ep 29 - Debating Blockchain Architectures (with Solana)

Transcript

[Music] welcome to web 3 with Asics and z a show about building the Next Generation the internet from the team at Asics and see crypto that includes me your host sonal choxie today's episode is one of our Deep dive Holloway style chats with Asics and Z crypto General partner Ali Yaya and deal team partner guy woollet in conversation with Solana co-founder and Solona Labs CEO Anatoly yakovenko who also worked at Qualcomm for over a decade where he was a senior engineer and engineering manager among other things I mentioned that because we discussed engineering hiring systems optimization and more in this episode we actually kick off the first half of the episode discussing and debating blockchain architectures including salon and ethereum monolithic versus modular

Etc and really what are the trade-offs and what to optimize for or not depending on what you're building and to make crypto happen at scale now and in the future we discuss how Innovation plays out in theory versus practice and in the second half of the episode we cover company community and eco system building including the nuances of leadership in open source so we also briefly discussed devro governance some of the Solana backstory and the differences between hardware and software Innovation including touching on the Solana phone as a reminder none of the following should be taken as investment legal business or tax advice please see asicsnz.com disclosures for more important information including a link to a list of our investments be sure to also check out the show notes

for links to related episodes or resources on some of the tech topics that come up especially on the evolution of programming role of AI and vdfs or verifiable delay functions the first price you'll hear is Ali's followed by toles and then guides well totally it's uh it's really great to have you here you are one of the most original and contrarian thinkers about blockchain architecture and the technical aspects of how to build a blockchain and decentralized computation generally I think the overarching theme is the big question of what the end game of blockchain scalability is and there are there's obviously so many different approaches there are very various

trade-offs to the different ways of doing things and so to start off I would just want to start with how do you see the end game for decentralized computation what do blockchain architectures look like then maybe we can take it from there well thank you for having me I'll be a bit more contrarian that I actually am you know I'm gonna take a very extreme position here I think that settlement will become less and less important commoditized like it has in traditional finance and you still need stuff for supplement guarantees but those can be achieved many different ways and I think what's really hard is what's really valuable to the world is having synchronized piece of state that's globally distributed and

globally synchronized and you can think of this as like Google spanner does this for Google but in a lot of ways NASDAQ naisy they do that for financial markets they guarantee that there is a certain like chunk of memory that you can rely on that gives you the best price for stuff and from a very high level blockchain systems they're permissionless marketplaces they're very programmable very open you can do lots of stuff for them but ultimately at the bottom of the stack is some kind of marketplace and it's very valuable for all those marketplaces to be completely synchronized as close to the speed of light as possible around the world because then everybody can use that as a reference point you can still run local

markets and do stuff but if you know that this is like the global price for X Y and Z it makes all of the world's finest more efficient and I think that's the end state of blockchain and synchronize as much State as you can at the speed of light that would be the end game one way to ground this conversation is to is to align on what we think the most interesting applications will be at the end game to imagine like now there are a hundred million or billions of users who are using crypto in some way or another and are interacting with blockchains what do you think the biggest drivers of activity on a blockchain will be once we reach that that world my belief is that the kind of applications that you will see that are

actually touching a billion users are very similar to the web 2 kind but it's more transparent it's true to that vision of a long tail internet with a lot of different smaller companies that are able to control their data they're not just being rapidly commoditized by Spotify and all these all these giant Ben like aggregators even though those companies are building great products and do great work but I think the long term is like you have create Leaders with more control more self-publishing and like real internet-wide distribution and markets that drive all of that like under the hood like as fast as possible another interesting way of framing this question is the set of trade-offs you mentioned you think settlement will

become less important in the future I'd be curious when you think about the end game and Solana as the place that a lot of global Commerce specifically Finance happens what are some of the auxiliary components or perhaps compliments to that well that system is not designed as a store of value that's very different problem it doesn't try to actually be as tolerant to internet failure because it's actually trying to be as fast as possible so you use all the available resources on the internet it actually relies on the world where you have mostly free cross-border communication and finance it's not like bunker coins you know and I think there is a need in the world to have a bunker coin that will survive God forbid you know like

Global Wars and like like cold Wars and things that really bisect the world and like these massive like geopolitical challenges something needs to survive that and that's a valuable thing in and of itself but I think that a very optimistic role that I believe is happening right like things are getting more and more connected I think we're going to see 10 gigabit connectivity between us like while you're driving in the US to a person and driving in China I think that's happening within like five to ten years in that world where like you have this completely interconnected world I think this globally synchronized State machine is something that absorbs a lot of the execution side of it and then you have a

lot of places where settlement can occur because the settlement part is very easy to guarantee empirically this is my again very I'm taking this position for a discussion we've seen I don't know so it's 2017. hundreds of different even kinds of proof of stake networks like design wise and like many many different instances of them and we basically have seen zero Quorum failures because settlement is very easy to get right once you have Byzantine fault tolerance and over 21 participants that are fairly decentralized you actually don't see settlement failures all of these other problems that are auxiliary to them I

think are already solved like tendermint I think is a wonderful system that has been unbreakable empirically despite having to go through like a luna-sized crash it wasn't it wasn't the Quorum that failed cell that actually works without any faults at all so I think we are overspending on settlement in terms of security and resources and engineering and way under spending on research and RD on the execution side which is the place where most of Finance makes money and I feel personally that if these Technologies are going to impact and touch the world at scale like a truly like global scale it has to be better than traditional Finance by price by fairness by speed even so this is

where we need to like Focus all our r d on and compete there you're highlighting the fact that settlement is one of the things that you can choose to optimize a blockchain and that actually maybe people over optimize blockchains to do settlement and that there are other things so you can also optimize for that maybe are more important there are certain things along the lines of performance like say throughput and latency as well as composability and things that are also important and often are in opposition against the security of settlement yeah one thing which might be very helpful is is just a high level description of of uh the Solana architecture basically like your vision for how a hyper scalable blockchain looks okay let me give you a brief

overview of what Solara is from like an architecture it's actually pretty simple people it's designed to transmit information around the world as fast as possible to all the participants in the network at the same time so there's no sharding there's no complicated consensus protocol so to speak we've actually tried to keep things pretty vanilla or like we got lucky I guess I would say with one hard computer science problem and this was box synchronization and this is the idea for using a verifiable delay function as a source of time in the network and you can think of that as if you have two radio towers that transmit at the same time or the same frequency as you get noise and the first protocol that people thought were

they started building cellular networks is that why don't we give each Tower a clock and they alternate by time when they can transmit and literally FCC is like a truck full of goons that will drive out to your cell tower and shut it down if it goes out of sync and you can't do that in an open permissionless network like around the world so the idea for Salon actually came from that kind of Eureka moment that you can use a verifiable delay function to schedule block producers so they don't Collide because in a network like Bitcoin if you have two block producers that produce a block at the same time you have a fork and that's the same thing as the noise in the cellular network so if we can force all of them to alternate by time

you get this nice time division protocol where every block producer takes a turn that's scheduled and they never collude and therefore Forks never occur and the network never gets into that noisy State and then everything else that we did after that has been pretty standard fed optimizations that folks have done in operating systems and databases you know for 20 30 years we transmit the blocks around the worlds like BitTorrent so in pieces with a ratio coding to a bunch of different machines actually that are ratio coding ends up looking very similar to data availability sampling and kind of has the same effects and daily transmit the bits to each other and reconstruct the blocks now they'll leave out and it just kind of

continuously keeps doing that and the main takeaway from how Solana is designed is that we try to really make sure that every process that the network or the the code base and the way that it's engineered is designed such that we can scale it look more coarse but if you know two in two years we get twice as many cores per dollar that we can literally tune it and say okay we now have twice as many shreds per block or twice as twice as much compute per block and therefore the network can do twice as much stuff and that happens naturally really without any architecture changes and that's that's kind of the the main goal that we really wanted to achieve

because this is based on my experience like welcome I was there from 2003 to like 2014. every year we saw Hardware Improvement and architecture Improvement on the mobile devices and if you didn't write your software as such that you don't have to rewrite it the next year you're kind of you're kind of screwed as an engineer because you would get a device to switch as fast with twice as big cindy lanes and you'd have to like okay I have to rewrite this code now just to take advantage of this so if you really have to think ahead that everything that you're building is only going to get faster and faster and that works out pretty well and that's been like the the biggest learning experience for my in my engineer career is that you

can pick an algorithm that's very very well tuned and it can be the wrong thing because the benefits of using that algorithm become very marginal as Hardware improves and scales and the complexity of implementing it right now becomes like a major waste of time so if you can just do the very simple wave thing and just scale the course you actually are like probably 95 percent of the way there that idea the idea of proof of history as a way of synchronizing time across validators that was a very groundbreaking idea that was like a fundamental difference in how to implement consensus in a permissionless decentralized network and so that has like the core Insight of

what makes Solana so fundamentally different from all other permissionless consensus protocols is is a very important it's an important kind of core part that was actually like it's one of the pieces in the andel's law like when you're trying to optimize and this is why it's so hard I think for folks to replicate Solana in terms of both node count and latency and throughput is because uh classical consensus implementations they're very kind of Step function based you have a whole network like tenormit or even hot stuff they all have to agree on what the current block is before they go to the next one and such power still do that they don't do a round of communication and say okay we did one 20 millisecond slot between like you know 10 000

participants let's all go to the next one they actually already have a schedule and you just transmit and go and because you don't have this required step function around you can actually run the network very quickly I think it was like asynchronously but I don't know if that's the right term but you have a set of leaders so they just continue transmitting at their slot and you never stop and wait for for consensus to run and that's only possible because we have this strict concrete understanding of time I honestly think that maybe even redundant there are clock synchronization protocols that you can build they're just very very hard like I remember working with them with like the

CDMA team and it's a massive undertaking to to have reliable clock synchronization and like Google uses you know permissioned Atomic locks and that's just not something that's feasible on a permissionless network so that's the philosophy behind Solana that's very interesting before I started building Solana I was like trading an interactive brokers and owonda and like forex.com like algorithmic trading for fun none of it ever made any money but this was at a time where Flash Boys was prevalent across traditional Finance yeah and whenever I thought that I had a good algorithm my orders would be a little late I could literally see like it would take much longer for that order to make it to the market and the data

would arrive slower so I was fighting both sides of this kind of like black box and I think the fundamental goal for these open commercial systems if we want to disrupt Finance is to make it impossible for that to ever happen that the system is open anyone can participate in it it's clear how to get access to the right to to get like prioritized rights or Fair rights whoever you want to construct it and it's clear how to get access to the data without any delays or without any anyone in the middle and for that to all happen as fast as physics allowed as fast as Engineers can make it I think that's the fundamental problem that if blockchains can solve it would be very like

impactful to the rest of the world and you know would benefit a lot of people globally and that could be a building block that then you can use to disrupt you know the ad exchange monetization model on the web like make it all about three with micro payments and attention tokens and yada yada I think there's an important distinction between pure latency and atomicity especially within a single state machine perhaps you can expand on which you think is important and why yeah part of this is that well it's impossible to make the entire State Atomic because that kind of that that kind of implies a single Global right lock for the entire State and applies a very sequential slow system so you need Atomic access to State and you need that

to be guaranteed because it's really really hard to build software that acts on you know remote state that isn't Atomic because you don't know what the side effects are on any of your computations so uh this idea of like being able to submit a transaction either all of it executes or it entirely fails without any side effects it's just part of what the properties of these computers they need to have otherwise I think it's pretty impossible to uh to write Reliable Software for them you just can't build any kind of reliable logic or definitely not Financial you might be able to build like things that are like eventually consistent systems but that's I think a different kind of

software so that there's always kind of that tension between making the system Atomic and performant because if you guarantee this optimicity means that you are picking a particular writer in any given moment globally to that particular part of the state and that resolution requires you know like for you to have a single sequencer and something to linearize those events and that creates points of where you can extract value and attack the fairness of the system there's a lot of trade-offs there and solving them I think is it's a really really really hard problem that we've seen not just Solana face those but ethereum in uh with flashbots and things like that one of the things that

is often debated especially in the ethereum community is the notion of verifiability of execution and at least from the perspective of the ethereum community is very important for end users or people who don't have extremely powerful machines to be able to verify the activity in the network I'm curious what your view is I think the end game for both systems is very similar if you look at the end of the ethereum road map and you have like either dank charging subcommittees or shards or whatever you want to call them the idea there is that the overall total Network bandwidth is greater than any individual node so the network is already Computing or like processing more jobs that any individual node was configured for you have to take

into account the Safety and Security considerations of such a system and you know there's protocols to post fraud proofs and things like that in data availabilities like sampling schemes all those are actually applicable to Solana as well so if you take one step back it doesn't actually look that much different you have a system like a black box that creates so much bandwidth that it's not very practical for a random user that's not a sophisticated user to go verify it so they're going to rely on the sampling techniques to guarantee that the data is there and you know like

the very robust gossip networks they guarantee that like notes cannot be eclipsed and they can propagate fraud proofs to all the clients and stuff like that and that that that's the same kind of guarantees between salon and ethereum the main difference I think is ethereum is very much constrained by Bitcoin and this idea of being Global money and really really holding on to that narrative trying to compete with Bitcoin in a city of store value and I think there is some narrative reasons there to have very small nodes that use our camp even if they're partially participating in the network they can they can still do that versus having the network all run by professionals that kind of makes

sense you know honest honestly I think it like it's a fair thing to optimize for like if you don't care about execution all you care about is settlement why not minimize the no requirements to their lowest possible and allow people to partially participate in the activities of the network I don't think that creates a system that's any more trust minimized or secure at the end of the day for the vast majority of the world like at the end of the day you're still going to rely on data availability sampling deaths and fraud proofs and the user to validate that the chain has done something wrong only needs to execute a bunch of signatures from the majority of the chain I'm Solana that single transaction describes all the individual pieces of state that that transaction

touches and that actually runs on any device like a browser in a mobile phone can easily execute that one transaction that the majority side that the invalid State transaction then because everything is specified ahead of time it's a lot of it's actually even easier to construct consultant than but like you VM or any any smart contract can touch any state and randomly jump between them during execution in some way that's almost simpler but I think from a very high level at the end of the day users are going to rely on this Das plus Rob proofs and that's kind of like the same for for all designs at this point so I think the biggest difference or distinction between what I see as the

long-term ethereum roadmap at the long-term Solana roadmap is around zero knowledge proofs and and validity proof specifically as opposed to just fraud proofs and the idea that even if a user does have to have some level of trust in a centralized RPC and throughput reaches the point to where no individual user perhaps even in a data center can run enough Hardware to fully verify the chain the idea that someone is creating a proof or you have aggregate proving together that is then distributed to individual users who can verify it and instead of putting your trust in crypto economic mechanism you're putting your trust in the the ZK proof systems and I remember when we had discussed this a little bit when we were at crypto startup school and you were kind enough

to come speak you seem to think the Ezekiel EVMS were almost impossible to audit and that they would not be robust for several years and so it wasn't even really worth trying to go down that path today because perhaps it was too early I don't want to misrepresent your views but I would be curious to ask you broadly why Solana has not prioritized ZK proofs and validity proofs the same way it seems like ethereum has I think there's two challenges there the way that we have prioritized them is there's a company called light protocol that's building application specifics your knowledge proofs and these are hand optimized for the specific program code so the proving times are fast the user doesn't really notice them during their interaction with the with the chain and

you can actually compose them you can have a single Salon of transaction that calls five different CK programs so that that is like a environment Island that saves some compute resources or creates privacy for the user but that's not really verifying the entire chain and the reason why it's really really hard to verify the entire chain is that zero knowledge systems don't do a very good job with a lot of sequential State dependencies you know like the classic example is the vdf or verifiable delay function every time I talk to you know look at zero knowledge performance you know you can really see it fall apart when you try to prove a sequential shot recursive shot to 56 it's because the

that sequential State dependencies on execution really blows up the amount of constraints that the system has to have and verification and proving takes a very long time I don't know if this is the best in the industry but latest thing that I saw on Twitter is about like 60 milliseconds for 256 byte straw that's a very long time for a single Shaw instruction right when like a normal Intel chip with a native shelter 56 instructions can do like 11 million shots per second so for sequential computation classical computation is going to be necessary anyways and in an environment that's designed for execution with like lots of marketplaces you actually have a lot of sequential

dependencies you have markets that are very hot everyone's submitting data directly into just that like one pair and every and everything that's triangulating around that pair depends on that pair so like your execution like that's sequentially dependent actually blows up and that becomes a very long system to prove and Solana does not prohibit somebody running a zero knowledge prover in the recursive batch verified way against the entire computation and that just maybe happens to run once every two hours or or eight days or whatever on some gigantic machine that's fine if it's feasible but what users need is if I'm trading I need that guarantee that my information was

written to the chain but then like microseconds right milliseconds and that I'm getting the state and some guarantees on that state also within milliseconds and like that's that's where like the money is going to be made so I think we need to solve that problem and that needs to be actually uh competitive of traditional finance and if that's true then you can start working on zero knowledge and figure out how do we like provide these guarantees to to users that are basically don't want to verify the chain and don't want to rely on these events but maybe we can at least do it once every 24 hours or something like that I just think that there's two different use cases and the primary one is we gotta like actually solve the the mechanical like markets

problem first not the problem for like the long tail of users that don't want to run a note so it sounds like what you're saying is uh proofs validity proof CK proofs are excellent at settlement but they don't really help you for execution because the latency is too long and the performance is too poor and that's likely to be true and the longer term it's been true so far unless you guys disagree and that's my gut instinct simply because the more active the chain gets the more sequential kind of hot spots and state dependencies you get it's not like these perfectly parallelizable things that never talked to each other that was true I think we would be done by now it's just you have a bunch of really hard to paralyze like really hot spotty code that Spikes all

over the place well the other counter argument may just be that the curve of improvement of zero knowledge proves in particular on the proverb side is exponentially uh improving maybe five years from now 10 years from now the overhead might go down from something like a thousand X on top of just performing the computation itself to then generate a proof which is what it is today down to something that is much much more feasible especially with the help of dedicated hardware and I I would love your perspective here given that you you come from a background Hardware engineering but it could be that we live in a world where zero knowledge proofs are so good because the demand for zero knowledge proofs in the world is so high

that it drives a lot of energy towards optimizing them at every level at the algorithmic level at the code optimization level and at the hardware level to make them very very competitive to the point at which it might actually be more efficient to have one node perform the computation and generate a proof and distribute that proof to everyone else as opposed to having every node having to run the computation themselves so I think that Trend will actually work well for the program optimized zero knowledge systems because then you can look at the kinds of applications that aren't experiencing the other effect which is there's more and more stuff happening on chain and the number of constraints blows up faster than you can then you can add

hardware and this is based on my gut my feeling is that as the demand side gets bigger like as you see more and more computation happening on chain it's going to be harder and harder for zero knowledge systems to keep up in a low latency way not 100 sure that it's even feasible but I think it's really possible that you can build systems that take very very large recursive batches but you still have to run classical execution and you snapshot it like every second and then you throw like an hour's worth of compute in a big parallel form that does verification between every snapshot and that does recurses from there but that will take time and a big pile of

Hardware that's going to run behind right the actual execution of the chain I think that's the challenge it's not clear to me that ZK can't catch up unless demand flattens out like I think eventually demand will flatten out and then like obviously if Moore's Law stops we we're kind of screwed as Humanity to begin with but assuming Moore's Law continues and how it keeps improving and at some point crypto demand is saturated like Google searches per second are probably saturated at this point then you'll start seeing that flip happening I think we're just far from that the other big big difference between the two models which is related to this point is that the kind of the ethereum roll-up

Centric vision of the world is one where you you do essentially Shard computation use Shard data availability you Shard bandwidth and network activity uh and as a result it's conceivable that you could end up with a greater amount of throughput in aggregate because you can add Roll-Ups on top of a single little one almost indefinitely but of course you'll you'll suffer on latency to finality and so part of the question then is what ends up being more important is it's throughput overall throughput down the line or is it latency and it may be the case that both are important but for different applications it could be that both architectures actually fill in a niche for certain kinds of applications that

maybe require One Thing versus the other and both of them therefore continue to exist uh side by side I think actually you can do pretty good latency to finality even in the nut system if you if you take like the subcommittee latency as the subcommittee finality as as good enough that really depends on like the overall performance how often does the subcommittee's block get rolled back maybe it's maybe it's designed such that it's never that you know could be almost possible but but it may be that you're still that you're still constrained by the throughput that you can get given that you have to generate zero knowledge proofs so that the L1 can verify them I think the problem with that environment there's like a couple of problems but I think the the main one

is that you have your roll up and your sequencer and this is where I think people are going to try to extract value is in the sequencer and the role of construction and they're going to build like a trade fire style system where you have sequencers that are more or less collocated they extract all the nav from that environment and they operate no different than uh you know Citadel or jump or broker dealer or whatever you want to call it that's routing orders and and putting stuff in the middle of them and that's fine but those systems already exist and this design doesn't actually break open that whole Monopoly what I think is better is if you had like a totally open permissionless

system where you it wasn't possible for those middlemen to actually even Place themselves to and to start extracting value that it that it was truly synchronized Global State machine even if that state machine has lower Network bandwidth then this massively sharded system it's still providing more value and it's very likely that it's actually going to be cheaper to use because the sharded approach is like creating a bunch of different small pipes and generally pricing in any given pipe is going to be based on how much capacity is left in that single pipe that overall network capacity it's very hard to build a system where all the

network bandwidth is completely shared you can try like there's designs for for Roll-Ups that will try to to put blocks and whatever is available but they're all going to be competing and bidding and it's not going to be as simple as here's one giant pipe the capacity is priced so the price is based on the remaining capacity of that one gen pipe and because it's one aggregated source of bandwidth I think it's going to get much much lower pricing and then speed and like performance at the end of the day I think at one point I I had heard you say that you don't believe that the demand for Block space is infinite yeah which which is a great point and it's it's very interesting a point with implications for how you would architect

the blockchain so I'm curious what do you think the the demand for blockchain is for Block space is at the limit it right at equilibrium once we have billions of people on web3 I mean like imagine if at Qualcomm the engineers were told the demand for cellular bandwidth is infinite code designed for Infinity they would be absurd right they're just like there's like a Target that people think is reasonable and it's like one gigabit or 10 gigabit and it's 100 times more what people are currently using and if you can design for that demand you don't have to think about every arbitrary long tail design choice you can actually look at how much Hardware you need to decide do I need to Shard or not like what's the simplest implementation and how fast is it going

to like the pricing for for deploying it is going to get to be so low that it's negligible my gut feeling is that there is some limit for 99.999 most valuable transactions will probably fit within the first 100 000 TPS it's my gut guess and a hundred thousand TPS system is actually fairly doable like today's Hardware can do it yeah Salon is soft architecture can do it and you can see that from like what fire dancer folks are posting the labs client right now I think would struggle it would require an elab experiment maybe to prove it but I think 100 000 TPS my guess is all you

need for probably the next 20 years of blockchain space well there is this effect of induced demand it is very possible that if block space is so affordable and so trivially is easy to use that the demand for blogspace will Skyrocket just because people will want to use it for all sorts of things that are maybe not the 100 000 most valuable transactions but are still transactions that people would want there's still some floor Price Right the price per transaction has to cover the bandwidth costs to every validator like egress cost or they're gonna dominate validator costs and if you have 10 000 nodes you roughly need to price the per byte usage of the network to be like 10 000 times

actually normal egress costs but that sounds expensive unless you Shard right I guess this would be the question do you do you imagine that at some point Solana hits a limit and becomes a shorter architecture down the line or do you think that the monolithic architecture is enough indefinitely the thing is the thing that like so far of people have done with sharding is that they build systems at a much lower bandwidth than Solana and Charlotte does so they hit those like kind of capacity limits and they started to get getting into like bidding wars to access the bandwidth and that greatly exceeds the egress cost already and like at 10 000 nodes of egress right just if we purely price egress I think the last time I

checked it's like one dollar per terabyte of egress like to to every validator in Solana it comes out to even like with 4X overhead something like four times ten to the minus six per transaction it's basically there is a floor price you can't use it to stream video but the price is so low that like you can use it for search like you could you could probably you could basically have every search hit the chain and then get the results from like your search engines uh as replies on Unchained I think this is actually an interesting point because we start the podcast with the question of what is the end game for blockchain scalability implying that blockchain scalability is the most

important problem and I I think Chris has used this analogy before the idea that a lot of the advances in Ai and ml over the last decade can be attributed fairly to much better hardware and that that was really what unlocked a lot of this and I think occasionally we talk about blockchain scalability as serving the same purpose that if we can just get to ten thousand a hundred thousand million TPS like finally everything will work but an interesting Counterpoint to that is today ethereum does 12 transactions per second and ethereum is still itself still does more uh throughput than any of the individual l2s at fairly high prices usually dollar transaction fees well on Solana a lot of at least the simple ball transfer transaction fees or fractions of a penny and when we talk about this usually it's

under the guise of if we get the next order to magnitude of throughput there will be a lot of new applications that we can't reason about or can't think about today and to some extent Solana has for the last couple of years been the place where you would be seeing those applications be built and yet a lot of the things built on Solana look very similar to a lot of the things built on on ethereum so I'm curious your thoughts whether greater increases and and throughput or reduced latency do unlock new applications that we can't reason about today or whether most of the things we'll want to build on a blockchain over the next 10 years will be pretty similar to the designs we've already proposed I actually think that most of the things will be pretty similar I could be surprised like I think every use case that I've had any

kind of traction has been like a Marketplace or NASA like an nft or a Marketplace for nfts and what's more interesting is how do you build business models that around that those kinds of properties in web3 for users on the data that users actually get back as part of participating in those networks and you have somehow I'll compete in growth like the big tech companies like this is like the the harder not to crack like how do you apply these new tools but I think the tools have basically already been discovered and the reason why ethereum transactions are so expensive is because the state is so valuable like ethereum represents very valuable state in this

chain and the markets of liquidity so then when you have that state available and anyone can write to it they will bid up up to the economic like opportunity costs for them to be first to write to that state and then all effectively Spike the fees and that that's what creates the valuable kind of transaction fees on ethereum for that to happen on Solano applications need to create that kind of valuable state like whether a set of teas or whatever our opportunities exist in those bits that are now for sale for anybody to to write to that that needs to happen often enough to where hot spots occur and like people start bidding up the the fees I'll give a Counterpoint here I I think

it's very easy to underestimate the creativity of a whole planet filled with creative developers and entrepreneurs and if you actually just historically if you if you look at for example the first wave of the of the web and the internet like starting in the 90s it took quite a long time in order for the main sort of drivers of interesting applications and value to really be developed and in the case of something like crypto we've only really had programmable blockchains since ethereum starting in around 2014 and something like Solana has only really existed for you know four years or something like since 2019 or whenever Solana initially launched which is not really all that much time for people to

fully explore the design space and then separate kind of related and important fact is that the number of developers in the space is extremely small still right we've got probably tens of thousands of developers who know how to write smart contracts and who really understand the promise of blockchains as computers so I would actually bet that we are very early on in the development of interesting ideas for what it is that you can do on top of a blockchain and the fact that you have this general purpose computer that can now support programs that make commitments about how they behave in the future the design space that that creates is so broad that that I would bet that we will be very surprised for what kinds of things people build and they likely won't just

be things that have to do with trading or marketplaces or Finance it could be that they take the shape of things like shared data structures that are very valuable but play a role that's not Financial in nature so a good example of this could be a decentralized social network where the social graph is on chain as a public good which enables all sorts of other entrepreneurs and people building technology to build on top of it and that graph because it's on a blockchain and because it's open and universally accessible to all of those developers becomes a very valuable piece of state that a blockchain maintains uh and in that world you could imagine people wanting to issue a very large number of transactions for any sort of

reason like to update this data structure in real time and if those transactions are cheap enough I imagine developers will figure out a way to leverage them in the way that they have historically like whenever computers get faster developers figure out a way to use that additional computational capacity for some improvement to their application it's never the case that we have enough computational power people always want more and I think the same will happen with blockchain computers as well and there will be no limit maybe it's not infinite but I think that the ceiling on demand for Block space must be much much higher than what we imagine it is like the Counterpoint is that the internet use cases were actually discovered pretty early on uh search and even social graphs and e-commerce were

like pretty early on I think in in the 90s there were things like for example uh ride sharing was something that would have been very difficult to predict the form that searched actually in the ended up taking would have been very difficult to predict I mean social networking broadly instead of streaming of videos and things like that would have also been inconceivable at the beginning and I think like same here we can think of some of the applications that people might build on top of a blockchain but but some of them feel impossible given the constraints the infrastructure constraints today but we could run through a list of the likely things that will likely become very big down the line once those constraints are lifted and once more people come into the space to build them so I think it's actually quite analogous and we might just be

surprised by by how big it all gets if we just let it run there's an interesting card game called dot bomb that actually Chris mentioned like five years ago he ended up sending me a picture of it after the first time we met but the the point dot bomb is a card game from 2004 where the goal is to lose money as slow as possible you can't actually win or ever make money it's you're running a bunch of different startups with internet ideas from from the 90s and without exception every idea that was supposedly a terrible idea including magic internet money online grocery delivery and online pet stores became at least a billion dollar business sometime after 2010. and so I do think a lot of these ideas that start out potentially being very bad or failing in their initial uh

implementation do come around and become very well adopted in in the future it's very interesting that all those ideas actually caught a bit of fire in the 90s that they're all plausible people could all imagine them and like that's actually a pretty decent signal that people want that problem to be solved but they're willing to pay money for it just the intros in there so the question then is if it's not scalability that's that's the blocker is it just cultural acceptance of blockchains is it privacy is it user experience audience what do you think Anatoly is the big unlock that will take adoption from where it is today to you know internet level mainstream this is kind of colored by my experience of the internet itself I

remember being on the web and going bulletin boards and all this other stuff and there was like this shift where at least I went to college and got an email address and everybody working had an email address and they would start receiving Links of like memes and funny things and like they were constantly online and then they were constantly on the web and you saw like web ux got a little better like actually Hotmail started working on like Facebook took off and because of that people had a mental shift where they understood what the web was I think initially it's just very very hard for people to even grok what a URL bar is what a URL is what does it mean to even click on something go to a server like I have to teach my parents that you know and we have the

same issue with self-custody and for people to actually understand what those keys mean what that seat phrase means what does it mean to own a wallet and transact with it there's a mental shift that needs to happen it is slowly happening I think with everyone that eventually buys crypto and deposits it on their own self-custody wallet they get it once they have that experience but like how many people have had that experience so far you know like it's not a very large number on that note you guys built a phone perhaps you can tell us the inspiration for building a phone and how you think the rollout has gone so far yeah well because of my experience at qualcommagne was a very kind of constrained problem that we could solve

and it was gonna not take the entire company to Pivot into phone making business it was literally like we need to hire these five people I'm very close friends with like two of them they've all worked at Google and Qualcomm they know exactly how to go do this and we can see what happens so it was uh an opportunity that had low marginal cost to us as a company and that seemed like a big swing that could potentially change the shape of crypto or like the mobile industry and and see what happens so it seemed like a worthwhile thing to take we built a device with an oem partner they were planning to release the hind phone and when we worked with them to add crypto specific features in it and we had awesome reviews from folks

and developers find it like as a alternative app store as a pretty compelling thing we're still not sure whether there is again macro conditions Plus is there an application and crypto that is so compelling that people are willing to switch from iOS to Android some people are but not like massive numbers of people yet it's very very hard to launch a device basically every device that has been launched outside of like Samsung or or apple have failed like people would be surprised to know I don't think the pixel sells even a million devices a year and that's Google right it's very very hard to compete and the reason for that is because the

pipelines have been so optimized for manufacturing those devices with both Samsung and Apple is that any new startup that tries to build a device they're going to be late on the hardware by the time they launch because both Samsung and Apple will do a remission and how quickly they can do it it's just very very hard to compete so you need to have something that is like a religious reason for people to switch maybe crypto is that reason we haven't we haven't proven that out yet but we haven't like disprovened it yet either like we still haven't seen the breakout use case where self-castity is like the key feature that people need and they're willing to change their behaviors but what's cool is the attack is there like there's other oems that can go enable it and I think we started seeing you know at

least Google changed their nft policies after stock has announced so if we had anything to do with that decision I'll take 100 credits you're one of the few Founders who's actually both build hardware and built a decentralized network building a decentralized protocol our network is often compared to building Hardware because of how complicated it can be do you think this analogy holds and makes sense what is the same what is different Hardware's Hardware we're still playing on easy mode it's just everything goes wrong like we've had issues like they have like qualca where like I mean it's a tape out that goes bad costs the company like tens of millions of dollars per day like it's it could be very catastrophic you have like bugs that like break subsystems that you

discover like six months like after everything's done and you have to like go jump through like crazy hoops and software to fix them it's it's much much harder like I I think just the stakes are much higher at a at like at a big ship company then something like software where you can still quickly go discover you can do you know 24-hour patch is still doable in software which makes it a lot easier it's nice to hear it's not as hard as we think I'm curious whether it's faster to change Solana or faster to change a core part of Qualcomm architecture faster to chase Solana for sure okay we're talking about how to bring the next million billion developers into crypto Solana has done an excellent job I think of

building uh its own community and has a really strong one I'm curious some of the approaches that you guys took as you were thinking about not just building the company but the ecosystem so there was a let's say a bit of luck involved there we started Solana labs in 2018 and this was at the tail end of the previous cycle and a lot of our competitors actually raised multiples more than than we did and we had a small team that was a bunch of ex-welcome folks that have worked for almost a decade with and um we just didn't have the funds to go like build and optimize EDM so we we built a runtime that we thought could demonstrate this key feature that there's such a thing as a very scalable

blockchain that's not limited number of nodes or bad weather latency we really wanted to hit all three of those and if we were to also try to go use CVM if we used vanilla Gap it would have been very slow and all our improvements would not have been even noticeable do you have to spend a lot of engineering time rewriting evm and like making it work so we focused on building this very very fast Network and nothing else and we when we launched we had a very Bare Bones Explorer command line wallet but the network goes fast and this is what the the hook was for for the developers because there weren't any other

alternative fast cheat networks at the time or programmable none that could give you that kind of speed and latency and throughput and the fact that it was rust actually ended up being kind of the score thing that allowed a developer Community to flourish because a lot of folks couldn't copy and paste solidity code they had to build everything from scratch and that building process is actually like how you get onboarded as an engineer like if you can build The Primitives that you're used to in stack a and stack B you could have learned stack B inside out and you can if you start appreciating getting the trade-offs and you become its Champion right you can then go be the religious

salad that says you know like this particular stack is better than this other one like writing those programs is like writing a kernel driver the spec was very similar and that was like something that I didn't anticipate and probably would have made the mistake of trying to build a VM compatibility had we had more funding the fact that we had limited Runway limited engineering time that was the forcing function that forced us to prioritize only for the key important thing which was performance of this one state machine my intuition was that like if we can unblock developers with that one thing give a very very fast cheap Network that they can't get anywhere else they can unblock themselves they can build all the stuff

that we're missing over the weekend or whatever and that actually happened and that was again surprising and and awesome that it did I'm not sure if it would have happened if the timing like the macro environment wasn't right like this was 2020s when we launched uh literally we announced March 12th uh then March 16th the stock market and then crypto markets collapsed by 70 and then three days later we were alive I'm so glad that we were alive three days after that collapse and not three days before like I think that timing those three days might have saved us that's a very helpful description of building through a church of winter another important factor here is is how we win

the developers so this is something that puts kind of a counterintuitive to understand is that you have to go to Glass to build your first salon and program and that requires people to actually put in the time and we call the chewing glass doves that overcome the chewing glass part not all of them but enough once enough of them do they will build the libraries and tools that kind of make it easier for the next developer and that is actually kind of a work of Pride for those developers to do that it becomes a library that they share or something like that and kind of it's just a natural thing that Engineers do and then open source ecosystems so you know software naturally scales I think this is something that we really wanted

to let the developer Community go and build and kind of chew glass and then turn it into something useful for other devs without us being super opinionated knitted about it because that really allows those folks to own it and really makes them feel like they have true ownership of the ecosystem so we try to work on problems that they couldn't solve over the weekend which are like these long protocol problems I think that's kind of like where that ethos comes from like you're willing to chew glass because you're getting something in return out of it you get ownership of the ecosystem and we're able to focus on making the you know the the protocol cheaper and faster and more reliable I'm sure is what your thoughts are on the

developer experience and and the role that programming languages Play Down the Line as as the space gains more mainstream adoption part of it is that it's quite difficult to become onboarded onto the space and to learn how to use the tools and to learn how to think in this new paradigm programming languages might have a big role to play on that front because smart contract security has become such an important part of what an engineer in the space has to do and the stakes are so big that it's it's just a very high barrier for anyone to cross and ideally we will end up in a world where the programming language helps you much more than it currently does through tools like maybe formal verification but just maybe maybe more help from the compiler more help from

automated tools that give you a sense for your whether your code has any chance of being correct in bug free yeah formal verification is I think going to be a must for any kind of D5 application and this is where I think a lot of the Innovation happens is the execution side of D5 like building new new marketplaces and those are the things that are under the most threat of hacks and those are the things that really really need formal verification and tools like that and obviously artists and all these other things I think there's a lot of other applications that kind of quickly Converge on a single node implementation that becomes kind of trusted through Lindy effects once you kind of establish

a single kind of canonical thing for a class of problems that's actually easier than a new startup that's building a new D5 protocol that has to go take out a bunch of implementation risks because no one has coded that before and then get people to trust that code and put money at risk inside that protocol and that's where you need all the tools like I don't know llms that can do unit testing and fuzzing I think could be really helpful formal verification for sure better compilers like move I think is a pretty good language that's been developed by the ex-libra folks but there's also like rust itself like is getting

um there's like two different Brewers that folks have been working at like Connie and brusty that are getting better and better with with every year I completely agree with that it feels to me like the world of programming is changing in a very interesting way because it used to be that most programming was just a traditional kind of imperative programming that people do I think the things that are very javascript-like and are very wet too and the model is you you write some code and it's likely Incorrect and it breaks and then you fix it and it's fine because the stakes are low but increasingly there are more and more applications that are mission critical and for those applications you need a very different programming model one that gives you greater guarantees that the code that you have written is actually correct but then on the other side it's also it's

also interesting because there's another kind of programming that's emerging which is machine learning like using data to synthesize programs that then do interesting things especially when the stakes are not nearly as high and both of these two things are kind of eating the original form of imperative programming from either side and there will be less and less just normal vanilla JavaScript code in the world there will be more code that's written by Machine learning algorithms from data and then there will be more code that's written through much more formal techniques that look more like math and formal verification for the things that are more Mission and life critical and crypto very much follows in the bladder category but it may also benefit from machine learning tools to to be able to sort of find bugs and be able to debug

or sort of get to the point where you have a better ability to determine whether a program has any vulnerabilities yeah I mean I could even imagine at some point prover optimize languages like or actor Idris being the language that you write your smart contracts in and then you tell and llm they'll translate this to solidity or Solana I don't care and then you run the approver against implementation that could be very powerful and seems now feasible like I I probably wouldn't have believed it two years ago but yeah there was a very much step function that happened with gpt4 I love that idea you can use an llm to generate the spec for a program that satisfies some formal

verification tools requirements then you can ask the same llm to generate the program itself and then you can run the formal verification tool on this pack and the program and see if it actually satisfies the spec and if it doesn't it'll give you an error which you can feed back into the llam and tell it to try again then you can do this recursively until eventually the yellow lamp with the help of a formal verification tool generates what's hopefully a verifiable formally verified program and you can describe the program in English which is amazing yeah hopefully the complexity of this pack isn't so large that a human can't understand that's right exactly because then the spec might have bugs that's the issue with normal verification yeah we're talking about how to build called

a strong ecosystem in public but one of the other things that at least to me was remarkable at the time a lot of blockchains would be launched and almost immediately decentralized to the extent that the core team stopped participating in discussions on on forums or stopped trying to help other partners participate and you guys seem to take almost from the network launch BD and go to market very seriously and I think that was probably a pretty big advantage in building the the Solana ecosystem this is like a apology quote so he said that decentralization is not the absence of leadership it's the abundance of leadership like I was like a Linux nerd in high school so I remember like how hard it was to get Linux taken seriously

in in like a Qualcomm like at a company that like actually even proposed the idea of running Linux on a mobile ship seemed ludicrous like when I when I first joined and it took a very large community-wide effort to go convinced that show that open source was meaningful and I think that's something that I felt that we needed to do as well that the network needs to be decentralized but that doesn't mean it's leaderless you actually need a lot of experts that can continuously tell people about like the benefits of using this specific Network and and its architecture and like continuously get more people on board and create more of those leaders that can be evangelists across the entire world but that doesn't

mean that everything happens Under One Roof if the network is permissionless and the code is open and anyone can contribute it and run with it it should naturally actually decentralize you should naturally see leadership spring up out of places that you never expected and that we saw happen as like folks jumped into our hackathon and like maybe they were the team that figured out how to build uh like an object framework like anchor and that became like coral and Armani's team that's building wallets and nfts and all this other thing and is as much of an influence in Solana right now as I like it probably people might delete listen to him more than me unlike a lot of the ux things for sure so that that kind of effort is

just something that we didn't plan for but we didn't plan to ever stop the goal for us is just to grow everything else around us that our voice becomes just one of many versus like we need to become quiet and exit or something like that one of the criticisms that we used to get is that we don't do enough PD and there's kind of uh this trend in crypto that when you bootstrap an ecosystem you give a bunch of Grants to big branded companies to go build products and stuff and that sometimes works but it's very rare and like internally like we've almost had the belief that we should be actually getting paid if we

like if we work with a big company they should pay us at least costs to go build an integration for them because that's at least a signal that somebody internally is desperate to solve a problem for that company right and they want to go to market as fast as they can and they're willing to pay us to do that and that means that there's a very strong internal driver to launch something at that big company and like get to pmf or product look at fit and this is why I think Solana probably has done a lot fewer Big Brand announcements or Big Brand Partnerships instead we've massively focused on uh hackathons like Regional hacker houses and grounds like ecosystem building trying to connect

devs between each other trying to get them to participate into this Loop where you know they maybe their entry point is a hacker house at a city and they meet some other devs that have raised capital and Solana that that's been like the flywheel we try to get people connected around the world and connected to developers like as one-on-one as much as we can and then get them all into hackathon they all compete and that competition drives them to go build their first or maybe their second product and their second active out and then get noticed and actually like make a pitch deck go from like an idea that they pitch in a hack up out to like this is a fundable company and like go through that whole process and that's where a lot of the traditional BD folks

of like Foundation or Labs have been focused on is like that cycle trying to get as many companies started in the ecosystem because there's very few products that are onboarding Venture scalable numbers of users right now on crypto in my mind it means that we just don't have enough idea is we don't have enough Founders that are taking shots on goal and figuring out what are the business models that are actually scalable to millions of users so we need to like work in the top of the funnel you need 20 different startups I think maybe one to two of them will figure out something with pmf and that just gets them into that like next stage and that stage is just as brutal you need 20 of

those people in like the next stage before one of them breaks out and to be honest I don't think we've in crypto we've seen anyone break out of the out of that next stage yet that's been I think like the biggest challenge a related question here is is is how how do you engage the community to develop parts of the core protocol itself I mean that's like one of the trickiest balance says to strike for any blockchain ecosystem like on the one hand you can engage the community actively but then you maybe have less flexibility in the process of governance involves many more people and coordinating is hard and then on the other hand you could control things more in a more top-down way and evolve more quickly as a result but then sort of suffer in your community

engagement and we're curious how you strike that balance so generally the way that I've seen it work I'm not at the foundation anymore but while I was there well we saw people that were already actively contributing into something that they wanted to work on and then they go through a proposal process there's like an RFP and they there's a grant or whatever attached to that and that that works out because you got to find Engineers that are capable it's very similar to an interview process like when I hire somebody at Labs if they can up like push PR's into the code base within 90 days some some kind of mismatch happened right it could be that a culture doesn't align well with that person or whatever doesn't mean that

that person's bad it's just something's not working and you really want to like make sure that those events are few and far in between then the grants process kind of runs the same way it's like you find Engineers are already committing then contributing to the code base they already know how culturally to get their code merged and how to deal with the open source direction of like an open source repo grind It's a Grind like how does lightest manage all those committers somebody has to do that job and get stuff and put it together it works well when when you find people that are able to figure it out on their own and then you can give them brands or whatever to take on bigger projects and those grants are important because

a lot of these problems they're like L6 L7 engineer problems you need like one or two really smart people that are willing to work on this for you know over a year and it takes them you know six months to get up to speed and generally those folks don't want to move jobs unless they have a three to five year kind of light of sight unlike a commitment straight to work in the same project so those grants kind of all have to account for that you find really good people people that can commit code and are willing to work for like you know three to five years on this one hard thing and it's only what do you think is the best way to run decentralized governance for a protocol today for like a layer one the approach

that we took that seems to work is very much like Linux it's it's kind of moving forward by avoiding vetoes from any like from any participants it's just like it's taking the path of least videos and honestly there's a lot of participants that can veto any kind of change it's validators it's Partners like Circle or coinbase so that's just running notes I can say hey this change is bad we can't run it for X Y and Z reasons and that approach actually works really well because the core function of protocol development is to improve the protocol and that's a very Almost Boring thing we got to make the system faster more reliable use less Ram no one's going to

object to any of those changes there's not any like really major controversial changes outside of maybe like the fee Market mechanics and things like that and when those happen ideally you have we have a process called the simdi process Salama Improvement documents you post these designs and then people talk about it and there's tasks and you run data and eventually gets into master and that's three months out before it actually gets demand so there's a lot of opportunities for everyone to go look at this code and decide whether it's a good or bad idea before it even gets merged it's kind of a slow process but it's actually not that slow if you've ever worked at a job company basically fairly similar to how you'd get like a big

architectural software change into like you know at a place like Google or Qualcomm you got to talk to a lot of people you got to push it you got to communicate hated make sure all the key partner like key people that are touching that code base are comfortable with it and then like slowly kind of get it through it's harder to make big sweeping changes like I think that's the reason why the you know proof of work to POS merged took a very long time at ethereum yeah is that those changes are even if everyone agrees on them you just kind of need need to bake them it's like baking time for any like because like a lot of smart people are looking at the same thing may actually realize something that could be a bug or some

design like something that you missed you know three months later like after everyone agreed and you need time for those concerns to prop up and actually go like through those arguments three four five times before everything actually finally crosses the line some tactical advice on how to hire contributors for the labs entity would would be great how do you think about hiring people to work internally yeah engineering wise which to have a pretty high bar at least we hire fairly senior people and this is what I would recommend to a lot of startups do the way that I hired is early on is I would be working on something and I would know exactly how to do it and I would hand it off to the new hire and tell them this is how I would do it and it'll expect him to either finish it without 90 days

or surpass me like tell me okay you're wrong you should actually be doing X Y and Z and the reason why I did that is because I could actually evaluate them during the interview like I could test them hey this is not a problem I'm working on I need somebody to take it over so then I can go do the unknown thing and it's very hard to do that in a startup if you're a CEO and you like give somebody an unknown problem you don't know if they come back hey I couldn't do this or not whether they're a good hire or not we we have a several function and we added it after kind of the ecosystem grew a bit think of it as a PM for the ecosystem and you need the ecosystem to grow a bit before you actually have a bad role to fill and we hired one when I

was spending too much time answering questions and being undiscored you know until two in the morning helping unblock devs as well I was like okay somebody else could be doing this I know I know what this job is now so I always thought of hiring in those early days of scaling specifically myself which is kind of a very selfish thing to think of but it's weirdly like I think the right thing to do for the for a CEO I think there's an interesting question when you're building a community you can build a community around an idea but a lot of communities form around a person or a set of people and you were talking earlier about how decentralization is not a lack of leadership but an abundance of leadership and at least from what seems to have happened over the last decade in crypto every

successful project has a messiah or a profit so to speak is spreading the good word and ever and and especially the big ecosystems have at least one of these people I'm curious the direction of causality here whether you think a successful project makes the original contributors profits or whether the original profits of Messiahs make the project successful I I think it's a successful project forces itself um to somebody that doesn't like Fall Apart under the pressure that's what happens it's very very hard I think to create that artificially you need to have something that is catching some pmf some fire and everyone's pouring in you know

I happen to be in the place that I needed to organize all these folks and contributors of like tell them what to do right like just at least suggest what they should be working on that's how it starts and I think it's the idea that drives it and then eventually an engineer can contribute maybe like I'm a really good you know principal engineer whatever and I can drive like big projects but I quickly became just one of many and like my contribution as like a single engineer become smaller and smaller marginally less important so maybe it's not the person but it's the idea of the person and Satoshi epitomizes this well that even more than a decade after they stop actively contributing to bitcoin people

still Revere Satoshi as almost a god-like figure and I do think this is interesting especially relative to the fact that with the exception of Satoshi almost all project Founders are actually just one person and and docs uh and I I'm interested in your sort of experience going through this I think it's actually becoming less so as time goes on there's kind of a spike in that happening and that it wasn't something that I was used to but and as people got to know me and started contributing to the code base and everything else and they kind of started growing on its own to normal levels maybe you're right that it decreases it decreases over time but I I also think that you're giving

yourself far less credit for galvanizing a movement from the very beginning just like coming up with the idea of proof of history and then evangelizing that idea catalyzing activity from the initial group of developers getting funding initially to start building all of this spreading the narrative making making Solana a publicly known uh idea for how it is that you can build a decentralized blockchain and ecosystem I think all of that is is essential in the early days and sort of you and vitalik and other kind of uh Messiahs or prophets for the construction ecosystems actually play a role as a Satoshi played that for for Bitcoin in the early days and I honestly

think that Bitcoin is probably the exception and it can only really work in that way because it was the only game in town it was the first and nothing else in the space really existed and now I think in a world where the space is much bigger and there are many other competing projects that draw people's attention I think that the sort of the leadership of a set of Visionaries at the very start is important but of course as the as the ecosystem becomes more decentralized that does become less less of a factor and and I think then the ideas themselves are what Drive the activity maybe the answer is you just have to be very humble to be a messiah and anyone could stand up and say it's because of me is disqualifying themselves exactly so we'll move to a

bit of a lightning round now totally how important do you think privacy will be for blockchains in the future I think there will be kind of like a shift like if we had with SSL where it was a few forward-thinking people using it and then all of a sudden I think a big payments company or whoever will adopt it it'll become the standard I think there'll be a shift it's just I'm not sure one it'll take it'll take a break up a little bit I think you needed to become a feature that if you don't have it you're just not able to compete we're just not quite at that level of maturity in the market but I think we will get there you know once we get like 20 30 40 million people all using blockchains on a daily basis

every Merchant in the world needs p l privacy that's just like the bare minimum well we get privacy through cryptography or through trusted Hardware I think cryptography like I think cryptography is like fast enough and good enough we don't need sgx like already like we had one of the been lucky to hire one of the tamponade students and he's built bulletproofs that are already fast enough unscalable enough to for us to handle privacy for like the world for payments like absolutely cryptography is is good enough already so the next next lightning round question what are the implications of a Solana style architecture for Mev the does the leader have too much power to reorder transactions this is the the big question so the way that we originally

designed this was that you can actually schedule more than one leader per slot and the theory there is that if we shrink this loss to be as close to speed of light around the world as possible which is like roughly 120 milliseconds and you have multiple of these that you you end up with something that is like a discrete batch time auction that's happening every 120 milliseconds around the world and the user can pick the block producer out of the ones that are all available either one that's geographically closest or the one that is offering the best rebate and we believe that that's going to be basically the best you can do like theoretically that's probably the closest most efficient way for finance

to run it's either a pick latency and it sent to the closest block producer or I pick the best rebate because I'm doing that latency dollar trade-off that was the theory we haven't tested multiple leaders per slot yet but we're getting close to where I think that might be doable maybe next year and then we'll see if if that is robust from reliability perspective but I think once you have it you really do have a very A system that basically forces Mev competition to be at its optimal to where the amount of extraction is as small as possible what is your favorite systems level optimization in the Solana architecture um turbine this I would say This was um how we propagate blocks uh and this

is was an idea that we had early on and it was uh one of the things that we really needed to we wanted to prove as early as possible to demonstrate that that each node in the system that we can scale them the Network's a very large number of nodes and we can transmit a lot of data but the amount of egress that each node has to share like the amount of load egress that each node has to take on is fixed that there's a cap at how much egress they have to do so that's uh if you want to think of it from a high level every leader when they create a block they chop it up into shruts and they create a ratio codes for those shrouds and then they transmit one shred to a single node and then that node sprays it to the rest

of the network so because it's all intermix for the ratio coding use somebody that's receiving that that data you have a very high reliability guarantees on that data because your number of nodes that are going to propagate the data to use very very large and the probability of 50 of them failing is very very small it basically requires fifty percent of the network to be down and then fifty percent of the network is down you're not going to be making progress anyways on on anything so that was a very cool optimization and the way it's designed it's very low overhead and very performant and even after we had like the fire dancer team actually from the ground up we told them

hey if you guys find any algorithm or anything better it's a stylus will will we can definitely change the the core code to to make sure that things scale better turbine is is like pretty close to us most as efficient as possible so very happy about that one to conclude Anatoly I'm curious how you see uh crypto adoption evolving over the next 10 years in 2033 how do you imagine most users namely users today who have no idea what a blockchain even is interacting with applications deployed on Solana I think what we should see are some breakout applications and payments because there's clear advantages to using crypto like usdc and digital

dollars for payments especially interest paying dollars and all these other things you can do programmable one is way way better than Current financial system I think once regulation you know Congress passes a couple bills I think payments is going to be a breakout use case once we have payments the other side of it I think that that's also growing are like social apps so applications that kind of bridge nfts and social either messaging apps or social graph applications those are slowly growing right now I feel like they're primed to take off and actually get to real sizable numbers I think between both of those you should start seeing use cases with 10 20 30 million

people once you get to that stage it's actually possible to iterate on what actually people want out of D5 and give them those products what's missing for everything else to break out is we need an active user base of 20 30 million people that you can sell products to that will actually use them not for the token but for the actual usability like utility of that product so that's my guess payments and then social super apps maybe something like that what advice would you give to builders in the space or Builders outside of it who are maybe curious about crypto and web3 yeah I would say that this is like the best time to do it like it's when the markets are a little bit like macro suppressed it's before

any big cycle there's not a lot of noise you can actually go focus on product Market fit and anything that you actually discover with regards to product Market fit is what's going to accelerate massively but macro turns around and you want to do that before that happens if you want to work in AI like I don't think people should be afraid to go start AI companies right now or crypto companies or whatever then go try it and go go build those ideas but I would say that like folks should try to build things that are bigger ideas instead of copying something that exists the best analogy that I've ever heard for this is that when people discovered cement everyone was focusing and building bricks the cement and then one person was like I can make a

skyscraper and they figure it out or combine rebar and the building that like nobody could have imagined was possible so these are the new tools that we have are basically cements you gotta you gotta go figure out what that skyscraper is and go build up I honestly think that was a great way to end that's a great a great answer and it's only and uh it's always inspiring to hear you talk about uh being a Founder so thank you so much for being with us here today and uh helping light the way for for future Founders thank you for having me thank you so much guys thank you thank you for listening to web3 with A6 and Z you can find show notes with links to resources books or papers discussed transcripts and more at

asicsencycrypto.com this episode was produced and edited by sonal toxy that's me this episode was technically edited by our audio editor seven Morris credit also to moonshot design for the art and all thanks to support from Asics and Z crypto to follow more of our work and get updates resources from us and from others be sure to subscribe to our web 3 Weekly Newsletter you can find it on our website at asicsnzcrypto.com thank you for listening and for subscribing let's go foreign