1 hours 15 minutes 13 seconds
🇬🇧 English
Speaker 1
00:02
If we're limited by bandwidth right now, in 2 years, there's going to be twice as much. Maybe we'll need twice as many cores to drive it because there's twice as much data to sign and cryptographically verify. But these are all Moore's law problems that just kind of solve themselves. But what we need to make sure in the design is that we can always increase the bandwidth.
Speaker 1
00:21
Like if the network capacities go to 100 gigabit in like 4 or 5 years at every data center, they kind of are in a bunch of them already, then Solana could run at 100 gigabit. We don't really care. Like we shouldn't be limited by the design, we should be limited by the hardware. If you kind of design around that, you have a very, very scalable system.
Speaker 1
00:41
It's very elastic to demand up to that limit of hardware. That's really the final bottleneck. And I think we've done a pretty good job with the design because what I see is that the Firedancer folks, without having to make any design changes, are showing much higher throughput by just doing straightforward, okay, you guys had to ship this thing as fast as you could, let us rewrite it in a better, more scalable kind of just raw software architecture way. And they can really demonstrate that all the components can scale very elastically with hardware.
Speaker 2
01:19
What's up, everyone? Welcome to another episode of Lightspeed. Today, we are joined by Anatoly from Solana.
Speaker 2
01:25
Anatoly, thanks for joining us.
Speaker 1
01:27
Thank you for having me.
Speaker 2
01:29
Yeah, yeah, We are pumped to have you on. I've been studying more and more about Solana and every time I do, I get more excited. So I wanna start high level to set the scene and really frame the conversation.
Speaker 2
01:39
So when looking at all the different design choices in crypto, it's easy to really get caught up on what you're building or what's being built and not why it's being built. So I want to start with the end with Solana. So like, what is the Solana thesis? Like, what is it built to enable?
Speaker 2
01:51
And I think that will then lead to why you went with an integrated global state machine and not a chain that relies on charting state with roll-ups, but instead relies on like ever increasing bandwidth and compute.
Speaker 1
02:03
So the, what I thought was really cool about crypto was that it enabled permissionless markets. And if you've ever traded in any permission markets, like I traded like as a total amateur. I had connections to interactive brokers and a bunch of other places and I've run my algorithms.
Speaker 1
02:23
Every time I thought that I had alpha, my data would arrive a little slower and my orders would be submitted a little slower and just felt like everybody was always getting ahead of me. And that's because those are not fair markets. And what makes them unfair is that there is centralized actors that get the information before you, they get to submit the information ahead of you. And that creates massive amounts of value capture for them and losses for somebody an amateur like me.
Speaker 1
02:55
So what I think is really cool about permissionless systems is that they can create fairness. And The whole purpose of Solana that I thought was really, at least it would be valuable to the world, is that we can create fair markets at the core information level, meaning that bits are transmitted everywhere around the world as equally and fairly to everybody concurrently at the same time as physics allow. And that submission is also as fair and competitive as physics allow. So those are kind of the 2 problems that we set out to solve.
Speaker 1
03:30
And I think this is very different than most of, probably every other crypto project that started out at that time, because they started from either the store of value narrative of Bitcoin, we want to build a decentralized money, or settlement guarantees. We want to build a decentralized settlement layer. And none of those really care about when the information arrives, who gets to submit it. Like all those problems are secondary and they're interesting problems.
Speaker 1
04:02
But I think what most people don't realize is if you build the first, if you build this perfect information system, settlement is free. It's just inherently part of it. So solving this much harder problem actually solves all the other ones. So that's really like, I think, kind of our goal, Solana's vision is to be that like perfect single state machine that everyone can rely on for fair access to the data and kind of see it as a source of truth for the world.
Speaker 3
04:31
So obviously the information sinking globally at the speed of light is kind of the big, let's say KPI that you're optimizing for. It's not obviously KPI, but so, and with that said, the first pitch for Salon, the elevator pitch, the catchy slogan was NASDAQ at the speed of light. So my first question is that-
Speaker 1
04:47
Blockchain at NASDAQ speed.
Speaker 3
04:49
Blockchain at
Speaker 1
04:50
NASDAQ speed. OK. I came up with it.
Speaker 3
04:56
Is that still, you think, how people should talk about Solana?
Speaker 1
05:03
I think there is like kind of a bit of a trap there to think of it only in terms of trading. This is something that I realized as building this is that the building a NASDAQ competitor is, and going out there and trying to get stocks listed, I think that could be a thing that people do, but I think that's a very small part of Web3. I think the fact that you have open permissionless markets really disrupts how the web is monetized, how people coordinate globally, and all the kind of really cool web 3 things that you want to do, that I think are much more revolutionary than reducing the amount of take that Citadel and NASDAQ has.
Speaker 1
05:48
They're just much higher impact than that. If everybody in the world has a cryptographic key and they know what to do with it, they know how to sign and coordinate in the DAO, that's effectively full global democracy, full global direct democracy. People can direct funds with that. They can all deposit into a DAO that says, we're going to buy every coal plant on the planet and shut them down.
Speaker 1
06:11
If people know that that will actually happen, I think they will actually throw a thousand bucks in there. And that very, very quickly can become a very meaningful change that is enacted through this. So I think crypto is just much bigger than markets. But what makes crypto interesting, I think, is the fact that people have cryptography.
Speaker 1
06:33
This is, I think Vitalik had a quote that I think cryptography is the first time that defense has an asymmetric advantage over offense. It's not just like you cannot break it, but an atomic bomb you cannot break at 2519 signature. This is kind of like the really cool thing about it. It will take infinite energy to go break that key.
Speaker 1
06:57
And that's really, really cool. And I think That's a very empowering thing. And I honestly think that if we could give humans the cryptographic keys, it doesn't really... The networks will just emerge from that.
Speaker 1
07:10
Like the blockchains, all the stuff, I think those are implementation details to a large degree. The way that I'm interested in attacking the coordination problem is through the thing that I'm good at, which is like low latency optimizations, making things fast. And I think a chain that focuses on that, on synchronizing state, the least amount of overhead as possible has a place in that world where everyone has cryptographic keys and they're doing stuff on chain with them.
Speaker 3
07:42
So my second question, and you kind of answered it as well, is, so you mentioned that the other chains maybe start from a place of like settlement, which I believe you think is the most boring part of finance, store of value like Bitcoin, the sovereign aspect, but then yours is globally shared state, which is actually a pretty abstract concept. I know some people who aren't too familiar with maybe what that means.
Speaker 1
08:07
It's hard to sell it. Like, how do I sell it to like somebody that's like, this is a digital gold. I'm like, that's cool.
Speaker 3
08:16
Exactly. So that leads me to my question, which is how do you rally people around globally shared state? How do you know there's PMF? Do people care?
Speaker 3
08:28
What makes you think that's the right direction to pursue?
Speaker 1
08:34
Well, I know that it eliminates ways for this arbitrage, I think parasitic value capture. When the system gets faster, lower latency, more competitive in terms of who can write. Like the amount of opportunities for somebody to extract unfair value goes away.
Speaker 1
08:54
And that's really important to like all the DeFi nerds. I think the reason why Solana has like a resilient DeFi community that's been able to take these major hits is because they get that vision, that they know that if I build a project here, I'm not going to get screwed by all these intermediaries that can extract massive amounts of math. And they inherently don't want that. That's bad for their users.
Speaker 1
09:17
Math seems like a good thing, right? Like on Ethereum, because it's like, there's like this idea that fees are good, but that's actually value that's extracted from the user. There's 0 chance that Senator Warren of the SEC would allow NASDAQ to take the kind of fees that Ethereum does. They would shut that down like instantly.
Speaker 1
09:41
Like all those regulators have like created everything around finance in the United States to prevent that kind of fee extraction from occurring. So I think for cryptos to be competitive, it has to be better than the existing regulatory systems, not just better because it's open and permissionless, but actually better. It's got to be cheaper, faster, more open, And then I think it can kind of like start replacing those systems and like be a value to the world, you know?
Speaker 2
10:16
Yeah, a hundred percent agree. Let's talk about some of the ways that Slana is actually going about doing that. You said that the final bottleneck is DA bandwidth.
Speaker 2
10:25
What do you mean by that and how does it apply to how performant Slana is trying to be?
Speaker 1
10:29
So like, this was an intuition that I had early on. We didn't even call it data availability bandwidth. It was really just this idea that the chain, you have a quorum, right?
Speaker 1
10:42
This network of nodes that all have to synchronize the state. That's the current active set of nodes in the quorum. The fastest they can synchronize that state is going to be limited by the slowest 1 third of the network, your Byzantine threshold, whatever, however many need you need to get to consensus. That slowest 1 third is going to limit the throughput.
Speaker 1
11:04
So you have to basically optimize and increase that throughput. And what's cool is that when you increase that throughput, you can increase the quorum, because what's very, very trivial to build, A very dumb consensus thing you can build is that every node votes and propagates that vote to everyone else as just part of the way the chain normally schedules all the transactions and all the events. This is why Solana consensus is implemented as a native Solana program, like a smart contract. And that's why votes are normal transactions, is because when you increase TPS and capacity, and therefore the data availability bandwidth at each node, you can actually just add more nodes and increase your quorum.
Speaker 1
11:45
And therefore, that number of users that are getting the data that's around the world increases and you have a fairer and more open system. It becomes harder for you to delay the data from any part of the world and therefore extract value, Information value from that. So this is a very, very kind of brute force approach because it's not algorithmic. It's very much based on if we are limited by bandwidth right now, In 2 years, there's going to be twice as much.
Speaker 1
12:17
Maybe we'll need twice as many cores to drive it because there's twice as much data to sign and cryptographically verify. But these are all Moore's law problems that just kind of solve themselves. But what we need to make sure in the design is that we can always increase the bandwidth. Like if the network capacities go to 100 gigabit in like 4 or 5 years at every data center, there kind of are a bunch of them already, then Solana could run at 100 gigabit.
Speaker 1
12:43
We don't really care. Like We shouldn't be limited by the design, we should be limited by the hardware. And if you design around that, you have a very, very scalable system. It's very elastic to demand up to that limited hardware.
Speaker 1
13:00
So that's really the final bottleneck. And I think we've done a pretty good job with the design, because what I see is that the Firedancer folks, without having to make any design changes, are showing much higher throughput by just doing straightforward, okay, you guys had to ship this as fast as you could, let us rewrite it in a better, more scalable kind of just raw software architecture way and and they can really demonstrate that all the components can scale very elastically with hardware.
Speaker 3
13:31
I'm sure you saw this recently on X.com that, you know, people think that all blockchains and obviously all computers do benefit from hardware advancements and bandwidth advancements and whatnot. But you know, what I said was that Solana benefits, you know, like the benefit that you get from each chain is not proportional, like they're not the same thing. Can you describe, like, how you think about that?
Speaker 3
13:59
Like, if when the hardware increases, the improvements are made, do all blockchains benefit the same? Why not? How should we think about it?
Speaker 1
14:09
It really depends on... So like that kind of idea of like, I kind of made a response to your post saying that it shouldn't be that Solana is scaling with hardware, it's that hardware is all you need. And this really came from my experience at Qualcomm.
Speaker 1
14:29
I was there from like 2003 until like 2015 or something like that. And this was when we went from flip phones to smartphones. And literally every year, we had an architecture change at the CPU and a process change and ARM was getting better. It's just, it felt like Moore's Law wasn't even going at 18-month interval, it was like a 12-month interval time.
Speaker 1
14:53
And it was pretty crazy. And if you didn't write your software in such a way that it's just going to get better by the next chip, you have to rewrite it. So it was like a waste of time to build it in a way that doesn't scale well with more hardware. And that was just really burned into me as I was doing the software development, trying to keep up with like the hardware improvements.
Speaker 1
15:17
That's kind of like the idea there is that, because hardware can scale at this exponential pace, if you design your software such that it just naturally takes advantage of it, whatever inefficiencies you have in the algorithm, Maybe there's an algorithm that's just way more efficient in some parameter. They just shrink. They just become marginally irrelevant. So all the time you're wasting trying to optimize for, you know, like all, like squeeze out extra percentages right now, you should just do analysis, figure out where you're not scaling with the number of cores and fix that because literally in 2 years, you're going to have twice as many cores and those other optimizations are going to be half as important.
Speaker 1
16:00
And this is kind of this exponential curve that humanity's on. My personal belief is if that ever stops, we should all be working on bunker coin. We've hit the limit of human civilization. Things are going to get really bad.
Speaker 1
16:18
So I'm an optimist. I don't think we'll ever get there. I personally believe that the number of cores per system per dollar is going to continue to go up because we're not even stacking them in 3 dimensions yet. These are still single layer, single CPU, not everything.
Speaker 1
16:37
Some phones do have stack RAM, but this is still super early in terms of how much normal Silicon can be pushed. So a lot of other chains do get improvements from hardware. And I don't want to say that, you know, like, I don't think that's what you meant with your quote. It's just Solana and all of our engineering is specifically tailored around this idea that we need to parallelize and maximize the efficiency across the number of disks, across the number of cores, number of network cards, to make sure that in 2 years when the stuff gets half the price, that everything just elastically gets better.
Speaker 1
17:16
And if we achieve that, we're kind of like done with the software architecture, we're done. Like we can kind of rest easy or probably not. We have to work on all the other stuff in the application layer, which is infinite amount of work.
Speaker 2
17:29
When you do that, you get to rely on like Intel and AMD and ISP providers with more bandwidth, right? 1 thing that Solana has been known for is parallelizable execution. Why is it that something like Ethereum didn't originally start with parallel execution?
Speaker 2
17:45
And Does part of that go into the fact that Solana separates state and programs?
Speaker 1
17:51
Yeah, so this is, again, from my experience working on embedded systems, like working on DSPs and things like that, if you've ever had to program a DSP or a GPU, or even like an AVX pipeline on Intel, it's a quite different environment. You have to lay out all the memory and all the lanes for a program ahead of time. So you have to know all the dependencies of all the memory accesses ahead.
Speaker 1
18:15
And then you tell the CPU, here's the code that you're going to execute across these lanes of memory, and then it just zips through them. Because once it knows the locations of all the dependencies, it can prefetch all the memory, thus eliminating any kind of speed hit to a missed memory load or something like that. Everything just runs through the system as fast as possible. And that's really translating, like, making the system be as efficient as possible in terms of execution.
Speaker 1
18:50
But that requires a developer to do this work ahead of time. They have to know all the dependencies of all the memory locations, and yada, yada, yada. And that obviously eliminates features like global memory that you can just go and call random memory or jump to a random memory location. So the way Solana's runtime is designed is from that idea.
Speaker 1
19:09
The developer does the painful work of figuring out what are all the memory dependencies that the program has and conveys that in the transaction that they constructed. And then the runtime is a very easy job. It never basically has a cache miss. So then we can schedule all these events and execute everything in parallel.
Speaker 1
19:31
It's been like, I think, It's been interesting to see how hard it is to do that once you kind of open the Pandora's box and let developers call random locations and do whatever they want with the state. Because just the way that they compose programs once they're given that option is very different. And they'll take advantage of it, right? You give them the tools, they will like treat everything as a hammer.
Speaker 1
19:57
And that becomes very painful to rewrite those systems. So I think it's gonna be very, I think parallelism is going to be really, really hard for EVM. And it almost, given that the fact that the way EVM is designed, it almost makes sense that roll-ups are the way to go because there you're isolating state with tokens and NFTs through a bridge, and then you have a little EVM environment that runs in that state, that's all global, then you have another little EVM environment, another roll-up that does its own thing. And that's how you do your state separation.
Speaker 1
20:30
You've kind of created parallelism that way. But that's a much more, like kind of, it's far less granular, right? You can't like compose between these states very easily. Everything becomes much harder on the composition side.
Speaker 1
20:44
I'm
Speaker 3
20:44
gonna ask 2 questions which won't be that related. So I know you said you have like a theory that being able to process multiple things in your head is like improves the brain hardware or whatnot. First is having said that actually L2s are probably the better way for Ethereum to scale.
Speaker 3
21:02
I'm curious on your thoughts about Monad, which seeks to do parallel AVM. And second, so what's, you know, it feels like it should be obvious to most seasoned engineers that you want to scale with hardware. Yet, I would say you're like 1 of the, I think Solana is probably the first serious attempt at that. And some people will make the argument that, well, the reason is because you can't run these systems on retail internet, right?
Speaker 3
21:31
Like at my house, you need kind of very strong internet connections and whatnot so that people can maximize the number of nodes running. But then you say to that, well, it doesn't matter how low the requirements are, what matters is the incentives to run it. How come That seems to be an unpopular view.
Speaker 1
21:51
Well, I think you have to go back to what you believe is kind of like, what is decentralization? Which is a very philosophical question. And I'm a philosopher.
Speaker 1
22:05
And my belief is that all interesting theories have to be grounded in physical reality. You have to see actually something physical happen for this to be a worthwhile theory to even discuss. So, if we're talking about decentralization, how does it manifest itself in the physical world? And, from my perspective, the key components of decentralization is that the system is permissionless.
Speaker 1
22:34
Like anyone can participate in it. That means that I can physically create my own box. I don't really care about the cost there. Like the cost could be a barrier.
Speaker 1
22:42
We could argue about whether it's 400 a month is too much or $50 a month is like the required amount for decentralization. I think that's almost like, if we get to that, if the discussion ends up there, I've already won. Right? But I think The key part is that the system's permissionless.
Speaker 1
23:01
That means anyone can connect to it. Anyone can create a box. Anyone can create the state. And the other part to it is that there's enough people that give a shit that it's running, that even if something fails, it'll continue it.
Speaker 1
23:15
And that latter part is all about the incentive. So in a system that is very cheap to run and cheap to participate in, I think it works for Ethereum, but this is not going to work for anyone else. If you started a very slow, expensive chain now that no one's participating in, You're not going to get enough incentives for people to actually care that it's running, that they would go and continue it. So you'll fail on the latter part.
Speaker 1
23:39
The fact that Ethereum has succeeded, I think, is because it was early and they created the whole smart contract system and has a lot of very credible, awesome researchers working on it and really working on that towards that goal of settlement. And it has PMF in what it's doing, but there's no way it's like, if somebody else tried to copy Ethereum and launch it, they're gonna fail for the same reason that if I tried to launch a Bitcoin, like slightly, but slightly bigger blocks, slightly smaller blocks, whatever, right? Like it's just somebody already captured that part of PMF. You need to find another incentive structure that will guarantee that people actually care about the state and continue running it.
Speaker 2
24:20
So Solana is a very performant, technically sound chain. And I would say it had a lot of updates in 2022 that went under the radar. And I think a lot of that was with SPF and some network outages.
Speaker 2
24:32
And some of those were quick and stakeway to QoS, which we can go into that, but it's really just like network stability. And you had fees, you added priority fee markets, isolated fee markets, which I would like to touch on and how that kind of corresponds to L2s. You also had things like Solang introduced where you can do Solidity on Solana, you can even do Move now. I'm curious with all that work that you put behind it, how easy is it for, despite this is me being not technical, how easy is it for another ecosystem to almost just copy all those updates that you've had?
Speaker 2
24:59
Because to me, Solana is shipping faster than any other chain out there? And is it really just the community that's the part that's not able to be forked?
Speaker 1
25:09
It's like, you know, BSD had jails, like I think 10 years before Linux had namespaces. It takes even this, like those, all those engineers knew each other. They like, literally we're all friends and we'll get beers together.
Speaker 1
25:24
It's actually really, really hard to yank stuff from 1 part of the, like 1 operating system to another 1. So I think some things could be somewhat transferable. I think you could take SVM and run it as a rollup on Ethereum and encourage people to do that. I think Eclipse is 1 of the teams that's doing that.
Speaker 1
25:44
I think we just need more SVM environments out there. It sounds counterintuitive. People think, oh, it's competitive with Solana. No, it's great.
Speaker 1
25:53
It's like more people that give a shit about SVM. They'll create more value for the entire Solana ecosystem, like Linux. I think those kinds of pieces can be ported over. I think the other pieces, you really have to understand what impact they have and apply them directly in your design.
Speaker 1
26:11
And the real challenge to what I found in building the system is that there's a very... There's almost like an Amdahl's fight. If you're familiar with Amdahl's law, it means that you have a complex system and you look at all the components that create friction or create a cost, a compute cost or latency cost, and you try to optimize 1 of them, even if you make that thing 100 times faster because of all the other parts of the system, the overall impact is very marginal. So you have that effect across everything.
Speaker 1
26:47
You can take Solana's turbine, and if you stuck it in like Ethereum's like block propagation, it'd have very minor impacts overall on like performance. You actually have to go and like take all of these pieces and implement them all at the same time to actually see the overall improvement. And that's really, really tough. I think that's just really, really hard.
Speaker 1
27:08
A lot of, like, especially a lot of the... Like I'd say, nearly all the chains that are running are coming in from this kind of design that's very much like Tendermint-esque. They have like small quorums that do like agreement on a single block, then they move in the sequential kind of step function forward. It's very, very hard to optimize those with the kind of techniques that we did.
Speaker 3
27:37
Yeah, I can't even like when I'm building a React app, I can't even swap my build system, we all own a blockchain. So I can't even imagine.
Speaker 4
27:47
Everyone, We'll get back to Empire in just a minute, but before we do that, I want to let you know that we have Permissionless coming up. Permissionless is a big conference that Blockworks and Bankless put on together. It is the biggest, the best DeFi conference in crypto.
Speaker 4
28:03
This year, it is in Austin, Texas, September 11th through 13th. If you've been in crypto for a while, you know that bear market conferences are the best kind of conferences. We have a phenomenal lineup of speakers. A lot of the guests that you hear on Empire are both going to be speaking there.
Speaker 4
28:19
You will have the opportunity to meet them there. And a lot of the topics that we cover on Empire, ZK tech, roll-ups, account abstraction, MEV, AppChain thesis, a lot of that kind of stuff, that will all be discussed at Permissionless this year. So because you are a listener of Empire, you get a special discount. That's right, Santi and I negotiated with our marketing team.
Speaker 4
28:41
You get 30% off if you go to blockworks.com forward slash permissionless. Empire 30 is going to get you 30% off your ticket. Today when I'm recording this, that's about $300 off your ticket. So type in Empire 30 when buying your permissionless ticket, you get about 300 bucks off.
Speaker 4
28:57
Click the link at the bottom of this episode. It's in the show notes. Do it quickly because prices go up all the time. And if you are going to permissionless, hit me up, let me know, shoot me a DM on Twitter.
Speaker 4
29:08
I would love to meet up with you there.
Speaker 3
29:12
You just mentioned SVM rollups and L2, So I'll use that as my opportunistic segue here. I have been trying to get you to write a blog post on this comment. I'm just going to read it out loud.
Speaker 3
29:24
So you said on Twitter, there aren't any breakout products yet in crypto that depend on low-cost fast networks. That's just it. If you don't believe it will ever happen, ETH L1 is all you need. If you believe it will, L2s aren't going to work.
Speaker 3
29:39
What do you mean by this? Why won't they work?
Speaker 1
29:42
Well, because they fragment the capacity, that they fragment the user base. So like the UX becomes very, very complicated and you kind of, I don't know if you ever worked at a big web 2 company like Dropbox, we had 1 giant MySQL database for all the folders because as soon as you start fragmenting them, links become very hard to create links between different users of different folders. Tracking consistency between those, between 2 different databases, is a major pain in the ass.
Speaker 1
30:13
You would have to synchronize everything through the L1. And that just happens at a large enough scale that that fragmentation creates massive composability, like either UX problems, because you're like, oh, you're not in this rollup, go through another rollup kind of thing. Or you have to go re-sync through the L1 and it's going to create the same kind of costs. And that's very, very difficult to deal with.
Speaker 1
30:42
And you can think of it as like, NFTs are like a perfect example of this. You can't have an NFT in every rollup. It actually can only be bridged to 1. Tokens, you can fudge with this and create different markets, but if I want a specific NFT, the marketplace that it's sold is the marketplace that I need to buy in, and it can only exist in 1 of those.
Speaker 1
31:01
And then therefore, you've broken up composability of just the marketplaces for that 1 single NFT by creating these different states. So that's, I think, fundamental challenge with L2s is their trade-off in this. You get performance benefit because you create this single global lock around a state and then you can go work in it asynchronously from everything else, but you're creating a very hard composability challenge.
Speaker 2
31:31
And so you had a tweet also talking about like the hypothetical example of an L2 building on Solana. Now you said that wouldn't actually affect the fees because it's not that you're competing for bandwidth or compute, it's actually the state contention. So can you maybe talk about that and say how isolated fee markets play into
Speaker 3
31:47
it?
Speaker 1
31:47
This obviously there's limits to this. If Solana is over capacity, fees will go up. But that means it's really bandwidth saturated.
Speaker 1
31:55
And then the dollars per byte that you're submitting to the chain are now going up. There could be other ways that it's saturated through compute and other things, but those actually do tie into state contention, I think a lot closer. So assuming that Solana itself is not saturated, what an L2 does is that you can think of it as a bridge into something that asynchronously process some state. And the way that that bridge works is that it runs a bunch of orders, transactions, and then submits those transactions to the L1.
Speaker 1
32:31
And that submission part, you can think of it as settlement, that can occur in batches asynchronously, like at whatever the cheapest opportunity time is. And the computation part is occurring locally in that system. And what's going to create fees in that state is if you have 1, 000 people that all want the same NFT, or 1, 000 people that all want to take the same liquidation or trade, they will bid up to access that 1 economic opportunity up to the cost of that opportunity. If I'm going to make 100 bucks in this liquidation, I'm going to bid $99 to go be first, because at least I'll make a dollar.
Speaker 1
33:11
That obviously doesn't always perfectly work out like that, but you can think of it at the most perfect, efficient market. Somebody will always bid up to that to be first to like take the smallest amount of profit. So those fees are spiking, not because anything's saturated, it's just simply because there's economic access to state. And that state, you cannot partition that.
Speaker 1
33:34
If you only have 1 liquidation that's on sale, you cannot put it into 2 different rollups and then have half as many feeds. It's simply because there's this 1 liquidation. It's like an NFT. It's unique, and it's a unique opportunity.
Speaker 1
33:47
It doesn't matter if it occurs on the Solana 1 or in a rollout. This obviously breaks down as if Solana itself is saturated. There's so many economic opportunities happening in Solana that all at the same time that you cannot fit them all into the block. And now they're fighting each other to be first.
Speaker 1
34:08
And at that point, what the validator should do is they should double the cores in the network. We should all like, you know, I will personally fly out to 400 data centers or however many. Mert will do it. Everyone will do it.
Speaker 1
34:23
I'm like, go double your course. The network is saturated. At that point, this is kind of, I think, the philosophical difference between Ethereum and Solana is that we're like, OK, we're saturated, just double the cores, and we're good to go. And then prices then go back to what they should be based on the economic opportunity.
Speaker 3
34:43
So 1 thing I want to touch on here, just to make sure we do the L2 folks justice here. You said that the reason why L2s wouldn't work is because of fragmentation and probably talking about complexity as a developer as well. Well, so 2 things to that.
Speaker 3
35:02
1 is that, so are you saying like, are you like bearish that people can come up with abstractions to remedy that? And 2, more importantly, some would argue that you actually only need 1 L2 And then, you know, you don't need all these L2s. What do you say to that?
Speaker 1
35:19
So then that system where you have, like, basically you take SVM that runs as many things in parallel as it can, like up to the limit of the, you know, it always increases hardware capacity to meet the demand. And then you take all that data and you dump it into a Dink sharding, like perfect implementation of a DA layer that tries to maximize its DA system. So you have 1 L2, 1 bandwidth optimized system.
Speaker 1
35:46
That's basically what Solana is. There's some implementation differences there, which I think we're all slowly working on. They're just not a major fire. But Solana itself can asynchronously execute all the programs and then pick forks separately in a separate pipeline than execution of the programs.
Speaker 1
36:10
So that kind of gives you that, like, data's really quickly confirmed, all the forks are quickly picked, but then these bigger systems can go execute on the programs. You can then do like batch ZK verification if that's what you need. Like all that stuff is all doable and so on. This is effectively like that single L2 that can dominate and take up all 100% of the DA bandwidth, that design, you have 1 L2 that's able to take 100% of the DA bandwidth.
Speaker 1
36:40
That L2 needs to implement localized fee markets because if a single use case can saturate the fees for everyone. There's a need for 2 old tubes, right? That breaks down. So then you're broken composability.
Speaker 1
36:52
So it's got to implement exactly the same system as Solana does. That's great. I mean, this is where if you look at the conversations me and Colluding Node have, we end up, okay, everyone's working towards the same goal. I think there's still kind of like a fundamental difference between Solana and something like that.
Speaker 1
37:17
I think like there's not going to be a demand for Solana to do dang sharding because that really breaks that idea of information synchrony across the world. This is the 1 aspect that Solana was started with, and I think is building towards, that I think is really important for a permissionless sub-blockchain to implement. And I think since we're the ones doing it, we should keep doing it. I don't want to split the data availability, even though you might be able to get an improvement in bandwidth there, it actually does still have a trade-off.
Speaker 1
37:53
You still need a system that is working for that use case, that when I submit 1 bit, I hear that it's equally observed in Singapore, in Brazil, everywhere in the world as fast as physics allow, because that actually creates value for the world. That reduces the information asymmetry between players. It allows for fair markets to exist. I think that's the 1 core use case that you really, really can't optimize away with like all these other systems.
Speaker 1
38:22
L2 itself could actually implement that. This is kind of like the L2 that's running in this 1DA layer could be implemented at like perfect synchrony, but it's effectively Solana. It's running the same as fast as possible consensus, and then it's dumping its data to Ethereum because it wants that data availability bridge. So like, sure.
Speaker 3
38:45
So you're basically saying, if I understand correctly, because I might be a little slow today. So you're saying that, well, L2s or the ETH and modularity roadmap and Solana actually have the same endgame. And they would basically have to do what Solana is doing today to scale anyways.
Speaker 3
39:02
Solana just does it in an integrated way, and they don't want to compromise on the information sync being asynchronous around the globe near the speed of light. Whereas Ethereum will have to compromise on that unless they add the sharding layer to the L2 itself?
Speaker 1
39:23
The data availability roadmap for Ethereum, the way I understand it is like the end state is like these ding sharding subcommittees that can all sample each other, they will be able to increase the bandwidth because part of the sub, 1 subcommittee is getting different blocks than another 1. So that implies that you don't have a single state that with perfect synchrony, right? Like you effectively are splitting the system.
Speaker 1
39:51
So it's just that you cannot take that optimization without the sacrifice, right? And if that sacrifice is really, really important to you, then you would design something like Solana where everything is built around how do we propagate the data as fast as possible to as many nodes as possible. Turbine and all these things are the solutions to that. But The idea of having a single roll-up that can take up all of the bandwidth of this data availability layer, whether it's Solana or Celestia or Ethereum, I think that idea is something that I think a lot of the engineers agree is probably the more efficient design.
Speaker 1
40:35
If you have a single rollout that can take up all the data availability bandwidth, you eliminate all of these composition bottlenecks and all this other stuff. But that's very hard, I think, politically to swallow because what are you going to do? Say, arbitrariness is a roll up or optimism. The way that free markets work is the faster, we get faster development when there are free markets and there's competition, we don't have the single enshrined thing.
Speaker 1
41:01
So it's good for Ethereum to actually not enshrine anything, at least for probably this decade.
Speaker 2
41:09
What is the key trade-off that Solana is making?
Speaker 1
41:13
Hardware is more expensive. So If you need more bandwidth, you got to increase the hardware. If you need more bandwidth in Ethereum, you have more subcommittees.
Speaker 1
41:23
The subcommittee hardware is fixed, but that means that each subcommittee is not observing the state at the same time. That's like you cannot solve both of those problems at the same time. There is a fundamental trade-off there. If you care about having state everywhere equally, then you need more hardware.
Speaker 2
41:41
Yeah. It sounds like that's going to happen no matter what, though. If you think about Vitalik's post that he had, it's like block production is going to use more hardware. It's going to be more expensive.
Speaker 2
41:49
It's like Solana is just having a path-dependent way where they're embracing it from day 1, instead of working back to what you see in Ethereum. I think 1 good thing about L2s is it allows for a lot of experimentation, but that's something that you can see with the SVM and L2s as well. So is that, if you're thinking about the SVM and how that benefits Solana, is that kind of the key thing? You might see experiments there that you can adopt, or is it just building out the tooling or something else?
Speaker 1
42:15
I think there's, if you're talking about the application space itself, that's kind of just a never ending problem because like every 2 years like developers are expected to learn a new stack. It's just kind of been part of my career, like you shouldn't. And this is where a lot of the work that we did in SVM to use LLVM and these other components is to allow that kind of experimentation that we can take move or solidity and start building tooling to integrate better with it.
Speaker 1
42:43
But it takes work. Somebody has to drive it and push it and do the iteration. It's not always clear that there's even demand for that because a big part of me thinks that smart contracts are actually not a language that is a distribution language. It's not like Swift for iOS, where you get access to their distribution network, right?
Speaker 1
43:10
Or JavaScript for the web. It's a backend system that's kind of more like building on top of Postgres or MySQL or whatever. And do you really care? Like, do you need that many languages for that or like that many solutions?
Speaker 1
43:26
Like you kind of end up with like, this is like the 1, 2 engineers that write kernel code, your driver code. And once your drivers are built, you have all the other people on top that are building all the applications and everything else. But once those drivers are built, I don't know how much innovation there is. Or it doesn't seem as big of a surface area for innovation as the application layer itself.
Speaker 3
43:53
So I want to zoom out now. I think we've had a little nerd moment, which I enjoyed. So I mean, having said that, Solana and L2s and ETH probably have relatively similar endgames and it's maybe the path that matters.
Speaker 3
44:09
1 thing that comes to mind and something we've, I don't want to say we, but Solana has been heavily criticized for is BD and you know, whatever that means. How do you think about Solana winning this BD battle and this getting just more users and developers on the chain?
Speaker 1
44:30
Oftentimes, like what do you believe is the BD criticism? What is the specific 1 that we did not get some brand or logo to say that they're building on Solana?
Speaker 3
44:44
So, I'm just trying to echo the sentiment that I've heard. I personally don't think there's such a thing. But so, I mean, like people will say like, well, how come Polygon's getting these deals and Solana's not, et cetera, et cetera.
Speaker 3
44:57
Like, how do you think about that? How should the community think about it?
Speaker 1
45:01
So like the way that I run labs, and this is, and like why a lot of the deals we've said no to is because I want labs to get paid for the work instead of paying a customer. So if we want to do an integration, it's got to, sure, we'll do whatever if you pay our engineers to do it. And this is very different from a lot of how other crypto-like startups operate, is actually pay the customer.
Speaker 1
45:34
Like, yeah, we'll do this integration. We'll give you X amount of millions of dollars to get an announcement and stuff like that out. There's some value to that because I think that does establish like brand trust, But I generally feel like that's a lot of money out the door for very little guaranteed usage, because there's no internal driver at these companies that's like, this is a problem that I need solved today and I'm willing to pay to solve it. If you can get somebody like a customer to say that, you know that you have PMF.
Speaker 1
46:08
So I am always searching for PMF and I think it's very important to make sure that all the work that we're doing is being driven by some external demand that's legitimate. This may be surprising to a lot of listeners, but literally like the way that a lot of these deals are structured is that you would pay like brand X to go launch in your chain. And that means that you've not like, you're paying your customers, it feels really, really backwards. So from my perspective is like when we have an opportunity cost, like do we focus on paying some brand or do we pour it into like, like I'd rather throw a party at a hacker house, honestly.
Speaker 1
46:52
At least like our devs have a good time. Or do we put it into like funding? Like we have a Solana Ventures arm that's done like 150 investments or so far across the ecosystem. There's at least like a CEO is like, hey, I'm gonna build this product in Solana and grind for PMF.
Speaker 1
47:09
There's at least a chance of PMF there. So that's been like kind of my, how I've been thinking about it. I don't know if it's right. I think, honestly, looking at Polygon, I've had to think about this.
Speaker 1
47:25
Would it have helped if Solana had more brands, just more announcements or not? Maybe in the short term, not sure in the long term, I don't know. These are like really, really hard questions, right? Like you have X amount of months to every month is a swing kind of thing, you know.
Speaker 3
47:43
On that same train of thought, 2 related questions. 1 is, is that how Slana Labs tends to make money? Like, is that kind of the main business models of labs?
Speaker 3
47:54
And 2, there's also Slana Foundation, which presumably their role is a bit more, you know, getting people to build on Solana. Is that how people should think about it? What's the mental model there?
Speaker 1
48:07
Yeah, that's kind of the way like, foundation is there to like, make sure that there is like a, it's like the Linux foundation. There's a bunch of companies working on open source software. It's got to make sure that it actually all lands and moves forward and kind of like shepherd this whole process of validators and multiple core development teams now and things like that.
Speaker 1
48:29
It's a big pain in the ass, including just providing the hardware to tests, like large networks, stuff like that. There's just a lot of kind of nitty gritty that goes into that part of the development. Labs, We try to do that kind of work where we're like, hey, we'll build you software for money. We try to do investments.
Speaker 1
48:53
We try to find teams that need, like we built templates, like, hey, this is a product that I think should exist. Here's a smart contract for it. We will go out and do audits and then help those teams accelerate. And that could be tied into a venture, like a YC style thing.
Speaker 1
49:11
There's a bunch of different ways. We tried to launch a phone. We launched a phone. The phone's awesome.
Speaker 1
49:18
We got like really, really good reviews from users and trying to get more of those sold. That's like another way to make money in crypto. That 1 is the hardest 1.
Speaker 2
49:28
What would you like to see the Solana community lean into? I feel like a lot of it's organic. You saw NFTs take off in 2021 and then dpen.
Speaker 2
49:36
I don't know exactly how that started in Solana. You had Hivemapper, Helium came over from Polygon, and then you had Render come from the AppChain. And then there's also DeFi. So you kind of have like these different areas.
Speaker 2
49:46
What would you like Solana to lean into?
Speaker 1
49:50
It's honestly like if I was really good at picking what's going to have PMF, I would be more confident in giving that advice. I think that's really, really hard. I really like compressed NFTs because I think that reduction in cost created 2 new kinds of businesses.
Speaker 1
50:12
1 is drip, the other 1 is dialect. So I'd really like to see more folks lean into that, actually go try to imagine what if you could mint basically NFTs for your users for free because the cost of the infrastructure is so low. What are the kinds of business models that you can build that? There needs to be more iteration on those.
Speaker 1
50:35
We need 5 dialects. Not that I want competitive, like we invested in dialect, I don't want more. There's some hard trade-offs there. Do we want more competitors for existing things or do we want the 1 thing to grow?
Speaker 1
50:51
But I honestly think that there's different iterations you can take on that idea of messaging app with NFTs, that dialect, the paths that dialect isn't taking instead of a clone. So I'd rather see people go figure it out and try something new and different. Similarly, I think compression itself, you can apply it to tokens, you can apply it to generic programs. I know I've been asking Merck when compression for tokens, stuff like that.
Speaker 1
51:20
I don't know if that's going to open up the same kind of use cases as compressed NFTs did, but I think it's at least worth trying. Those are the opportunities that I feel like can scale to 10, 20 million users. Like if you build a social network with like a few million users, it probably would be really, really hard to get funding in traditional VCs with like 2, 3000000 actives on your social network, but it would be a runaway hit in crypto. That would be like the biggest thing in crypto right now.
Speaker 1
51:55
So I would like to see people take swings out
Speaker 3
51:59
of it. So you are the CEO of Solana Labs. Solana is a network.
Speaker 3
52:04
There's also Solana Ventures and then Solana Foundation. There's a lot of moving pieces here. And I know like running Helios, that's already pretty like full time, but you have like a lot of different things.
Speaker 1
52:15
Well, I'm only the CEO of Solana Labs. I don't do anything at the foundation. Ventures is part of labs, but it's run by an awesome person, Matt Back.
Speaker 1
52:28
So most of what I do is a lot of Twitter and product development and iteration. And you see me publish SIMDs and stuff like that to where I think the core engineering should go. But honestly, I'm not coding in the kernel anymore, so to speak. I would actually love to do that, but I would mean dropping everything else.
Speaker 1
52:55
This is the big sacrifice of somebody in my position is I can't do the thing that I love and I started this whole thing in the first place.
Speaker 2
53:04
When you're watching from afar, Natalia, is there anything going on in Solana that you would like to change?
Speaker 1
53:09
I think we need to basically fix ecosystem governance. Like I think that was like looking back at the last cycle, there's just been a bunch of failures that could have been prevented with just like some very common sense, basic structures that have existed in like corporate finance for hundreds of years probably at this point. And like, they're very easy to build with realms, DAO, or whatever, or squads, and just have delegation.
Speaker 1
53:39
The reason why proof-of-stake networks work really well, and John Carbono's talk on proof of governance, I think, really touched on this, is that you have well branded actors that are like people delegate to. So there's like an election, there's an epoch for that election, they're sticky, they have to get paid, right, to be a validator. And they care about that income. So they're all incentivized to keep the network on us running because if they all collude to go destroy, they're not all going to collude to go destroy their own value, right?
Speaker 1
54:11
Like the thing that's creating value for them. And that's very actually easy to replicate for governance, for DAOs and for like these token-based startups. And If I could wave a magic wand, I would have every company that launches on Solana, every token project that creates a token literally have a standard template, do elections out of validators, literally go tell the validators to be on these DAOs and run the DAOs and have very standardized templates for token economics. And that would eliminate say 99% of the kind of failures that we saw on both Ethereum or Solana in the last cycle.
Speaker 3
54:51
You just said before this that you wish you could still work on kernels, but you have to do some other responsibilities that a CEO has for Solana Labs. So that brings up kind of 2 questions. 1 is kind of what do you see your role in like the next 10 years in Solana?
Speaker 3
55:08
And 2, when would you ever deem Solana successful and you're just, it's time to pack
Speaker 1
55:15
it up? Time to
Speaker 3
55:15
pack it up.
Speaker 1
55:19
Dang. I have a drink. I would like to see the ecosystem grow so much that I don't have any impact on all the other things besides I can actually like my highest impact thing that I could do is be like a really strong Individual contributor to the code. That's that's what I would think it would be success and I I would just continue doing that 1 of you know my favorite like mentors that like through my career were like these like gray beards that were there from like the 80s that are just like principal engineers that just get to work on whatever really cool, interesting part of the code base that they want to.
Speaker 1
56:03
That would be like kind of my dream to get there. So slowly but surely, this is what retirement is to me, is that like I get to code and the parts of the system that I care about. I think Solana, like metric of success, it's really hard, but I would say that there's these vanity metrics like TVL or it's placed in the top 10 or like the market cap, which I think are like really don't capture like success at all. I think if crypto really grew, like to the point that it has impact to a lot of people, I don't really care about any of those metrics as long as Solana is actually coordinating hundreds of millions of people's private keys, that they're all actually transacting in the network and there's thousands of devs that are building applications.
Speaker 1
56:59
I don't really care about if we're 10th or 3rd or 20th or 50th even, as long as that's happening. I think that's the more important metric to focus on, is are we actually touching people's lives in a daily or even monthly way, like fashion. So that would be I think success for me. Like I hope crypto grows so fast that Solana is not in the top 10, but we also like are impacting like, you know, half a billion people on the planet.
Speaker 1
57:29
That would be like the biggest success that I could imagine.
Speaker 3
57:32
So you're a philosopher early on in the episode. I wanna ask a philosophical question, which is that, you know, you've, I think, I would say Solana Network has been quite successful and you've touched a lot of like lives and helped, you know, I mean, I build my business on Solana, etc, etc. Why are you still here?
Speaker 3
57:52
You know, like, what's the why?
Speaker 1
57:55
That's a good question. I don't know, like, I'm like a person that seeks stress too. Like, I like I have always taken on projects bigger than I can chew.
Speaker 1
58:11
So I think that's part of it. I want to see us get to that point where we do have 100 million people, all with public private keys that are all trading, doing something like coordinating DAOs, all on-chain. I honestly believe in that future. I think it would transform the world.
Speaker 1
58:33
When people talk about lack of democracy and transparency, if everyone in the world was able to transact on Chain, it would be global democracy. It would just really be transformative. So I don't know. There's something there that's interesting enough that that's why I get up every day.
Speaker 1
58:53
I wanna inch us closer and closer to that.
Speaker 2
58:59
Your last 2 answers, Anatoly, were poetic, like too well said, honestly. What I do, we were talking about messaging on top of all this, like for Solana. And I do think as new users come in, it's going to be a whole different story.
Speaker 2
59:11
And they're going to like listen to you, they're going to see what's being built. And I think The performance optimizations that Solana is working on is going to relate a lot more to like Web 2.0 developers. It's really going to hit that once we get past like this first wave of users. I am curious because how you just described Solana probably hits really well as someone that is a bit philosophical or technical.
Speaker 2
59:29
If You just sat next to me in a bar. Why would I use Solana? What would be your pitch to an average user?
Speaker 1
59:36
The basic thing that you can do now that's really cool, I think, are the permissionless payments just like Sun. Start using USDC on Solana. You get settled immediately.
Speaker 1
59:47
You have it. It's yours. And it's very, very strong guarantees on security. That I think is like after 2008 and the crypto collapses and the banking crisis in the US, I think everyone has a bit of understanding that owning your keys with private keys, like owning your dollars with private keys is much, much stronger than trusting some third party intermediary to store them.
Speaker 1
01:00:20
Same thing with I think crypto. I think that part of it is that idea of permissionless money is probably still the most important use case. I think Bitcoin or Ethereum are very important parts of that too. Beyond that, I would say I really like collecting NFTs.
Speaker 1
01:00:37
I would recommend that go find a really cool artist on Drip and go subscribe and get that experience of actually owning your art. Like DJNPOET got banned from Twitter for a hot minute. I don't know if he's back yet, but like...
Speaker 2
01:00:52
He's back.
Speaker 1
01:00:52
Yeah. No 1 was able to take your NFTs. You didn't lose the relationship, that connection that you have. And that's a really cool thing.
Speaker 1
01:01:00
Like, that is actually a totally open platform for artists and like fans to be able to connect to artists.
Speaker 3
01:01:08
Well totally we are going to get less philosophical and more lizard brain. I am going to just ask some rapid fire questions and just answer as quickly as you can, but if you want to expand, you know, feel free to expand, of course. All right.
Speaker 3
01:01:25
What do you and Raj disagree on?
Speaker 1
01:01:29
We actually argue all the time about like what should labs be focusing on what should we be building. And that's like usually things are good if we are able to disagree and like fight through it and get the result. It's bad like when 1 of us is like disengaged, and we don't get that and then the other person just drives it and that thing usually ends up like shit.
Speaker 1
01:01:58
So it's really important that we do disagree and that process is really important. I don't know what we, I'm gonna remember the last thing we disagreed on. He thought, oh yeah, I was like, we should run another NFT, like kind of like hackathon style event that we did. And he thought that it's cringe for us to do it, for labs to do it, it's better that the community is doing it.
Speaker 1
01:02:24
And like, you see that kind of activity of like artists being promoted through like natural, like kind of organic, like places like drip or whatever. So I don't know if he's right, maybe he's right. We'll let
Speaker 3
01:02:39
the audience decide. Yeah. What do you fear the most when it comes to Solana?
Speaker 1
01:02:46
Like a 0 day attack, like 0 day bug. That's like the scariest thing. I can't wait until Fire Dancer is out and there's 2 implementations of the runtime.
Speaker 1
01:02:55
But yeah, zero-day is like the scariest thing.
Speaker 3
01:03:00
What is 1 thing that you were super excited about in Solana, but that nobody else seemed to just quite pick up on?
Speaker 1
01:03:06
People do care about this, but like PDAs, I thought was like the coolest thing that I actually designed in the runtime. It's this idea of like, because it's very, very hard to build these like escrow systems that are able to take ownership of assets. And like, this is the whole account abstraction and all those other things.
Speaker 1
01:03:31
And you look at like, all these are the runtimes. And I had this eureka moment that there's a way that you can do it with by abstracting the public key field. So you didn't have to create an object or a type or whatever. You just simply said that anywhere in the entire Solana runtime, anyone that interprets a public key, 32 bytes as a public key, you can actually substitute it with a PDA, and it just makes composition so natural.
Speaker 1
01:04:01
Devs don't really have to think about any of this stuff. It's just naturally flow. It took a while for that to catch on. I would say a solid year, but I think people appreciate them now.
Speaker 2
01:04:14
And PDAs are program-derived accounts, by the way. Yeah.
Speaker 1
01:04:18
Addresses, not accounts.
Speaker 2
01:04:19
Addresses, my bad. Yeah, GPUs are really hot right now with everything in AI. I think Solana might use GPUs for signature verification.
Speaker 2
01:04:28
Do you see a future where GPUs are used more? Because I know in the initial tests when you were showing, I don't know if it's 500, 000 TPS, whatever it was, you were often quoting GPUs as being used at that time.
Speaker 1
01:04:37
Yeah. When we got to a million signatures per second, that was using pretty old GPUs at this point. I don't want to say 1080, but it might've been like an NVIDIA 1080. Yeah, it was a 1080.
Speaker 1
01:04:50
So like 4 1080s would get you to a million signatures per second. Like the thing is that like AI needs the latest and greatest for this kind of operations that the network needs. You can actually do with like pretty old, old GPUs to get a boost. It's actually really, really hard to use them for that because without something like asynchronous processing of the ledger, because the latency to load the data to the GPU is like in the orders of hundreds of milliseconds.
Speaker 1
01:05:22
So it actually eats into the block time. What the Firedancer guys have done is that they took a very, very old FPGA, like the FPGAs that run on AWS cloud, and they're showing that they can do a million signatures per second, and the latency is like in microseconds. It's like sub 1 millisecond latency. So that's really, really cool.
Speaker 1
01:05:44
So maybe not GPUs, maybe FPGAs. But these are like old FPGAs. We're talking like a few hundred bucks for 1 of these.
Speaker 3
01:05:56
How do you think about competition?
Speaker 1
01:06:04
There's some stuff that I think is good coming out of competition. I think Move is like a really well designed virtual machine. So I think that's cool.
Speaker 1
01:06:15
I try to like basically like the way that you should behave is you should be trying to competing against the best version of yourself. And like looking at the whole industry or like a game, what would the best version of me do? How do I beat them? And they would like gobble up all the best pieces out of everyone and integrate them to accelerate.
Speaker 1
01:06:35
I think stuff that has been done that's good, that's worth copying, is like, I think Move is a really good VM. I think Polygon show that you can get somewhere with actually getting brand deals and accelerating some of it. I think that's maybe important for the foundation to pay attention to. Not so much labs, but OP stack is a cool thing.
Speaker 1
01:07:04
Like the fact that there's like a, now like a kind of almost like a product category, OP stack, the fact that they were able to take this very complicated thing and create it into like a word, I think that is pretty cool. Like, I think it's worth maybe for Solana to do this. And like, we're trying to do this out of labs a bit with like the GameShift developer SDK where we basically aggregate all of these like service providers and APIs across all of these different companies into here's a single stack for game developers so you don't have to go try to figure out how to integrate with 10 different companies. So something like that.
Speaker 1
01:07:44
I think When I see competitors do something interesting, I try to adopt it in some way, at least at labs.
Speaker 3
01:07:53
What does everyone get wrong about Solana?
Speaker 1
01:07:56
That it's not decentralized. It's just dumb. It's just such a dumb thing, especially now, like, hell, we like survived the FTX collapse.
Speaker 1
01:08:06
The fact that the network went through that shit show, and like, has continued running and people continue building it. It's like the most decentralized thing possible. I honestly don't think that like, Ethereum could have survived consensus collapsing at a particular moment in time, that Ethereum would have died.
Speaker 3
01:08:26
What
Speaker 1
01:08:32
does everyone get wrong about crypto? That's a good question. That I think people have to understand that the protocols are not companies.
Speaker 1
01:08:52
And this is like a lot of that I see at a trade file, like this misconception. That's very, very like, they're just applying these old models and concepts to crypto. Like Solana's a bunch of open source code. Anyone can fork it.
Speaker 1
01:09:07
Anyone can fork Ethereum. You can go run your own instances of all these networks. And the token is just really representing the spam defense mechanism for these communications networks. That's it, it doesn't give you voting rights, it doesn't give you participation rights, it doesn't do anything else.
Speaker 1
01:09:27
The only reason why all these other things work is because we're able to coordinate our decision-making power with public keys and quickly come to decisions and build upon that. But it's a very network effect kind of thing. It's not a corporation or a company. It's a bunch of open source devs that are volunteering to work at a particular thing, but anyone can leave at any time.
Speaker 2
01:09:51
So what do you think about ETH? What do you think about ETH as ultrasound money? Do you think it's useful that people in that community want ETH to be a deflationary or is that just marketing?
Speaker 1
01:10:05
I think it's like a dangerous meme. I think it's a dangerous meme. I think the problem with it is that if you think of it in terms of that as an investable thing, the revenues on Ethereum are pinned to the value of Ethereum itself.
Speaker 1
01:10:25
If Ethereum drops, so do the revenues and you have this revenue death spiral. If that's what people are valuing, then it might end badly. Like, this is why I think like, I think it's kind of like, worst meme on store of value on like Bitcoin, because that idea is just like, digital gold. It doesn't promise anything.
Speaker 1
01:10:47
It's just like It's just there it's a dumb pet rock There's no promises of it being anything more anything else than that And I honestly think that's like a less dangerous meme like people I think should be very careful too, with how they approach this idea of like ultrasound money.
Speaker 3
01:11:07
Favorite 3 programming languages? Ooh, like
Speaker 1
01:11:10
I was a big Haskell nerd. Actually, like this is like, I would say Haskell, C and Rust. Those are my top 3.
Speaker 3
01:11:20
Okay, 2 more. Your most unpopular opinion, doesn't have to be crypto, just most unpopular opinion.
Speaker 1
01:11:31
Oh man. Shoot, What is my most unpopular opinion? I don't know.
Speaker 1
01:11:39
I think like within crypto, probably that the hardware costs don't matter. It's probably like that. That one's the most unpopular 1. I don't know besides that 1.
Speaker 1
01:11:50
What else that I have that's pretty unpopular. I Don't know. I'm surrounded by by yes, man
Speaker 3
01:12:01
Well Totally you're obviously a founder yourself for Solana and you're CEO of Solana Labs. There's obviously a lot of people that want to build the future of crypto, and especially on Solana. What advice would you give to aspiring founders?
Speaker 1
01:12:21
I think it's like really, really hard to build a company. Like you saw, I don't know if you saw like Mark Andreessen's post, the 4000 investments, 15 make it to 200 are like profitable or break even or better. 15 become companies that have 100 million more in revenues.
Speaker 1
01:12:45
So like, And those represent 97% of all the returns. So you can kind of think of it in terms of that. Like your probability of success as a founder in anything, even outside of crypto is 15 and 4, 000. And like that kind of dream success, right?
Speaker 1
01:13:05
Like 200 out of 4, 000 break even. That you can consider like a really solid founder story. Given that, I think it's awesome. I think this is the most transformative, amazing experience.
Speaker 1
01:13:22
If you like stress, I seek stress. If you seek stress, if you're 1 of those people that has low heart rate or whatever, It's an awesome, very rewarding, very insane experience. You will feel like you're on top of the world or like the depth of despair like every other week. And you meet like a lot of people that become really close friends from that shared experience that you otherwise wouldn't have.
Speaker 1
01:13:50
I think that kind of trauma really like brings people close together.
Speaker 2
01:13:55
You're really encouraging new people to come to the ecosystem, man.
Speaker 1
01:14:00
And I think the Solana community has awesome, like kind of awesome founder network. Really good, I think, we see is really good founders that help each other, really good businesses to work with. And it's, I think 1 of the best places to go actually start a business.
Speaker 1
01:14:16
I don't think your probability of success is going to really change with whether you're in crypto or not, or which layer 1 you pick or not. I think that's honestly like those are minor points at the end of the day.
Speaker 2
01:14:29
Yeah. Talking about community, it's just Solana, as you said, you chew glass at Solana, right? And it's like over the last year or so, it's like you've been punched in the face a number of times, but you keep coming back. And to me, the community is really stronger than it's ever been.
Speaker 2
01:14:43
And that's why I think this is so impressive. And we'll close this up now. But Anatoly, thanks so much for joining today. I hope this conversation helped anyone that's either ignored what's going on Solana or misjudged it.
Speaker 2
01:14:55
I think it's gonna be a really exciting year and it's a great community to be part of so Anatoly thanks so much for joining us.
Speaker 1
01:14:59
For sure. Great.
Speaker 2
01:15:01
All Alright, we'll see you next time.
Omnivision Solutions Ltd