Special: Solana (with CEO Anatoly Yakovenko)

Um, yeah. Should we do it? Just dive in? Yeah, let's do it. Welcome to this special episode of Acquired, the podcast about great technology companies and the stories and playbooks behind them. I'm Ben Gilbert, and I'm the co-founder and managing director of Seattle-based Pioneer Square Labs and our venture fund, PSL Ventures. And I'm David Rosenthal, and I am an angel investor based in San Francisco. And we are your hosts. All right, David, on our last episode, we covered the history of Ethereum, what it is, and of course, we speculated on its future. But a big outstanding point at the end of the episode was with all this excitement around DeFi and NFTs and decentralized apps. And I mean, this decentralized world computer has gotten really slow and really expensive.

It's really slow. Yeah. Yes, it is. Sure, the Ethereum community has plans for how they're going to fix this. But listeners, as many of you heard, there are other credible blockchains out there to build applications on top of. I was trying to do some DeFi stuff the other day, no joke. And like, gas fees were 50, 60 bucks. It was crazy. Brutal. That's like close to an all time high. It's wild. Well, one credible alternative blockchain that David and I have been particularly interested in is Solana. So on today's episode, we are joined by Anatoly Yakovenko, the founder and CEO of Solana Labs. Solana Labs is the company that spearheads the development of the Solana blockchain. And just to set some context, they recently raised $314 million, led by Andreessen Horowitz and Polychain

Capital. Yes, that is pi hundred million dollars. And they have a market cap of all their coins right now, right around $10 billion. Both so fun and so crazy. A couple of years ago, if you were going to say like, make a joke, oh, we're raising pi million dollars, you would raise like $3.14 million or maybe $31.4 million. Wow. Just the scale of all this is incredible. Yeah, it's wild. Well, if you are new here, you should join us in the Acquired Slack. And if you like this episode, you will love the community discussing crypto in our digital assets channel. You can join at acquired.fm slash slack. And thank you to listener and community member Austin Fedra, not only for

setting up that channel and curating that community and keeping it awesome, but for his work at Solana and introducing us to Anatoly to make this episode happen. So thanks so much to Austin. Huge shout out. All right, listeners, now is a great time to talk about a new partner of ours here on Acquired, Lagora, the agentic operating system that is redefining how the world's best legal teams work. Yep. It's sort of obvious that AI is going to completely change the legal industry. I bet most of you listening have dropped a contract into some sort of AI chatbot out there. Lagora took that insight and asked the question, what if you really built something with that power from the ground up

for the legal industry? So the founders did exactly what great founders do, operate with obsessive customer focus. They embedded inside a massive law firm for months. They sat with the lawyers just watching how the work really gets done. And that's how you get features that customers love, like tabular review, where you drop in a folder of hundreds of contracts and it pulls every key term into a grid a lawyer can actually work with. Lagora's bet here is interesting. Since it lets each lawyer handle more complexity, any given person can increase the quality of their work and do higher value work. And this means that the pie can grow even as each individual task takes less time.

And they recently launched Lagora agent offering greater intelligence and performance. The agent lets lawyers set an objective. Then it can handle the planning and the execution and delivery of the final product. Legal teams get to maintain full control and transparency since they're still involved where judgment is required. And Lagora works where you already work. You can use it within Microsoft Word while redlining or drafting. The early Lagora numbers essentially speak for themselves. When they have a head-to-head pilot with their top competitor, they win 70% of the time. Lagora now has over a hundred thousand lawyers on the platform from 1,200 legal teams in 50 countries. And crazily, they went from 1 million to 100 million in ARR in about 18 months.

Truly insane numbers. And that is the real test. Plenty of things demo well, but the question is whether a busy associate actually reaches for it during crunch time or whether a partner trusts it before going into a conversation with a major client. If your legal team wants to check it out, whether you're a law firm or you're in-house at a company, you can learn more at lagora.com slash acquired and just tell them that Ben and David sent you.

Well, listeners, as you know, this is not investment advice. We may hold, I think we even will talk about on the interview with Anatoly that we do hold Solana. And yeah, you should make your own independent research decisions before buying anything, whether it's a cryptocurrency or a stock or anything like that. And the show is for entertainment and informational purposes only. Now onto our conversation with Anatoly Yakovenko, the founder and CEO of Solana. Okay. Welcome Anatoly. We are so excited to have you here with us for so many reasons. We've been following Solana for a little bit now. Thanks mostly to Austin Federer in the acquired community, founder of the digital assets channel in the acquired Slack, which is a wonderful and vibrant

community in its own right that's developing. And Austin, of course, works with you guys at Solana now. You had some huge news last week where you raised pi hundred million dollars from entries in Horowitz and Polychain Capital, which we definitely want to get the story behind that. And I'm excited to tell your whole story here to the acquired community. Awesome. Yeah. Great to be here. Before we get into what Solana is, why it exists, what problems it's solving, why there's a need for other layer one sort of base layer protocols out there in crypto besides Ethereum. Maybe to start this round is obviously enormous. Lots of demand. You guys are doing wonderfully well. You know, you said that your path is not quite like a lot of other

crypto projects who raised tons and tons of money out of the gate. Maybe can you rewind and tell us a little bit about how you ended up sort of shipping first and what it was like when you first tried to raise capital and why you decided to take this path? Well, I wish it was like just active decisions versus just trying to survive. The project was started kind of end of 2017, start of 2018. And that was really like that right peak of that last boom cycle. But the crash that followed was really fast and brutal.

And this is right when we were trying to raise. And at that time, like the funds that were committed to us like blew up in front of our eyes and like, sorry, guys, we can't fund this. Wow. With these dedicated crypto funds? Yeah. Yeah. And nothing to like, you know, like not to blame the people involved or anything like that. It was just such a volatile time in the crypto industry. The things went from like, extremely promising, like everything's going to be magic and crypto in the span of a year to, oh my God, the world is falling apart. So yeah, we were born out of that fire. And we had enough funds to get a ridiculously amazing team together. I was able to pull a lot of like kind of principal engineer level folks,

senior directors at a Qualcomm that were just itching to go build the next big thing. The true builders there that were like super experienced, and it worked with me for over a decade, like all jump ship. That was really surprising to me too. They're like, as soon as I asked, they're like, yeah, where do I sign up? You were like Eric Yuan building Zoom out of WebEx. Yeah, exactly. Yeah. That was really awesome. And a huge confidence boost in the design of what we're doing was right, because there was a huge amount of technical risk, right? I came up with this idea of proof history. Nobody else was working on it. There were a few academics defining verifiable delay functions, but it was really like out there, you know, everybody else in the industry was going a

totally different direction. So that made it hard for us to get a lot of commitments to from investors who were more much more readily eager to invest in somebody with like huge pedigree out of a, you know, MIT or whatever, like an Ivy League. And Anatoly, I want to put a pin as we dive deeper down the crypto rabbit hole here. We have a lot of sort of non crypto native audience of sort of startup native audience.

So I want to highlight you just used a term proof of history. We're going to put a pin in that and come back to it later, because it's one of the things that makes Solana special. But I'm going to do the work of highlighting a, hey, I'm going to force you to define that here at some point, because a lot of us are very new to this. Yeah, totally. So we got the team together. And we had like, decent runway, like I think around two years. And we thought, oh, this is going to be done in a year. This is easy, we'll just ship it. It actually took a little longer, like two and a half years to do it. And cost more, we had to hire more folks than we expected.

And we did a couple raises after that. But we're always a little bit starving, under two years, under 18 months of runway, at any given time, bear market, trying to ship this thing as fast as we could. So we could get to the next stage, which is, let's get users, let's build the use case they're always dreaming of, let's actually get to the adoption stage. Yeah, it feels like when you were first starting, it sounds like you're in this window, right? Where like, the crypto native investors, either their funds blew up, or they just were focused on other technical paradigms. And then the non crypto native investors were probably like, this stuff is radioactive, we're not touching crypto at all right now, right?

For the most part, except for the few amazing folks like Foundation, like Slow Ventures, like Multicoin that actually backed us and stuck with us through every raise. And we also in that process discovered this amazing community of validators, like the people actually that running the nodes, that are crypto AGs that have been doing mining involved in Cosmos and every other network. And it became clear, like in a span of a year that like every round, these guys always came in first without asking any questions. They're like, yeah, let's do this. Like, you guys need more funding? Let's do this. And that was just awesome.

Because they were validators on the Ethereum network and seeing the problems there. It's just folks that are like super deep in the space at a technical level. And a lot of them would just run every network, every cool piece of tech that anybody would try to ship. They're just kind of true believers. And we found that community to be immensely supportive and really kind of like helping us out financially and really being our first customers. These are the first folks that we shipped software to that ran it. They saw how the sausage is made. It failed. They dealt with every failure. So that was really amazing to having built that part.

That's cool. Okay. Maybe so now with this kind of level set and also the level set, just so our listeners get a sense as we're talking, the total market cap of the Solana Protocol is about $10 billion. So you have come a long way from those days. Let's rewind. How did you get into crypto and blockchains? I get the sense that you weren't a 2009, 2010 running your own nodes, true believer from back then, or maybe I'm wrong.

I mean, I heard of Bitcoin right around that time during the 2008 crash. Definitely is what kind of opened me up to it. And I mined it a little bit on the CPU, lost those keys. I thought about like... Like everyone. Yeah. Like what if I wrote a GPU miner? Somebody shipped that as I was thinking about it. I was like, well, maybe I can do FPGAs. That came out like a month later. And then I remember this is my first true crypto experience. There was a company that was building ASICs, like CPU dedicated ASICs, like dedicated semiconductors that they designed solely for crypto mining. And I wanted to buy some, but they built them and delayed the actual shipping them to

customers for six months so they could mine with them first. Wow. Talk about front running your customers. It was beautiful example of what crypto is, which is it's totally wild west, peer-to-peer system where anything goes like true to its internet core. And you have to like understand that, but who is on the other side of this conversation on the internet, right? Like if you can't verify their keys, right? If it's not software that you can verify, then you can't assume or trust anything. So Anatoly, you were at Qualcomm though, right through all this, which is going to become very important here in a minute. Yeah. What were you working on there?

At that time, 2008, shoot, I was still working on Brew, which is this flip phone operating system that ran like, you know, that was the first mobile app store. And if you're a programmer, this was written in C with C++ compatible virtual tables that we wrote out by hand. So you literally like type, I built the sophisticated macro language to generate these. It like really made me an engineer, like that period of like working on these like dinky 16 bit ARM chips with like barely any memory. Like you had to like really think about every optimization, everything that you're doing as like close to the hardware as you could. That's great.

You teed up exactly the question that I want to ask when you called out scalability. And so in a pre-Solano world, what are the problems? Why is it that all the existing energy and heat, and I'm using those metaphorically in this sense, and attention with the existing systems, you know, Bitcoin, Ethereum, a lot of the altcoins that were being developed in the mid 2010s, like what are the real problems? At the technical level? Yeah. Let's zoom in on scalability and cost.

Yeah. So systems based on proof of work have this wonderful property that when you receive all of the data, you can verify its thermodynamic security from its starting point to its final point. And you can do so in this black box without talking to anyone else. And the subjective, objectively measurable kind of security is something that is really unique to proof of work. And that's what gives Bitcoin its kind of claim on store of value. You know, it's still not proven yet, but that's really the strong argument there. Can you have store of value? You need some objective way to measure this. If you think about gold, I can literally verify that this is the gold element and then measure how much of it exists and

then compute that I have X amount of it out of the known universe. So these systems depend on this property that I can get the entire data set and verify it from scratch. And if you put that limit on it, it means that the underlying chain just can't grow that quickly. It has to be somewhat limited in terms of the amount of data that gets put into the system. And aggressively, maybe you can double that size every two years. But people are very cautious about that because they're now sitting in a $1 trillion asset. And they're afraid that if we do increase the throughput of Bitcoin or something like that, or the gas capacity of Ethereum, that it's going to grow so quickly that it will

become unusable from that perspective and lose its kind of magic power. Magic being the really hardcore verifiability of all of the energy that it took to create the blockchain so far. Yeah, not just the energy, but also that the energy went into something that is consistent from its root point from day zero, that there was no invalid state transitions and nobody monkeyed with the amount of ETH or Bitcoin in existence. So that piece of it, I think, is really interesting. It's like a meme. It's a phenomenon, a religion, right?

These are like what you're talking about for folks who aren't as deep in the crypto community. These are like religious concepts, right? Yeah. So proof of stake is something that Vitalik kind of started publishing about pretty early on, I think, almost right after Ethereum, super early on, right? So this is not a new concept. But the idea with proof of stake is that instead of using this thermodynamic energy, you have a bunch of keys that represent some value in the native token itself. And based on that weighting of those keys, you can then verify the consistency of the chain. But because that token, right, is monopoly money, right? It's virtually manufactured inside this computer, right? It's not real. It doesn't have

the thermodynamic weight to it. So it's something that you have to establish this first trust. I have to find its first root of trust, kind of like my browser has VeriSign as its root of trust, and then everything's derived from that. And that first establishing a trust is this hard problem of me asking the world, hey, am I talking to the right network? Am I in the right network? And that first thing is not strongly objective, it's weakly subjective. And once you kind of bite that like apple of weak subjectivity...

You might as well go all the way and start farming the orchard. Exactly. So Solana, our design is like, to me, this was obvious. Okay, VeriSign, root of certificate, this is how the internet is built. Trust and first use is a standard security model. You verify once, then once you establish the session, it just continues running. If you go all in, then you can just start optimizing this thing to the limit. And this is kind of our philosophy is that break away from all of these chains of proof of work, this idea that you need this full objective verification, and just kind of go all in, make the fastest possible communications network. And then what we're thinking about isn't so much about store of value

or global money. It's more, do we have the most censorship resistant messaging network, this way of passing information throughout the world that guarantees that any two parties connecting to this network will always receive their messages, they can always synchronize, they can always have fair access that nobody can get ahead of them. And that's more of a financial system. It's more of a marketplace. It's literally what like NASDAQ provides you when you like jump through all the hoops and get your machine set up in the same room, server room as NASDAQ runs their markets. They literally give you an ethernet cable that's exactly the same length as everyone else's. So your information gets to the markets at the same speed as everyone else. That's censorship

resistance. So we kind of like bought into this idea. And this is probably because Qualcomm is a communication company. I always kind of think of these like old school terms of radio, Shannon Hartley theorem. Like how do we like, this is just information passing between people, right? I was going to say it's the same thing when, you know, I dial Ben on my on my cell phone and like, I'm pretty sure I'm going to get connected to Ben, no matter how many people are around me.

Right, exactly. All right, listeners, now is a great time to tell you about a longtime friend of the show, Vanta. AI has scrambled the whole security picture. It used to be that you proved that you were secure once a year on audit or a static PDF, then everyone would nod and you're done. But in an AI first world, that doesn't hold up anymore. Yep. Your risk surface changes every week now. A vendor turns on an AI feature or someone writes in a new model without telling IT and your posture is different than it was last week, let alone at your last audit. Vanta's own research found that around 70% of companies have this quote unquote shadow AI running with no security review at all.

Right. And that's where Vanta comes in. They're the leading agentic trust platform, meaning they've built the thing that closes the gap. And the way that they close that gap is Vanta agent. Think of it as a GRC engineer. That's governance, risk and compliance, except that it's software and it doesn't sleep. It finds the issues, drafts the fixes and cuts the time that you'd spend on vendor assessments in half. In half. Which is exactly why more than 16,000 companies today run on Vanta. Companies like Ramp, Cursor and Snowflake all stay audit ready and catch the risks that crop up between audits across every vendor, every AI tool, the whole environment.

And that's the real value. Trust has to be continuous now, which is why Vanta automates your security, your compliance and the work to earn and prove trust. We're huge fans of Vanta over here and literally hundreds of acquired listeners have become Vanta customers at their companies over the years. So you can get $1,000 off Vanta at vanta.com slash acquired. That's V-A-N-T-A.com slash acquired for $1,000 off and just tell them that Ben and David sent you. Anatoly, can I ask you to define a couple of terms? I've heard these terms settlement layer and execution layer. Can you go into that concept a little bit and explain where Solana could fit in?

Yeah. So settlement, I can't name you a single company in the world that does settlement. They were briefly in the spotlight during this GameStacks thing. But basically, this is what is the final state of the trade is when they look at the list of who owns what and they change the values around. And this is what Bitcoin does, right? I send you Bitcoin, an hour later, everyone in the world considers that settled. It's slow. It's official. It's like the very bottom layer.

Exactly. But it needs to be extremely secure. And that security in the traditional world has been commoditized to the level that the company that does it, there's only one of them and they make some small percentage of the things that they settle. But all the fun stuff, derivatives, options, spot markets, futures, perpetuals, whatever sophisticated financial thing that you can imagine, that runs at an exchange, which does execution and clearing. Some of them do clearing, but just think of it as like one layer, not to make it too complicated. And this is where we're talking about physics limits of speed of light and Shannon Hartley theorem of how much information can we observe in the world, compress it into a price for a particular asset, and then throw that price into a

market and have that be information that's then propagated to the rest of the world that can then work in this feedback loop. And those limits are like, there's theoretical speed of light limits on how fast this thing could do, like how big of a marketplace can we build for the entire world. And that's in itself a really interesting hard problem that has nothing to do with settlement. Right, right. And so would it be accurate to say that like, Ethereum is sort of supposed to be or could be this interesting execution layer, but actually, it's really slow, because it's still proof of work based?

Exactly. So Ethereum is like, Bitcoin is fully like, we don't want anything to do with anything that even potentially touches the store of value thermodynamic energy security, we want to be the simplest thing that's easy to verify, easy for an engineer like in college to build a client for. That's cool, right? Like, that's the simplest form of money, maybe that that's really what they're going for. Ethereum is trying to be in the middle where you can do some of that, at least in its current version, in the ETH one, you have proof of work, you have an execution environment, which is secure, but slow. And you can do like kind of sophisticated contracts, constant function market makers, those constraints on Ethereum have driven a ton of innovation. Like AMMs just didn't exist until

Ethereum really popularized them. And what's an AMM? Automatic market makers, Uniswap. Yeah. It's a beautiful piece of math that allows people to passively create markets with capital in like a passive way with just a single cryptographic signature, all of a sudden there's a market and anybody can trade with it. It's literally the New York Stock Exchange in code. Yeah. And in like a hundred lines of code. Yeah. It's amazing. But that has its own inefficiencies because it's not maximizing the amount of information that the world is synchronizing on.

Yeah. For a sense of scale, the whole Ethereum network, the global Ethereum network is limited to what, 65 transactions a second. Is that right? Something like that. So it depends on the gas limits and how complex the operations are. Some measurements go as low as 11, some go as high as 65. From what the last numbers I checked, Uniswap does about 100,000 to 200,000 transactions per day. So right there, that's a huge portion of the capacity of the whole Ethereum network.

And Serum, which is a central limit order book style exchange that runs on Solana, same kind of design as NASDAQ, that does 20 million transactions per day, but for much smaller amount of capital and volume. So if you think about like users, there's way more users in Uniswap. There's more capital in Uniswap. It's much more transaction efficient. But the synchronization of information on the central limit order book exchange on Solana, is just much tighter, much more message heavy. And like, I would consider it's closer to real prices, like it's much closer to the real market.

Yeah. So this is a great transition to like, tell us what is Solana and how does it actually solve these problems? And maybe just to like anchor it one more time, give us the transaction limit that exists today for Ethereum, for Bitcoin, because these are astonishingly low numbers. Because this thing is in the public consciousness, everyone assumes that, oh, they must, there must be tons of transactions. But like, it is incredibly limited in the way that these networks work today until something like Solana.

So what's cool is that seven transactions per second is what the Bitcoin theoretical limit is. It's actually quite a lot. Humans don't do that much stuff per day. If everybody was using it for payments, it would kind of fall over. But if we're talking about global settlement layer for this like, world's reserve thermodynamic money, it's actually pretty good, right? Yep. There's a heck of a lot more than seven credit card swipes globally per second. But in terms of settlement layer, yeah.

What we're talking about is like almost like the wire level. When you send like an international wire, how many of those per second occur, Bitcoin could probably handle that. And maybe that's fine. Ethereum is trying to do something more sophisticated, where you have tokens, you have governance, you have these communities, and they need to trade and move money between them. A lot of this activity occurs on centralized exchanges, people deposit their tokens there, they trade at a high frequency, and then they move them back to Ethereum. But also a bunch of it occurs on Uniswap and these automatic markets. And that's really like the cornerstone of DeFi is that trading bit on Uniswap. My understanding of what Ethereum is today, depending on gas costs,

is somewhere around 11 to maybe 20 transactions per second. And I think when we were seeing fees over 40 bucks a transaction, Ethereum was doing 1.6 million transactions per day. Now that fees have dropped to under a dollar per transaction, I think today it's under a dollar, and it's doing about 1.2 million transactions per day. So we're talking, it's like same order of magnitude as Bitcoin, way more complex transactions, because there are smart contracts that are executing. And there's just a lot more interesting use cases of this decentralized computer, but still relatively the same speed as Bitcoin.

I would say that it's twice as much. That's enough to do a ton of really cool stuff. That's enough for governance for a bunch of communities to form. So the growth that you're seeing in Ethereum is real, right? There's real people using the network just for what it was designed for. And that's really cool. But it really needs this change from either ETH1 to a sharded system or ETH1 to a bunch of roll-ups or layer twos or whoever you want to design that to really take on a billion users. That's just not going to happen there. And what we set out to build with Solana was, I think we just never thought of Solana as replacing digital gold, sovereign money or anything like that.

To store a value necessarily. Right. It was, how do we build a Byzantine fault-tolerant system that can synchronize information globally as fast as it can? And this is really kind of thinking of these systems in a very old school way. Like a morbid analogy is you have a bunch of sensors detecting a nuclear strike. How do you make sure that they act correctly is you need a high degree of independent verifiable sensors that is necessary to corrupt for the system to go wrong. Balaji described this as the Nakamoto coefficient.

So it's the minimum set of independent participants that if they all colluded would break the network. So, right, like I have my sensor array to detect a nuclear strike. How do I know that it's working? I need to maximize that minimum set that would collude against me. And to do that, to really scale that to a very large number, 20,000, 30,000, you need to make a system that just handles a lot of messages per second, does a lot more cryptography per second, uses a lot more bandwidth, requires more hardware. And I honestly thought that everyone else is going to do the same thing, because to me, this was the obvious kind of like, that's the hardest problem. We got to solve that. If we don't

solve that, everything else is BS. Yeah. So how do you solve that? And just because we're storytellers here on Acquired, like how does that stem out of the things that were known to be true at Qualcomm? It seems like it sort of comes out of the TDMA philosophy. Yeah. There's this, if you remember physics, I mean, this literally like these problems went back to like Cold War, how do we detect a nuclear strike reliably? Yeah. And building the first radio towers, if you have two radio towers that go up and they transmit over at the same time or the same frequency, those electromagnetic waves interfere and you get noise. So the first optimization that anyone's ever tried is why don't we give each tower a very synchronized clock and have them alternate?

So you have a bunch of towers. If all the clocks are very well synchronized, that they never overrun each other and they can all transmit information and you get this like, you know, information passes through. And this problem exists in Bitcoin and proof of work because these systems have these agents called block producers. And proof of work is this puzzle that you solve to be allowed to produce a block, right? Once you've proven that you found the puzzle for this particular challenge, you can construct a block that's signed along with this proof of this puzzle and you transmit that. And Nakamoto consensus encourages everyone else to build on top of your chain, like on top of your block. But if there's two blocks that occur at the same time,

there's now two forks because the participants don't know which block to build on next, right? They now have two options. And in the classic Bitcoin world, it would be like, well, which one's longer? I'll go with that one, right? But if two blocks are produced at the same time, then both are longer. Both are exactly the same way. Ah, I see. Right. Both have exactly the same weight and that's the noisy state. Right. So, somebody else has to pick one, right? The next block producer basically picks one at random or the one with the best fees or something like that. And that noisy state is a delay in the network that requires resolution. Same problem as radio. So, I had two coffees and a beer and I was

up to 4am and I had this like kind of eureka moment that puzzles similar to proof of work using the same shout to 56 pre-image resistant hash function. Instead of running it in this like massively parallel, use as much electricity as you can to find this answer as quickly as you can, you actually make that thing slow and recursive and force it to be running in a single core, on a single thread, on the computer. You can force it to prove that it's taking certain amount of time, that there's a forced delay before you do an action. So, I knew that I had this like arrow of time.

So, then you can start timing and assigning time blocks to transactions. Exactly. And this was like, I don't know, I was manic for like a week because... You were at Dropbox at the time, right? Yeah. This idea was like this high level idea of like, there is a way to construct an error of time. And it was so cool to me because there is no mathematical version of error of time. There's nothing that guarantees times for forwards or backwards in Einstein's equations, right? And digital systems are math systems. So, this is like a way that ties, is cryptography even possible in this universe, right? Like this is an open question or will all cryptographic systems be broken?

If they are possible, right? If there's some proof that... If we can get to a point where we can prove that cryptography is guaranteed to be secure, then we can construct an error of time, right? There's a way to move like things forward. And it was so cool to me that it's possible today with these very simple digital systems to do that. And so, can I maybe try and parrot back my understanding in a very simplified way here? And you can tell me if I'm right. You sort of turned this method of propagating the true correct ledger around all the other nodes from a real-time game to a turn-based game where nobody by being able to, you know, in the gaming world, hit keys faster or in the blockchain

world, throw more cores at it and more energy at it. That's not going to gain you an advantage. You have one set sort of turn on this single threaded, single core thing that you're constrained to accomplish your thing. Yeah. I would take it one step further. It went from like even a turn-based game to a scheduled game that you have a certain amount of time to take your turn. And if you don't take it, it's passed to somebody else. You're on the clock. It's like checks.

Right. Exactly. And that difference between going from turn-based to the scheduled system is why Solana is faster than Tendermint or all the other proof-of-stake networks. Is that those classical BFT systems? They are like Ben takes a turn to speak and everybody else checks. Did Ben speak? Did he not? And then David gets a turn to speak after those checks go through. In the schedule system, we literally just had like a button that passed between all the windows in our Zoom.

And if I don't speak during that moment, I can't speak. We need Zoom to ship that feature. Yeah, exactly. I see. So Solana effectively pre-schedules all the moments where I get my turn to speak. Exactly. And that reduces all this overhead of synchronization in every step and every block. And that is more of a real-time system where it's like voice. This is how voice works in cellular networks. This is how most radio networks are kind of the base layer of radio networks is this TDMA.

Oh, that's so cool. So is it then that the proof of history is tied to that, right? Yeah. And that later, more complex realization of how to tie this high-level idea of an era of time to this TDMA construct, that took months to develop. And working with the folks we hired at Qualcomm, that's really when that took root. But initially, I was like, holy shit, I have a mathematical era of time. I know that that's good enough for something.

So this is amazing. Okay. So with this, what's the throughput of the Solana network today? So we've benchmarked this at about 50,000 TPS with peaks over 65. Wow. So that's what Visa is, basically. Yeah. But there's a lot of caveats there because benchmarking is like, I worked at a hardware company, benchmarking is a game. So you can think of it as when TSMC claims three nanometer process, that's the smallest unit that they can make. So when we benchmark, you flood the system with as many non-overlapping, easy, paralyzed transactions as you can. And you can kind of see, okay, under these ideal conditions that we can, in theory, get this much throughput through. When something like Serum runs, it's a central limit order book. Each one of those markets right now, I think is implemented

as like a one megabyte buffer. So each order requires this one megabyte read, one megabyte write. Those are more complicated transactions that require kind of like more work in the system, but they're still parallelizable. And it's not that like there's limits there that can't exceed 65,000 TPS. It's just, if we start hitting those limits, then we go optimize. And then we see what the hardware limits are. And they're like, this is where benchmarking is hard because, you know, you're dealing with like problems in software that if somebody actually needs them solving, it's like three to four weeks of work to do it versus like everything's optimized to the theoretical limit of the hardware itself. And if you backtrack from the hardware specs,

like if you actually start, like, like we did a Qualcomm, we look at a chip, we're like, okay, this is what this DSP can do. That means we can support 64 frames per second at 4k resolution to do post-processing. You start like that's the limit. And then you write the software until the client is happy. But you can tell them the best we can do is 4k video at 64 frames per second. And maybe you ship at V0 24 frames. And they're like, that's good enough, right? Don't work on something else for now. If you start at the hardware level itself, one gigabit networks is really the bottleneck is like the bandwidth, you need one gigabit of input and output per validator

on this global network. And that can in theory support 700,000 transactions per second. So does that mean that this network requires pretty great internet connections in order to be Solana based versus operating a Bitcoin node? Exactly. And that's because this is maximizing the amount of information that anybody can synchronize on over as many parallel kind of like bits as we can. So if all of these events are parallelizable, right, if these are like different markets, right, there's 700,000 different markets, they all want to do one event per second, we can schedule all of them.

Right? That's kind of how you can think about it. And like that capacity is all over a single giant kind of like table of state that you can execute over. And as bandwidth gets cheaper, and as hardware gets better, you know, like AMD doubles the number of cores, NVIDIA doubles the number of GPU core, like lanes per GPU, PCI buses go from like PCI 4, like one terabit to two terabit in the next version or whatever. That means that we can then ingress, like inject more data at these data centers with like bandwidth that right now goes to 20 gigabit and process it and then synchronize it around the world. So at the kind of the base hardware layer, if you're just looking at the hardware specs, I think you can do tens of millions

transactions per second. But this is going to be kind of blood, sweat and tears to get the software to get to use all of those. But I mean, that's like, that's just such an order of magnitude higher than the current state of the Ethereum network. I mean, you could imagine why if you're building Uniswap now or the next Uniswap or the next whatever, we should we should talk about what all this enables. There's just kind of no question where you're going to run most of your transactions.

And it seems like so you mentioned 700,000 is sort of this like theoretical high watermark at this point transactions per second. Like if you wanted to be Visa, you want to be a credit card network, like a super high traffic, high frequency, what's their processing limit per second? So their capacity, I think they've published that they measured it up to 60,000, but probably they handle around like 5000 at most per second, like a steady state. So all of Twitter is 5000 messages per second. Wow. Humans are just really bad at doing like anything interesting.

NASDAQ does maybe two, 300 trades per second. Now there's 30 times more orders per trade. And there's 30 times more cancels per order. Yeah. Because bots are just flooding the system with like, no, it changed the state. That's the thing. This is really about machines. This is not about humans. Exactly. Wow. So Solana is already from a throughput perspective past the point where it could ever, like you would never need anything more out of a network of humans pretty much doing anything in the world. But the interesting part is here, like if we're building a financial system, the value creators right in this thing are like the machines that synthesize other world's information.

They look at all this arbitrary data and then pick a price for a particular asset and throw that information, right? That's value creation. That's a machine that's doing that as fast as it can. And we need that thing to be at the same network as the other value creators that are building products that humans want, right? They actually onboard humans. That's value creation. It's really hard to get humans to do anything. So if there's no intermediaries between those two, that's like the perfect financial system, right? That means that we've kind of like eliminated all of this gunk.

There's nobody taking 30 bips and every different, like every time some paper changes hands, everybody takes a little bit off the top. Like that's really what the stuff is all about. Like how do we eliminate all of that? And it's pervasive to the point that what I consider like a totally unfair market is Google and Facebook, where I go to a website, they steal my data, right? And feed me information I don't want and charge 0.2 cents to somebody that's, right? Trying to advertise to me. But 60% of what they're paying for is fraud, right? Like click fraud.

Like F that, like F all of that. Amen. All right, listeners. Now is a great time to thank our longtime friend of the show, ServiceNow. If you are running a large enterprise, AI agents are likely spread across every team and deploying them is no longer the hard part. Yeah. The hard part is knowing what permissions they have, what employees are using them for, or what decisions AI is making. AI security for an enterprise at scale is not a small concern. Like the risks are real.

Exactly. And the challenge with AI is governing it, securing it, measuring it, and making sure that it actually delivers value. That is why ServiceNow built the AI Control Tower. Yep. AI Control Tower gives enterprises a single place to see, manage, govern, and optimize AI across the entire business. And it works with any AI, not just theirs. Every device on your network, every permission across every system, every AI agent, visible and secure in one place. And ServiceNow can do this because they've spent more than 20 years building the operational backbone of the enterprise. The workflows, governance, approvals, security controls, and institutional knowledge that power how work actually gets done across IT, HR, customer service, finance, and security. ServiceNow already runs more than a hundred billion workflows annually and trillions

of transactions for more than 85% of the Fortune 500. So when companies need a place to govern AI at enterprise scale, they're building on a platform at the center of how their business already operates. And in a future that isn't going to be one AI, it's going to be thousands of AI agents working across every function of the company. But the question is, who's managing them all? So if you're trying to turn AI ambition into real business outcomes and make it work safely, securely at scale, go check out ServiceNow.com slash acquired and tell them that Ben and David sent you.

One of the, I just think, super cool use cases that I know people are building on Solana, obviously there's tons of DeFi stuff, but Audius, which is a decentralized Spotify and managing streams on the Solana blockchain. Either talk about that or maybe are there other projects that you're seeing out there that are just like, this is super cool stuff that could not exist. Yeah. So Audius, I think is also one of my favorites. And a lot of that comes from Reniel himself.

Like he's just an awesome founder. You guys should totally talk to him. What I love about it is that it is trying to like get away from traditional web to business models of let's steal your data, shove you like stuff you don't want, undercharge our artists. It's really like, here's a platform to connect artists to fans directly, like without any intermediaries. And the metadata that it's collecting, the plays, the fan relationships between users and their artists, it's all done kind of in the open and on a decentralized platform. So it doesn't need to have this web two model at all. Like it can just kind of run off of its token. And this stuff is really hard to imagine like what's going to take off, right? What's actually going to replace web two,

right? It's hard to imagine that, but music has always been like at the forefront of new tech, right? And I'm excited to see like them having like ridiculous traction, 4 million users, like using the stuff with like traction within a small... This is Audius. Yeah. There are 4 million people using Audius. It's amazing. And this is just like a single kind of like niche of like EDM artists. That's like who they went after. They had like all the best people kind of jump in this and love them from the start because music is like such a financially predatory environment for artists, right?

Well, shoot. I mean, rewind like every generation of the internet, music has kind of led the way. Like there's Napster, right? Like no Napster, no Skype, no crypto. Like, you know, it just like, it's all a lineage back to, you know, it's the ultimate peer-to-peer file sharing. But I think a lot of that comes from the fact that like the music industry itself is like predatory on like every financial part of like the ecosystem from the artists to the fans.

Like it's just really trying to squeeze everybody and bring them out. So what's the value proposition here? I always want to just keep playing this. I have this voice in the back of my head all the time. That's like, cool, you made the same thing, but you put it on the blockchain. And in the old world, it's like, well, it's worse because it's slower and it takes more energy. In a Solana world, you're like, okay, now at least it's as fast as a centralized world and doesn't use up a bunch of energy. But what's the reason that it's like better to be a decentralized version of Spotify than Spotify?

So it's peer-to-peer, right? Like that's kind of the cool thing about it. You're not really using Audius. You are directly connecting to like Boyz Noise or Deadmau5. And it's their private key cryptography that's signing those assets in Audius. It's your private key cryptography that's connecting to them directly. It's a direct line. So they know how many streams they're getting from whom and they're getting paid directly. Exactly. Yeah, it's this fascinating concept where in the abstract, if you were to tell me about the internet in 1985, I would believe that the natural way the internet would end up would be decentralized.

Like, wow, it's this thing where anyone can connect to anyone else and anybody can start a business and anybody can put up their own website and author their own HTML. And it's the sort of surprising thing that happened that it led to centralization where 1985, me would be shocked that there are five big companies. Like how, what? That's not what the technology does. And it's quite remarkable that with this next evolution moving from web two to web three of the blockchain that like now we actually do have a chance for 1985, me to be like, oh yeah, okay, that actually makes sense on how I would have guessed it ended up.

And this is why we don't even care about this like sovereign money narrative, right? Like if you think about like what we're solving for Audius, it's this piece of artists with a cryptographic key, a bunch of fans with cryptographic key, and we're guaranteeing that they can talk to each other. We're guaranteeing that when you just described that anybody can self-publish, do whatever the hell they want, that there's no interference there. That's it, right? Like the token itself is just there to prevent spam in this super connected world.

So Salada doesn't have a dog in this fight of, you know, should some country adopt this as their currency or change what fiat money is being used? I think like thinking it from like a information system only, kind of, I think it's actually more interesting to think of it. If we had a super connected world, you know, that's equivalent to like, if we all had like neutrino based lasers that could not be like interfered with that we could all talk to each other at any moment. That's kind of a science fiction thing that we're building, right? In itself.

It's like, I'm reading The City and the Stars by Arthur C. Clarke right now. And that's literally like the, that is the city. Like, let's do that. That's a cool enough thought, like cool enough thought experiment that if that exists, and it's a truly super connected world, what's possible in that world? Like that to me is like an interesting question, in many more ways more interesting to me than what is store value, what is gold, what is money?

Right. Well, Anatolio, I, so in that vein, I mean, there is a cryptocurrency called Sol, you know, people can buy it all over the place. I think on FTX is where I've bought it in the past. Like, how does that factor into this? Like, how does that work in the network that you've created? Yeah. So going back to that, like more of an example of nuclear strike, what we're really trying to do is guarantee that there isn't any central authority that can issue who are these like set of nodes that can guard the network. And if we were the root certificate, and we like issued a thousand nodes, and we could issue more at any time, then that effectively makes

us the point of failure, right? That censorship resistant piece is not good enough for prevent nuclear strikes, right? So if you like think of it from like- I mean, I trust you, but I don't know that I trust you very much. So to eliminate this like central certificate root, you limit the number of certificates that could be issued, issue all of them, and let them decide who owns what, which is effectively like what this token represents is a weight in the network in terms of how much vote you have, how much power to censor you have.

And maximizing that minimum set of nodes that can censor the network, you kind of need this thing to be freely transferable and freely reassignable. So it's just naturally how it evolves, right? If we were actually to try to keep some form of control over it, but we did a good job with everything else, then the network could decide to remove us from control at any point anyways. So it's sort of like the people who have the most soul, or they're sort of like, so you distributed all this soul among a group of people. And that group is sort of like the verisign of this network in a distributed way. And people sort of vote ratably on whether something is valid or not.

Right. And the goal is to grow that set as large as possible, right? And like make that minimum set that are up to 33% as large as possible. And that's our religion. Like if we're talking about store of value or something like as being the Bitcoin religion, the religion in like our little community is maximize censorship resistance, you gotta like, that's the number one focus everybody's working on. And that's how this is proof of stake based. When we talked about proof of work, it's like proof of work in the Bitcoin senses. How much thermodynamic energy did you use up in the past, everyone collectively to prove that this thing is real? Proof of stake, if I'm understanding correctly, is more like, well, you've got a bunch of

soul. And so you and all the other people who have the most of it have the most authority in deciding this, basically certifying the or verifying that this is the person you think it is. Correct. Cool. We want to grow that set to be as large as possible. Again, thinking from this like nuclear strike detection. Sensor ship resistant. Yeah. Okay. So this is a good transition to the last big topic we wanted to cover with you guys. Like all this is awesome. The innovation, the technology, just the like, just makes me smile thinking of like technology and liberal arts. You know, you're marrying different concepts from different fields, bringing them together here. None of that matters if you don't get adoption of the network, right?

Like you can have the most beautiful thing, but you got to have people using this. How did you, especially because you started in the, like the crypto winter, how did you get, I don't know, anything? How did you start like getting people involved in this? So like the validators were really like the first group of customers or users and they're like just super positive people that just want to play around with cool tech. And this idea of maximizing censorship resistance, it's just cool. People like kind of think it's cool. It's go try it, right? How cool would it be if it works?

So we kind of got like a really good community of builders that were willing to do whatever we asked them to do, which is like call up a data center, go install a machine there. Like this is not just like running systems at home. You actually need to like understand what you're doing. And that process really taught us that it's not about like reducing friction for users or devs. It's about giving them something that they don't have like anywhere else, like giving them like something so interesting and so cool that they're willing to do the hard work to build and like deploy. And the use case that we thought would be really interesting was what if we like ran NASDAQ on this thing, like this thing that

Serum runs, like price time queue, central limit order book. And through the winter, a lot of people were telling us that DEXs are dead, that we shouldn't even bother. Wow. DEX being a decentralized exchange. Yeah. But I fundamentally believe that if the system is possible, if we can be a world's price discovery engine, then that's like the Google will be the world's information library, like kind of style of like dream. Like that's like the world changing thing that we need to build for.

And you guys never had any ambition to build a DEX yourself, right? Or did you? Uh, we would have done it on our own. We had plans to do so. But then, um, right after we launched, so we briefly talked to FTX, like nine months before we launched and we were like, Hey, what if you guys, you guys are cool. You're building cool products up and coming exchange. What if we like did options or something like something interesting on chain? And they're like, is it live? Or like, if we think in about six months, we're like, okay, just, just, yeah. Good.

We don't have time for this, right? We got to ship stuff like tomorrow. So six months later, we actually go live and we built this little demo called break, like break.solano.com where you smash your keys and you see transactions fire off. And then you see them all clear, like on the screen, like you see them all confirmed. And these are smart contract transactions. You can go look at the code. And I demoed this to Sam and he was like, okay, everyone else needs to see this. And this engineer's. Sam's the founder of FTX.

Yeah. They like showed this to their like engineering folks. And they've been thinking about this from like a trader's point of view. If you can actually have a system that was fast enough to do full style central and mid order books of the stuff that they know, there's a chance that decentralized finance could potentially get 50% of the world's finance, maybe 25. Doesn't matter. It's big enough for them to like, holy shit, this is like a big opportunity.

Just to make a meta point here. And I think I'm understanding this, right? You could basically say, Hey, you want to run a quant fund on the regular stock market? Well, use our blockchain based software that can execute super fast and go trade regular securities. You don't have to trade blockchain assets. Yeah. You could trade anything, right? Because this is an open network. The tokens can represent whatever the settlement layer tells the chain that it's representing could be stocks, could be something like dollars, like at a USDC. But it's more the idea that if you have this open permissionless platform that is fast enough for this use case, will like things just kind of naturally roll downhill and like end up there anyways? It's yeah, right. It's kind of like separating the

pragmatism from the idealism where like, most of the people who actually bought Tesla's early bought it because it was a status symbol and like a really sexy car. And they said, I'm saving the world. And so like, this actually could be the best fast software to use for a high frequency system that we're describing that may or may not trade blockchain based assets. So you pick it because it's the best system. Oh, but also I have this idealistic view of this decentralized world, and that's the way it should be. And like, is the decentralized world the natural endpoint?

Right. Because it's easier, right? It's cheaper, it's faster, it's fault tolerant, you have stronger guarantees. Like if all those things are true, right? Like, then will things just naturally gravitate there? Like, and as soon as this thing exists, I mean, if you don't have to pay Visa 5% of every transaction, like, you wouldn't. If you don't have to pay, you know, NASDAQ, I think earns like 10 billion a year. Where does that money come from? Yeah. This comes from the people.

I want to rewind back. You talked about the validators. I think this is an important point for building a crypto, at least a protocol that I hadn't quite thought about. It's kind of like supply and demand. You needed the supply before the demand, before you could go get an audience, or you could go build a NASDAQ. You actually needed the network of validators that were going to run this, right? Yeah. And we started with like just 40 of these dedicated people that were just there through our first launch that crashed in 20 minutes. The second launch that crashed in two hours.

Like over and over, they like just came back. And I get what you're saying about them just being like, they're there because like, this is cool. Yeah. They were true believers, right? And like some of those folks were so early and like, we had bug bounties that they were finding these really hard to find critical bugs that they're now founding their own security firms because of this like early investment and like digging through our code, understanding like where we made a mistake and demonstrating it. And like that commitment really created a ton of value for everyone, for everyone involved. It was just awesome to have these folks early on working with us.

That's amazing. All right. So Anatoly, after all this, it sounds to me like Solana is just going to work. It's going to take over the world. Like it's going to be this super high frequency, perfectly optimized blockchain for exactly what it's doing. When I sort of turn my head and see the sort of elephant in the room of Ethereum sort of coming up with ETH2, which is not necessarily the tightest defined at this point, but there's going to be a system based on sharding and moving from purely proof of work to proof of stake.

Like how do you think about Ethereum and Solana? Does it coexist in the same world? Is it one or the other? Are they doing things right? Are they doing things wrong? So what I remember from open source world, like 90s, I was like a Linux geek. I built Linux from scratch and Microsoft was trying to kill Linux. And you just can't kill it. You can't kill an open source community. We can kill Ethereum just as easily as Ethereum can kill us, meaning it's impossible.

Like if you have people that are willing to work on something over the weekend, right? At their own time, just because they think it's fun. There's no killing that, right? Like it's just code that people want to write because they love to do that. But we're competing because there's overlapping features, right? At any given moment, there's a certain number of users. And that competition I think is awesome because we do something while people look at it and like, can my application running Ethereum adopt it, improve on it and vice versa. And we're going to see this kind of bouncing back and forth between any of these healthy communities that are growing, that are like doing the iteration cycle product market fit. And it's not just like, I think the kernel engineers, like honestly,

we're all friends. We all like hang out when we go to conferences, we end up like all at the same bar drinking beers, talking about like, you know, consensus or whatever, just like Windows and Linux kernel engineers hang out because it's just interesting problems, right? But the ecosystems, I think have this like, I think healthy flow of competition that I think is necessary for the entire space, the crypto space to onboard the next billion users. Like this world right now that we live in is just so small. How many people do you think self-custody keys? Do you guys self-custody keys?

Should. I know I should. You guys are in the crypto industry at large, but you don't self-custody keys. Yep. So not your keys, not your coins. You can't directly connect to an artist and audience without your own self-custody, right? You're not in this interconnected world. You're still in this web2 world where you're going through intermediary. Yeah. What percentage of people who think they own cryptocurrency self-custody? I don't know. I think my upper estimate for self-custody is maybe 30 million, but that's like a very generous estimate.

30 million people. Yeah. So DeFi Summer with 30, 40,000 people participating in these flash communities that would form overnight. Imagine that was 30, 40 million people that was rapidly coming together around an idea and pulling resources together around an idea just in an hour and a day around the world with no friction. That's the power of cryptography. That peer-to-peer part of us with keys all connecting. It's just kind of mind-blowing how important that is. The layer one tech is like, I think we're just kind of arguing about Intel versus AMD or like x86 versus MIPS. These are details that are important to us as engineers, but that end-to-end user, like super connected world is just so wild. It's so cool.

Yeah. Well, it's interesting to think about what you were saying a minute ago about you guys and Ethereum, you know, going out to a bar and having drinks and talking about technical problems. Without you guys, I wonder, do you think maybe you can't answer this, but like Ethereum didn't really have any, I mean, there were other, there are other competitors for sure. Let's not say you guys, all of the various Ethereum killers out there. If they didn't exist, Ethereum would just be like dumb, fat and happy, right? Like you need this motivation to like push things forward, right?

Maybe. I can't imagine a world where people wouldn't look at it and decide I want to build something different because the design space for these systems is just so interesting and so big that like as an engineer, it's like operating systems, right? You look at one, you're like, well, I want to make a bunch of different choices and do something different. Like, let's see what happens. Like there's like an urge to try it. Well, give me your pitch. So I'm an application developer. I'm coming with some heavy weight.

Maybe I've got money behind me or a great team behind me. And I'm saying, geez, if I write code that's compatible with Solana, it's not compatible with ETH2 and vice versa. Why should I write it on Solana instead of ETH2? Because we guarantee this super connected world for 1 billion users without shards, without intermediaries. This is where it's going to happen first. Like we're the, this is what we're building for. We don't care about anything else except connecting this billion people together. So you can go deal with shards, layer twos, all this other stuff if you want, or if your vision is bigger than x86 Intel, right? If you're building software and you don't care about the hardware, like you actually care about the user experience, like we're, we're the place to do that.

There's also an element right now too. Like if you want to do this, you're the only place you can do it, right? Like none of these other ETH2 isn't live yet. It's a theoretical concept. We also like, I think if you talk to any of the folks that worked with us, Raj and I are, like it's at the end of the day, it's like about the people. Like, I don't know what it is about Renil that just we instantly connected. I don't know what it is about Sam that there was this like instant trust. Like we're in it together. We'll just go do whatever it takes to, to make this work. There's like people like that in the world that like,

if you're like one of those and you talk to us and you're like, I have this massive vision, I don't know how I'm going to do it. Like we will just boil the ocean. We'll eat glass to make it happen for you without like actually thinking about like any of the, the details are not as important to us. Like right now, like the, the commercial parts of it, like none of that is as important as like actually building something cool.

Yeah. I love it. Love it. All right, listeners. Now is a great time to talk about one of our favorite companies, Statsig. Yes. Long time acquired partner. There is a reason why the best product teams at companies like OpenAI and Notion, Atlassian, Figma, Rippling, Brex, and more rely on Statsig, whether they are iterating on their core product features or shipping AI powered experiences at scale. Yep. In the crazy speed of today's AI world, shipping fast is just table stakes now.

It's basically trivial to build and deploy your app constantly. The real advantage is how quickly you learn what changes actually created value for customers and how fast you can use that signal to guide what you ship next. Whether it's a feature tweak, a pricing change, a performance improvement, or an AI update like a model change or prompt adjustment, they're not relying on instinct. They're measuring what actually moved engagement, retention, and ultimately revenue. And as more teams build with AI, that learning loop becomes even more important. Building with LLMs introduces non-determinism into your product experience. The same input doesn't always produce the same output, and behavior can shift in subtle ways in real world use.

So doing offline evals will give you part of the picture, but you can really only understand the impact once your product is live with real users, and then you can measure how their behavior actually changes. It's very different than the way that you would ship features in a pre-AI world where you knew exactly what the software was going to do in production. Yeah, exactly. So this is where Statsig comes in. It brings experimentation, feature flags, and product analytics into one unified system so teams can ship safely, test rigorously, and directly link what they changed to how users actually behaved. The result is a tighter feedback loop and learning that compounds over time so you don't just ship more, you ship better.

So if you want to make learning your competitive advantage, whether you're building new AI experiences or just evolving your existing core product, go to statsig.com slash acquired to get started. All right, well, to cap it all off, this is so exciting. Everything that's already happening, even just Audius itself is one example of like, this is live. It's happening today. What do you think the Solana ecosystem looks like in two, three, five years? I think, well, the ecosystem, I think like, you know, we have this dream of this being competitive with like something as big as like Google, like Android. There's 5,000 engineers working on Android. 5,000 engineers cost a billion dollars per year. We can't possibly raise that much to

compete with them. I don't know, in this environment, maybe. Right. There's no way, right? Like, so we actually have to have folks come into the space, build awesome products, love the tech, go like start being those one of those 5,000 engineers on their own and like build a community that's more grassroots than this top down, like Google style centralized system. There's just no other option. So like if we succeed, it's like this amazing set of engineers that are just working and building really cool shit and amazing set of validators that are building really cool hardware to make this stuff run as fast as possible. Like, and that's kind of the dream of it, you know, that there isn't this one single central authority that's telling

anybody what to do. Love it. Great spot to end on. Thank you for joining us. Where can people find you on the internet? Solana.com is where you can find all the information. Solana and Solana's Twitter handle, like go follow it on Twitter and follow me if you want to hear me talk about consensus and censorship resistance or Raj, if you want to, you know, he's far more active on Twitter about like ecosystem and like cool shit that's being built.

Are you guys on BitClout yet? No, no, we're not. There's a cool project called Wumbo that is like creator tokens that are part of the, that can like attach to any like social network that's like you're building on the hackathon. If you want to do some work and you're not an engineer, go look through the hackathon submissions and then tweet at us and like, what is good? Because the hardest, one of the hardest parts in like the space is actually synthesizing information, right? Go use your massive brain, the most advanced piece of technology in the universe. Go look at all this information of all these other teams that are building stuff and tell us what's good. What do you like? What's

actually high quality? Uh, that takes a ton of work. Amazing. All right. We'll put links in the show notes. Awesome. Anatoly. Thank you so much. Thank you. All right. Listeners. We'll see you next time. You, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you, you