in the world, I think, right?

Oh yeah, in terms of programmers and numbers of, I wonder.

By survey of number of programmers on Stack, Stack Overflow.

Oh yeah, but That's also the cutting edge, right?

Those are the people that are just like excitedly writing code.

I wonder if there's people that are just like maintaining gigantic code bases.

I feel like the amount of Java out there just running industrial systems has gotta be enormous.

And then of course in the banking industry, finance, it's like even older stuff, Cobalt and whatnot, but.

I've been actually looking for somebody to interview who represents Cobalt and Fortran, like who's the figure still, still there that holds the flag.

I did, you know, with Java, founder of Java, a creator of Java, creator of Python, creator of C++, but nobody wants to hold the flag for Cobalt and Fortran, even though some of the most important systems in the world still run on those.

Like power systems and infrastructure systems, which is fascinating.

Like a lot of stuff that we rely on that just works, and the reason we don't change it is because it works well.

It's written in languages that people don't use anymore.

Yeah, that'd be a cool series of interviews.

Get the stuff that's like tech that was invented 40, 50 years ago but still is being used widely.

Let me ask the big question of what are cryptocurrency exchanges and what's Coinbase?

Before, I'll ask even bigger questions, but it's just a nice kind of palate-cleansing question of what is Coinbase?

Coinbase is a cryptocurrency exchange, brokerage, custodian, basically.

We're the primary financial account for people in the crypto economy, how they buy crypto, how they store it, how they use it increasingly in different ways.

So yeah, we wanna be the way that a billion people hopefully access the open financial system globally.

How does it, what does Coinbase actually do?

Okay, so, you know, basically in any given market, there's some people who want to buy some people who want to sell, and there's, you keep an order book of all those prices.

And then, you know, if someone's willing to buy for more than the lowest price, someone is willing to sell, then, you know, you get a trade to execute.

That's kind of how an exchange works underneath.

And a brokerage is kind of simpler than that, even you don't have to know, look at the whole order book and everything.

But you just go in there and you say, I want to buy $100 of Bitcoin or whatever cryptocurrency, you get a quote, and if you like it, you can hit accept.

And you know, the core things that we do to make all that kind of just work, make it seamless, it sounds simple on the surface, is we got to we have to do payment integrations, you know, in a variety of places around the world to make it easy for people to get fiat currency into this ecosystem.

There's lots of hackers out there trying to break into our systems and steal crypto or to put stolen credit cards and bank accounts and things like that into these systems.

We have to integrate with the blockchains themselves, right, which are periodically getting updated and having various airdrops and all kinds of things.

So we're integrated with lots of different blockchains.

And then we have to store the crypto that people buy securely as well.

So crypto is kind of like storing, you store the private keys, essentially, And we've invented a lot of cool technology about how to do that securely.

That helps me sleep at night as 1 of the largest crypto custodians out there.

So those are some of the pieces that had to come together to get that early, simple buy sell experience to work.

And yeah, I mean, Coinbase actually has a lot of different products now.

So we have like an institutional product, we have Coinbase commerce, which is like merchant payments like Stripe for crypto.

We've got a self custodial wallet, which we can talk about, there's all kinds of cool applications people are building with web 3, and they can access it through that we just just launched an NFT product.

So we're sort of like a portfolio of crypto products.

Now we're big enough where we can do multiple things.

But yeah, that's the core thing we got started with.

And still, the majority of our revenue today is people just want to come in and buy and sell some crypto and we help them do that and make it simple and

I'll ask you about wallet and NFTs, about what is it called?

But order books and exchange, what's the difference between that and stocks, for example, which there's also order books?

Yes, I mean, stocks trade through order books too, so do commodities.

So when I wanna buy 1 Bitcoin and I see Coinbase say the price of that Bitcoin is say $40,000

When you press the button on your keyboard, like an electrical signal goes up the wire on your keyboard.

Well, that's also important, the timing, right?

That's, you know, there's a whole concept of like slippage and like, by the time the quote is executed, if the price has moved too much, like we may reject it.

And you know, there's there's various things like that.

But how do I mean, what's the simple version I can give you?

So we you know, we'll, we'll basically check the order book, give you a quote, it's good for some period of time, or for some amount of slippage.

And then what's happening is we're initiating a debit to your payment method, whether that's a credit card or bank account, or, you know, you're storing dollars or euros or something on our platform, there's various payment methods.

And then we're crediting you the crypto and we're taking a fee for it too.

So that's fundamentally what's happening underneath.

How do you calculate how much slippage is allowed?

The dynamics of that is pretty interesting from the little I know about it.

Yeah, so there's a lot of people, like traders who get super into this and like high frequency traders and arbitrage and all kinds of interesting topics.

Flash Boys was like an interesting book on this whole thing.

You want like access to information the fastest, sometimes even putting your, you know, your thing in the data center right next to the thing.

We don't we don't allow that colo stuff because we want it to be more democratized.

But you know, basically, you give a let's say we wanted to just keep it math simple, we want to charge a 1% fee.

So if you're buying $100 of Bitcoin, and we'll charge you

you know, we've presented you the amount of Bitcoin you're going to get for $100.

Now, let's say 10 seconds later, you hit accept, we go to we go to fill the order.

So it's going to be some error bound around that 1% fee, right.

And if we think we're actually losing money on the trade, I think we'll often reject it.

So some part of the fee, the slippage is incorporated into that, averaged over

I mean, just, it's fascinating, because even just that little detail probably requires a lot of experimentation.

Yeah, and It's kind of like a giant bug bounty out there because if you get it wrong, there's people who are gonna arbitrage that.

And we've had people sort of pen test our systems in really creative ways where like, they'll just fire, like programmatically with APIs, they'll fire off like a million different quotes and look for 1 of them that's out of bounds and then actually take that money right there and you know, we get people doing all kinds of crazy stuff.

How do you protect, so we'll talk about cybersecurity in interesting ways, but there's a lot of clever people trying to do clever things to earn, not even just to break into the system, but to earn an edge of some kind in the system.

There's no silver bullet is a bunch of lead bullets, right?

So it's, it's like, you know, 1 thing we do is we just have good test suites, right?

So you're, you're testing every piece of code that goes out.

That's like just common good best practice, but it's particularly important in financial services.

Another thing we do is we hire third party firms to try to audit this stuff and break in.

Another 1 we do is we have a bug bounty program.

So we basically pay white hat hackers to find this stuff before the black hats do.

And we've paid out lots of good bug bounties.

And occasionally you don't get it right, and you lose some money, and then you fix it, and

Let's talk about cybersecurity a little bit more.

You mentioned using stolen bank accounts.

So that's another 1, that's another interesting 1.

Okay, so fraud prevention, yeah, is a big topic.

So 1 of the, there's a lot of things people do, but 1 of the things they do is that use machine learning, right?

So what you want to do is kind of build up a labeled data set of all the different people who have turned out to be fraudulent and good actors, and hopefully collect as much data as you can.

And then you might feed hundreds or thousands of these factors into your machine learning model and it'll come back with a risk score.

So, you know, an example of like the kinds of factors people create or put in there, you know, obviously, I don't want to disclose too many of them, because it's, it's, it's a cat and mouse game, but Just kind of, I don't know, relatively well-known stuff might be, you know, you have device fingerprints, right?

A lot of people who are farming lots of these accounts, they're using emulators and like virtual machines and stuff.

They're not like, you know, an average person on that device.

And then you'll see sometimes like, 1 of my favorite metrics we track for for this was called like improbable travel velocity.

So we would you were tracking people's IPs, right?

And you might see someone who was 1 day in Austin, Texas, and then like an hour later, they were in London or something.

Because like, there's legitimate people who use VPNs too.

But If it's not possible for them to have gotten on a plane and gotten there that quickly, then that's usually they're like spoofing a device or IP.

But yeah, if you feed enough of these in, you will, oh, another fun 1 is like, real users will type their credit card like 1 number at a time.

Scammers have a list of them and they'll just paste in a whole number.

So you can look at like the number of milliseconds between keystrokes.

Like there's all kinds of stuff people have come up with.

You know, for travel velocity, you could probably incorporate VPNs too because there's probably a travel velocity for VPN switching too, that's human-like.

Like if you're using legitimately VPN for something else, that might be, there's like legitimate uses too.

Actually, you know, I feel embarrassed that I don't know this, probably should, but the I'm not a robot.

But how does that even work that well then?

It feels like it should be very crackable.

And yet a lot of high security places use that.

So I think they've, yeah, but it's using a lot of similar signals, like mouse movements, keystrokes, and then obviously all the stuff that comes over the wire with your browser.

So like what operating system, what fonts, you know, what headers are being sent over.

And there's actually, there's an old website.

It was kind of like Panoptic Click or Panopticon or something.

But it basically was like a proof of concept site that they would just show you all the data that was kind of getting sent over with your request and like say that there's only 1 person in the world who has this exact set of data, it's you.

And so it's almost like a workaround, a clever workaround to track somebody, identify a unique person, even if like there wasn't a cookie involved or something.

Where you can't see anybody, you're in the dark, and yet you have a lot of signal and you have to figure out who's a real person, who's not, who's a robot, who's not.

Let me step back, we're gonna jump around all over the place.

So just let's use Bitcoin as a measure of time.

You started Coinbase when Bitcoin was $10.

And you just mentioned an incredible system with security, with transactions, everything is thought through, there's a lot going on, but what was version 1 back in those early days, the first prototype of Coinbase?

Like what did it take to write it, to think through it, and make it work enough to at least make you believe that it's gonna work?

Well, I definitely didn't know if it was gonna work.

I mean, it was kind of, I felt like I was just following my gut.

So, I mean, I was working at Airbnb, I was a software engineer there, project manager, I was working on some fraud prevention stuff, for instance.

And I read the Bitcoin white paper in kind of December of 2010.

I started going to some Bitcoin meetups in the Bay Area, met lots of interesting people there, like crazy people, anarchists, you know, like really brilliant people, all the above.

And so I started nights and weekends trying to put together a prototype.

And my initial thought was, well, you know, SMTP is a protocol that runs email.

And you know, Git is a protocol for version control that people made like Gmail and GitHub, they don't, most people don't want to run their own email server, or even their own Git server, they just want to like, use a hosted thing that, you know, will do all the security and backups for them.

So the thought in my head at that time was bitcoins this new protocol, there's probably going to be somebody who makes a hosted service that does all the security and backups for you makes Bitcoin as a protocol easy to use.

So maybe I should, maybe I should make like a hosted Bitcoin wallet or something that was that was my it was gonna make Gmail for Bitcoin or something.

And a bunch of people told me that was a bad idea.

Like most of my smart friends who I told about it, they were like, well, first of all, I don't really get what you're doing at all.

Like Bitcoin sounds like a scam or something you've gotten involved in.

But then other people who understood what Bitcoin was told me they thought it was a dumb idea because they're like, dude, if you store all this Bitcoin, you're just going to get hacked like nobody, you know, why would you why would you do that?

And So I kind of had this thought like, you know what, I'm not going to go all in and like make a store everyone's Bitcoin.

And I'll tell people this is like a beta thing, like don't put any real money in it and just see if there's interest.

And if I feel like I'm onto something, maybe I'll go do this as a company because I did I really wanted to be an entrepreneur at that time.

And I was I always wanted to like start a company.

But I was, I was, you know, I wasn't yet I was an employee at a company that was great.

But so anyway, I had this prototype, I was hacking together nights and weekends, I actually wrote a whole Bitcoin node in Ruby, which turned out to be a maybe a weird decision in hindsight, because Ruby wasn't the most performant language, we've subsequently had to rebuild that many times.

But Yeah, I had this hosted Bitcoin wallet and the thing that I didn't have any users for it by the way, I applied to Y Combinator because I was like, maybe if somebody there writes me a check, this will like make it feel like a real company.

And I was trying to find a co-founder at that time unsuccessfully.

So I was basically just wandering in the desert.

I had a lot of self-doubt about this, because I was like, I don't know, all my friends don't think this is kind of dumb.

And maybe Bitcoin is just going to get shut down.

And like, this is all be some stupid thing.

So there was definitely a feeling of just wandering lost in the desert, lots of self doubt.

See all Lex Fridman transcripts on Youtube

Brian Armstrong: Coinbase, Cryptocurrency, and Government Regulation | Lex Fridman Podcast #307