Anatoly Yakovenko: Solana – at Breakpoint #418

Transcript

hello and welcome to epicenter the podcast where we interview crypto founders builders and thought leaders about the future of the crypto revolution my name is brian crane today we have a really special episode so i was speaking with anatoly yakuwenko the founder and ceo of solana labs um about solana so of course solana has had an absolutely uh phenomenal uh you know phenomenal growth this year has gone up tremendously in market cap and activity and we recorded this episode at the very first um big solano conference break point which took place in lisbon there was over 2 000 people there and it was

really very vibrant event so i'm excited about that conversation and i'm sure you'll enjoy it as well now before we go to the conversation with anatoly i want to talk briefly about the sponsors for this week and that's cal swap now dexes are great but they're vulnerable to problems like meb we talked about this a little bit in the podcast fail transactions and high gas costs on ethereum cowshop tackles these issues head on and offers a new kind of trading experience it's built by gnosis and cow swap is a meta dex aggregator that's right it's a dex aggregator aggregator it fights meb by matching overlapping orders directly and if no overlapping

orders are found the trades are settled on a variety of under lying amms and it chooses the best price wherever it finds it so give cows up a try and enjoy perks like no gas fees paid on failed transactions optimized transaction management for multi-six and dollars and some other fun and entertaining surprises so head over to cowswap.exchange and start swapping today and now with that let's go to a conversation with anatoly welcome to epicenter the podcast where we interview you know crypto founders and talk about you know the future of decentralized technology so this is episode 418 and i'm here today with anatoly jacobenko he's the founder and

ceo of solana um yeah so welcome anatoly awesome to be here yeah so we've we've done an episode before with you right so this was like 2019. uh pretty early on you had uh there was a blog post where you guys outlined sort of the seven i think it was seven uh you know eight was eight but one got dropped which right yeah but yeah it's basically like okay these are like this uh you know technological breakthrough or innovations that like solana is based on of course that was like early on and now here we're today at the solana conference breakpoint lisbon

uh it's been the last day yeah last day yeah so yeah how has has the conference been for you it's great yeah it was uh really unexpected how many people were going to show up and how much energy they were going to bring lisbon is a wonderful city so it's kind of easy to put on a good show here it's just a really nice place but um just really cool to see all the people that you don't know if their bots are not online right but you know there's people right people that love to code love to build stuff just amazing for you

what do you feel like makes the solana community what's it like i think yeah it's this idea that we're ready to eat glass i think kind of started permeating through this and this goes back to like the validator days when i was talking to you guys and like you know certus and all the other ones the the problem was that like hey we need you to go to a data center and put boxes up and everyone said that's too hard except for the for the few um yeah yeah yeah no i think that's definitely true right because i guess um so i mean of course our listeners right

will be well for aware of solana at this point right at least on a high level uh you know i think solana is i don't know five or six or something like that before i'm not sure but like somewhere someone's really high up yeah i don't care i don't think about it but maybe let's just step back a little bit and sort of start at the beginning like what was the original vision for you for solano so i'm like an operating system geek i worked on brew i worked on android and the linux kernel that's that was like the first interesting problem that i thought this is cool in tech i don't know why i thought it was cool it's kind of like a boring thing right you have hardware and you're just

writing a little shim on top to abstract the hardware to make it easy for doves but there's so many constant trade-offs that you're dealing with that it's kind of a nerd snipe right it's a rabbit hole how do you build the fastest possible system call and things like that become kind of interesting challenges so when i started solana you know i called a bunch of my friends from qualcomm was like okay this is a new operating system a new platform there's going to be applications we don't know what they look like but there's stuff running on ethereum we think that's like it was super early days with uh

2017 2018 just there was an idea of a smart contract but there were no nfts yet wallets barely worked then yeah it was just this kind of you know let's build another os but now with a different set of challenges maybe talk a little bit about this idea of like solana as an operating system like yeah the the thing that got me excited that i thought that i could make a difference was that um i never thought sharding was a viable option so again i i the reason the design is the way it is is because it's what i'm good at but i also think that it's the best

design all things considered and the if you don't have sharding you have one constrained environment like there's only one of these computers then you're dealing with effectively an embedded system and that's where you have constrained hardware and i'm good at finding those you know local minimums where you can squeeze the most performance out of out of a fixed set of you know you know memory silicon cores whatever so once you once you constrain it to like you only have this much that's all you're going to get how do you squeeze everything out that's that's kind of the problem that i'm good at and that was to me kind of the challenge so

the design and everything else that followed really tried to uh optimize for let's build the fastest possible os given the hardware as it is today and where it's going which was then still is cuda cores single instruction multiple data these massively parallel you know systems where you have um you know kind of a constraint like you know programming a gpu or avx is different than writing a bunch of code that does whatever it wants so you're you're already dealing with a little box you know yeah so how do you how do you make that little box the fundamental core part of the platform the developers target

right because like i think that was i remember you that was always where solana was kind of unique right where you have all of the other projects you know from like ethereum near you know many others were all focused on on somehow okay we have this one blockchain now it's getting full we gotta like paralyze it or like you know we've got to chart it and then we have the system that's how it scales right and you went the other way i think still so the reason for that was i looked at the data sheet of you know computers at the time and just google on the internet okay what how much how how fast can you do a signature verification how much of those can you

do in a gpu how fast can you move memory from the network card to the gpu to do these operations and they just kind of like laid it out and it looked like that with hardware you know in 2017 2018 not super expensive hardware just off-the-shelf stuff that you could find you know at fry's or amazon you could get to you could saturate one gigabit which was roughly 700 000 tps or 500 000 tps if the messages are 256 bytes so can you can you build software that like gets out of the way of the hardware so you can actually take in that many messages

figure out where the signatures are move all the data to the gpu run all the signature verification go fetch all the contract state go figure out which programs you want to call execute and dump the state transitions back into you know whatever disk or storage in theory on the data sheet you can do that in practice it's really really hard because the operating system you're running on will it has a scheduler and it decides when stuff goes you have like caches that are not uniform um you take a lock on one thing and it stalls like a thousand other things and then it becomes that that's that's the that kind of like

meat and potatoes problem where you're just looking at grafana metrics and benchmarks and heat maps that's that was my bread and butter for a decade so that that's where i thought that i could have an edge and so if you look at it today what are the bottlenecks it is like um so if you take a modern system um i think you guys ran some benchmarks but uh there's other validators like uh and buzzer and like a bunch of folks in the community that they just spin up whatever the latest and greatest they can get and on modern like epics with you know i think 128 cores or plus without going to gpus you can break 200 000 tps in a single system

so that's just this runtime fetch a transaction from a linux kernel ddp buffer do signature verification go find some state do a change in that state and write it back but all the states are parallelizable they never touch each other so in the most kind of trivial you know flip one bit unchained right flip as many bits in parallel as you can atomically uh you can scale it up without going to gpus to about 200 000 and maybe with gpus maybe 250 because the bottleneck is no longer signature verification it's memory network cards i o latency you're

dealing with very small packets which is a huge pain in the butt um hardware doesn't like small things it likes give me one giant chunk of stuff so when we outline those eight pro eight solutions it was really like eight eight challenges these are the things that we think we need to solve and this is how we think is the best best bet to solve them um and you know the one that we dropped was the kind of the least important one in a way it was storing the entire history of the ledger because of uh you saw filecoin and rwev and a bunch of other folks solved that problem and you now have robust networks that can do persistent storage forever

of history yeah yeah and and like how far do you think this one can take this like do you think solana will be capable of like you know handling sort of yeah all the crypto blockchain activity that people want to create uh like link line rate which means that as fast as packets arrive to the network card before the next one arrives you can already handle the full the full loop um i think it's doable honestly i think it's scalable to whatever the bandwidth you can connect right so if you have enough if the latest and greatest bandwidth is 20 gigabit

uh you can go to fpgas go to hardware like start doing this directly on the on the neck on the nic i think it's doable but that requires kind of a huge investment and both in terms of financially because you need to hire a bunch of engineers you need to contract out with like you know the foxconns of the world you'll need to build the reference platform that's like kind of going yeah the solana signature box so that means also like validators running like very different setups and very different hardware and it would be yeah this is kind of going to the what i think will probably be the

eventual future of the stuff if there is such a thing as a global price discovery engine like what is blockchain for i think it's for running markets right and tracking ownership but tracking ownership is settlement right and settlement's part of running a market so if you have one giant global you know state machine for running all of the world's markets for you know from nfts and sidestar atlas to you know btc usdt or whatever the most liquid pair is every every market in the world all in the same thing it really feels like it should be optimized to whatever the physics do you think are there mechanisms that

will or like you know what do you think will drive you know in a way right for this kind of progression to happen right well you what you would also want is that like the different validators and people running validators have like an economic incentive to do that right they know in a way like okay if someone increases their capacity like you know they economically benefit from that that is the the tragedy a commons problem right i honestly think it's overstated just just being here right at this conference and the people you meet that's the vast super majority of the validators and they just want to like they want to see the whole thing succeed

but the number of people taking advantage of the tragedy of commons problem in a open decentralized network in an open source community right it's like the people that you know complain about the linux box but never never tried to fix them or never try to help the dev debug them it's a small number it just can't they just get filtered out he's like why would you be doing this right there's there's plenty other ways to make money like you're doing this because you have some passion something that is like okay we're gonna decentralize the world yeah sure i agree with that right for now but still you will need to like

put a lot of investment into that right too for sure um but it's not like too hard right it's just you know the opening ceremony it talks about decentralization the importance of the behavior change to maximize censorship resistance not a technology problem it's a really like a community like needs to be aware that it's important yeah you mentioned a lot the idea of censorship resistance i'm curious like what kind of censorship are you most afraid of um it's actually not that i'm most afraid of it's the one that i wanna that i feel like will provide the most benefit so like bitcoin right is censorship

resistant but it doesn't guarantee when and um that is actually true about any system any bft system right you send a message solana like goes into fork death because there's a denial service and bgp routers get screwed up and the network can't come to finality right um is the network censorship resistant or is it being censored well if in 24 hours the validators figure out how to reroute around that failure and continue right from the last point of finality and then your message gets accepted then

it got accepted just within 24 hours and not 400 milliseconds so the network is no matter what censorship resistant as long as there's at least one validator that's out there with a valid copy of the ledger that they can tell everybody hey look this is the right this is the valid copy you can all run your tools locally to verify it everybody sign this thing right so in that sense if you don't put a time constraint on it it's censorship resistant but what people want for the benefit of actually using the stuff in practice is they want those guarantees within a very small amount of time because when i like click a button you know an audience that says i like this song i

don't want to wait 24 hours i don't want to pay an arbitrary amount i don't want to pay an unknown amount of money to get that thing uh accepted right i want to pay a very small amount of money and have that be confirmed in a very short amount of time so when you constrain that then it becomes an embedded systems problem which is fun yeah i mean of course that definitely makes sense like of course yeah you want to i want to be able to send a transaction i want to be confirmed like you know pretty much as fast as it can possible cheap you know i think solana actually does that today does it well um but that doesn't but censorship is i mean i guess what what consensus should

be right i mean of course you could have like maybe governments trying to kind of like crack down on blockchain networks and like so as long as long as there's one copy right right this is why when you look at like the validators app uh brian and all these other folks they're like okay well what are the data centers what are the asns let's map this out and that's track where stake weight's concentrated and then let's go talk to those people and get them to decentralize it more and you see that process run in the community and people are like given the information everybody's making like the right decision they're like okay my notes are all in hetzner let's move them out all we had to do was just tell people hey

look there's too many nodes and heads there yeah it's it's a it could like cause the network to stall right if tesla has a fire right and people just did it on the right so i agree with that right so of course there's definitely like a risk right around like geographic concentration of nodes or maybe if you have like a few violators that have like you know a very large amount of stake and then like maybe there's some sort of although even there right you have today i think like 19 validators right that are like the superman already so but like the censorship thing feels like a different so that is why so to me that it's the same question because you're kind of looking at the nakamoto coefficient

so you take some parameter data centers or asns and then you start assigning given whatever it is data centers you know geographic like you know jurisdictions where's the concentration of stake and node and then you identify that bottleneck that's above whatever threshold you don't like 33 percent or 15 or 5 and then you go talk to those people like hey look this this is like a worrisome thing can you guys move your nodes and people do it that's the cool part but that that again the reason why you want to do that is because that has impact on real-time censorship resistance like i don't want to i don't

want to have this like guarantee that if if i send a message and that's the thing that fails right the amazon asn goes down but now i have to wait maybe 24 hours before that message accepted what i want is when i send a message within 400 milliseconds and for a fraction of a cent it always gets accepted i mean to me that feels like sort of around the resilience of the network right a lot so it's the same thing right it's just like you start going to objectively measurable parameters what is what can we what's objectively measurable like any any one of those things you can then decide this is what we're looking

at we think the failure rate is x then we need to go and make sure that there's at least you know why number of uh these things before the network has to go through this like recovery procedure what do you think are the biggest challenges ahead barcelona i mean realistically the going to like link line rates being unbounded by the software right by the implementation it's a really hard problem it's like a really hard engineering problem um so that that's kind of the biggest challenge it seems like the salon is fast right right cheap and fast but there's 300 million cryptocurrency holders out there that's the latest number i saw

when we actually talked to the folks doing analysis of how many of those do stuff on all the chains combined it's like maybe three four million at most yeah actually people individual humans signing stuff it's a very small number there's a lot of like you know people arbing uh uniswap or whatever markets on ethereum but uh those are mostly bots and it's not 10 million people it's a very small number of people generating like half a million whatever events per day so yeah the real challenges are like the engineering ones everything else i feel

like is the it's it's like the the no never if we do that right like everyone else in the community will solve all the other problems it's just that's the that's the challenge that you can't fix over a weekend [Music] with uh the fact that our developer tooling requires you to on by hand pack and unpack data structures developer can do you know figure out how to write a library for that over spend a week or two and armani did right so that's not a problem that we actually need to solve the hard one is if it takes two years to design hardware that can actually handle

unbounded bandwidth like go up to 20 gigabits and we really believe that that's what the world needs right for the world to be fully decentralized then that's the thing that uh that's the biggest challenge that we need to tackle yeah one thing i'm also curious about because i feel like it's it's a topic that's like discussed a lot in some other kind of crypto communities and i think it's like lesser in solana is like the topic of governance and i'm curious like what do you how do you think solana governance should work i mean maybe today but also in the long term yeah like i think um [Music] i'm i'm like a i think it should have like a very clear constrained

mission that makes sense and then the only thing that the governance is doing is hey are you executing in this mission or not like it's more just like a backs back check on on the work that's being done and like uh because if you have a like like it's almost like a constitutional dictatorship or whatever right you gotta like you gotta pick what you're gonna do right and that's that's the thing that's the most important and then you invest in it over a long period of time and it's not open to question like whether you change the mission it's that are you like screwing up on execution or not

then making that big bat is the is the hard part and the bet that everyone is making i think internally at labs the foundation and for the vast majority i think for basically everyone that i've ever talked to is that that real-time censorship resistance piece that's an important thing and we should we should basically do whatever we can to maximize it okay uh but then like let's say they are like different interpretations different visions differences that arise like do you feel like so so like inflation was a great example yeah what is the optimum rate for inflation nobody knows right

and does it impact censorship resistance well if you don't have it it does if you have too much of it it does but that's again like a very wide decision thing yeah and within the community there was a proposal process everybody that wanted to submitted a pr that said it should be axed they tried to defend their reasoning for it uh there were people that wanted to not have inflation at all there's a forum post on this and then there was a validator effectively vote and that's where it got right so that that kind of like how do we pick a parameter right like that's that that feels like again like a thing that you can ask the community and have it figure it out

go through that process but do you feel like there will be a need at some point for like a sort of uh explicit on-chain governance system or like you prefer that kind of thing to happen uh well i mean validators are always an explicit on-chain governance for them to take an upgrade for them to sign a feature up a feature upgrade right is a explicit thing um because they're signing those messages with their validator keys sure that's true yeah now of course validators have that role with upgrades that's true yeah so but like for layer one that is the layer where anything important with regards to governance happens is that the validator side

i don't think you can escape it yeah i don't think you can like push it up i mean it's interesting that cosmos kind of did that and there's like a pass-through mechanism to the stakeholders yeah you can of course you can never escape that right that's always there i think that's true right i mean i guess maybe tazos was the one that we're in like further sort of trying to do that right with these automated upgrades but still but like about that you can always fork the software like run something different so to me to me it always felt like that's the heartbeat the validator community those are the people actually running it they have the most understanding of the day-to-day and that's almost like you gotta convince them to do anything

yeah if you fail to convince them you're doing something wrong right maybe you can talk a bit about uh well i'm curious about meb like what so this is becoming a big topic on ethereum i guess most importantly where and maybe for listeners who are like not too familiar with it basically it revolves around the ability of miners to like to decide on the order of transactions and then there's like economic value in that like for example there's like an arbitrage opportunity and then someone can can you know kind of so do the arbitrage make some money and then of course like if you can decide like

who which transaction does that there's a value in that so solano's blockchain and nasdaq speed and we never thought of mvv but i'm just kidding yeah i mean uh so i always thought that that that was actually the feature for for blockchain this is how they should work they should have meb and if you come from that perspective it means that you need to maximize the um almost the market for it like maximize the competition from out so how you do that is you reduce the block times to a smallest amount possible so we're at 400 milliseconds and you have a liter scheduled four times in a

row so imagine it's 200 milliseconds and the you have a switch in every block from one leader to the next and solana is actually designed where you can have more than one leader assigned to the same slot producing the same block if you look at the white paper i don't know if people got it there's this thing where you kind of mix the things and you can then tell the relative time of events um it's a b minus white paper i get it but all the ideas are there so if you have multiple block producers now in that 200 milliseconds me as a trader i'm willing to wait you know two seconds and i want to maximize the number of

bids that i see from the validators the real time guarantees in solana is that if i'm a validator and i'm assigned slot 10 then i'm guaranteed to be able to transmit a block in slot 10 and as long as no more than a you know a third of the network is actively trying to censor me then because of how turbine works because of proof of history is this forced delay function i'm have extremely high confidence i'll be able to actually submit this block and have it confirmed that means that i can then bid for this order to the user that's willing to wait until my blog therefore i can give him the highest rebate out of all the other validators

right so at this like lowest layer by minimizing the block time maximizing how often we switch maximizing how many block producers work on a single block together right we can actually create the biggest fairest most competitive market for mev and then you then you have real value creation right because what is what is mev it's actually i'm observing all the world's information every exchange out there twitter feeds etc i'm creating a model for what the future price of anything is going to be right so i am doing value creation right and predicting the future then i'm betting

you know bidding for order flow hey users i have the best model predictor this is what i think the fairest price is going to be right by the time you submit your order give me your orders i'll give you the biggest kickback yeah so i i definitely agree if you write that like short block times i mean first of all i guess mbb is obviously harder on solana than on ethereum right i think that that's already in a way i think it has been uh reduced i think solana has kind of reduced the scope of it and maybe the issue around it uh but but it is value creation if you can feed it back to the user yeah because you are effectively trying to predict the future right yeah

right yeah and and yeah i think in a way you will see it feeding back to users right because i mean first of all right if validators uh make more money with that then it's also something where they can for example say oh like you know hardware right yeah let's reduce fees exactly we do fees right exactly right maybe you have any negative fees you know which is what i mean by rebates right and you can you can do it via like a global negative fee rate right if you want to socialize it or you can do it a competitive way and i think the competition is actually a pretty good approach because you know like folks like susquehanna like jump they have their own models

right this is where they differentiate and this is where they compete and i think you know in a perfect world they're the ones that are like running alongside that they give validators some software hey run our models and like send those prices to the users and and get you know everybody get everybody wins basically and so i mean if you look at least sort of the debate around me and ethereum right there's also different types of med that people think of differently like in particular you know like let's say this is the example of arbitrary so you have like you change a change b and the prices are different and someone can make money by like bringing them in equilibrium i think that's like unequivocally good thing right because i

have his user then can we basically say i can just trade on whatever amm i don't have to worry about it because the prices should be the same right so that i think that's clearly a great thing then there are other things you know like this sandwich attacks right which are basically right i mean how does it work right so let's say user basically says okay i want to trade you know this amount of coin for this other amount of coin and i'm willing to accept you know a price up to some amount right where uh and then somebody can basically push the price of that amount take it back and and sort of i get the worst possible price i was okay with so who right so who takes a loss there right the lp's or the user

do you the trade like whoever makes the trade takes the loss right so if you have at least on the first order right so if you have a competition for user flow then everyone's going to say well i can maximize the sandwich attacks yeah and i'll give you 50 of the profit and i keep 50 right and some other validator says well i'll give you 75 and keep 25 percent yeah and then the user wins right well i mean if we can improve the stake system right i guess the yeah so the reaction could be right that yeah validators basically try to like maximize that in some way and then that profit flows back to uh there's so holders right the stakers

in the end yeah and like not entirely right because you to be able to maximize that you would need to i think give a return to the users as well there's you know the person training yeah there's some equilibrium because i think so because the user always has a choice to trade on solana or anywhere else yeah there's no there's never like a monopoly there's no modes right so you're always going to be dealing with a user that has the option to make that trade in phantom right or they can actually go to ftx yeah yeah i mean of course the other i think sort of the other maybe answer to that might also be just that well you can you can certainly design amms that are resistant to that right to

different degrees right so i think to some extent or people do that then well people will stop using that kind of amm right that i think they're actually going to be slower and have worse spreads compared to a competitive mvp market okay so the latencies are going to be worse because you're trying to like basically stop information from flowing as fast as possible you're trying to control when when events happen and force and ordering or like randomized stuff right and it's maybe perceived as more fair but i think the result is a slower kind of a slower system that's going to be more expensive right so your take is like okay it

should be exploited to the maximum and then return to the user right to the to the stakeholders whoever right i'm just thinking like practically uh i wonder how that actually flows back to who to the user well like uh what i imagine when none of this works yet is that at the wallet itself right you can have a button that says wait x x seconds to get wide wide kickback for your trade yeah yeah and maybe you always stay fastest right like there's like get fastest cheapest whatever slowest gas meter how much you pay right in this case it's

how much you want back right so wallet could basically say okay we are going to kind of like you know sell the transaction right and then uh mev is taken from that and then some of that goes back to the to the wallet and then they can distribute that to the users yeah yeah i think it's going to be very interesting to see all of this yeah there's so many pieces there it's such a yeah that is like a whole financial system just for trading pictures of of dogs [Laughter] the entire world's financial system is going to sit on top of a collection of a

monkey nfts yeah yeah but that in a way is kind of the amazing thing about like what's kind of happening in krypton is that you have to like speculative markets around everything and then people like optimize things you know to to you know make it more efficient make more money or something like that and then i think it drives forward all of the technology alone yeah so this is where like you know if we do the really hard stuff make this as close to physics as possible then all this stuff becomes more efficient and more value can be returned to the world and this has kind of been my my main focus and my biggest gripe with this idea of ultrasound money

is that that meme is not it is not maximizing the amount of value returned to the users it's not minimizing the value extraction like these networks vitalik's idea was that this is a public good right right this is a public open to the world's like infrastructure it should be minimally extractive right it should minimize the amount of value it takes so to be as close to the physics the you know maximum capacity cheapest cost yeah nothing else matters no no other memes doesn't matter if it's ultrasound you know money or not or it's store of value

it's like is it actually generating like a positive not some good for the world and is it taking the least amount of value to do so yeah yeah i mean it is interesting you know how there has been i think a sort of shift in the theorem narrative where it was uh ether was photo it was gas right in the beginning right it's just okay it's just to pay for transactions right and then i think there's definitely has been a sort of shift towards even even this eip1559 right it's like burning the fees because people are willing to pay for that now i think they've been doing it

for so long that maybe the mindset has shifted that it's acceptable like in my mind as an engineer like no this is wrong the system is wrong you gotta like throw it all away if that's what it's gonna adopt i mean i think today right the theorem is just unusable right i mean this is crazy that you have like hundreds of dollars of yeah you have to pay for a transaction right so that's it actually becomes kind of what bitcoin actually even bitcoin's cheaper now right bitcoin consent pretty easily now but i think always uh one of the ways that uh bitcoiners were sort of dealing with

okay the system's really not scaling and it's clearly not gonna scale unless like you you radically change something was like okay but then it becomes this kind of settlement layer and bitcoin transactions can be very expensive and you have like layer two and things and everyone can kind of hold bitcoin even if they don't directly do so and now i feel like sour value is like at least some something you can look at and be like okay i don't really care about how long it takes for my transaction to settle yeah right as long as within some like two weeks it does right so within two weeks i could find a cheap enough time to do so so so what are your thoughts on like roll-ups and the kind of ethereum scaling direction um they're not as uh they don't so they

don't optimize the censorship resistance piece right they actually just reduce they batch stuff so there's some batching optimizations so they may reduce some cost to users but the majority of those costs are simply because evm is too inefficient to execute right so you if you get rid of evm execution that's what effectively a rollup does right is it delays executing a vm until um like you need to run a fraud proof but if you don't use vm use like bpf x86 any other like efficient virtual machine um it's not a there's no real savings there

so you say it doesn't reduce censorship resistance because oh it doesn't maximize your censorship resistance because um it only optimizes this one thing that's extremely slow but you don't need to do anyways and uh if you let's say you reduce fees to users but you still have the settlement layer that extremely expensive to use so that thing let's say that thing is generating 80 million dollars a day in revenue is that is it providing censorship resistance decentralization at the cost of 80 million dollars a day no it's not if you look add up all the

costs of running an ethereum node and add up all the nodes and how it's sliced right in terms of uh censorship resistance like real time it's uh i think like 4 000 x off you mean that the fees collected right by are much higher than the cost to run so so it is not a minimally extractive public good right it's just a thing that is has like captive markets for the services it's providing and it's actually kind of like squeezing yeah right you could you know reduce some of those costs to users and even if you can that allows you to increase the number of users that will use the system

but it's still not a efficient system for the value it's creating for the world yeah no i think that's true right right i think there's obviously like is an extent where you okay you have this ethereum ecosystem and there's a lot of interoperability all the things there so there's a lot of like value to like being in that ecosystem and then but at the moment it's sort of like okay because there's this capacity constrained right and just candle everything right it tracks extracts the body can to the ethereum the values in the cryptography yeah and it's in the hands of the users and a user that is holding a bat token

that now holds it from a suck p to 56k one key in metamask and uh mt5519 key and phantom for the same token they're still part of that bat community that likes brave yeah yeah so that just you know got uh that token went through mobile right there's some risk there to use a bridge always but uh you can analyze that risk you can minimize it and now you've reduced significantly reduced fees and now you can kind of minimizing the amount of value that the network itself extracts from that community do you think it's essential or do you think yeah i wanted to what extent is it just

it's just a crucial thing that you sort of stay ahead that sort of the network you know throughput the capacity of the network stays uh ahead of the actual load but like it is a big enough factor yeah because you're like kind of like uh costs go up when there's like a burst really like most things like what any kind of web scale system you're paying usually for the worst case because you have like a rush right and you don't want to be able to handle it so just like in those systems you're typically going to have to be way over capacity to handle a extremely high spike and when you get that spike persistently that's when fees need to go

up to force whatever is effectively spamming the network to back up yeah well let's talk a bit about sort of you know we are still at you you were saying before right i think okay you were saying maybe 300 million accounts so like pets.com stockholders yeah 300 million actual users.com you know three yeah so where do you think we are in terms of like mainstream adoption do you think we are going to see the kind of like web 2 applications uh yeah like what time frame do you think those are

if if next year we were looking at like phantom adding a million users a month um that would feel like it seems real right that seems possible yeah even a million a week it's like seems possible but that would be like oh man the weird this is it we've we've hit real world adoption um yeah that that's kind of like i don't know when that happens could be next year could be two three years from now it's hard to say right yeah yeah and next year feels just as likely as two years or three years

yeah i mean probably a lot depends on the market right because if things go up then interest just goes up so much i guess it's two things maybe right like i think the crypto market is like a huge factor yeah it just drives so much interest and then maybe another thing is like okay people are actually putting out applications that are just really great and they're unique because they leverage crypto and then people who just want to use the application they don't care about the market and the prices and then they kind of switch that way so what we see now i guess that's different in the in this cycle is that you're not investing in ideas or white papers it's in in like actual products

you almost don't even i haven't read any white papers a lot in like the last six months it's like give me the mvp that works and i kind of look at it the design the ux and sure there's like sim innovation around the risk engine or the the thing that the you know the new fancy d5 market or whatever right and i almost don't care [Laughter] that part is like not as important as this there's like good ux and it feels like humans are going to use it yeah so the the thing that makes me sort of like pause a little bit here is because i remember you know in 2013 there was this bull market and then we had 2017-18

in this bull market and personally like in each of those past bull markets i was pretty convinced that like that's it there's like mainstream adoptions coming right because it's obvious that this is like we're gonna be superior in so many ways and then like oh people understand that they're like adopted and why would that ever stop right like why would it stop until it's actually like there and then of course it kind of did in both times so i'm i'm wondering is it is it the same now is it also like okay we have we in this bubble have this idea because because it's obviously in the end is going to happen right in the end you know it's all going to be on crypto and like

i don't know i think it's too long yeah yeah i think so yeah and yeah it's kind of scary so you think it's it's like full it's not gonna like be easy because you still need to build applications you need to like iterate on the parts where you have to see drop-off that users like lose interest yeah right and you need to build really good applications that people want to use that have some some true like reason to be cryptographic right to have cryptography yeah not just like we stuck crypto inside you know facebook yeah

like it has to like actually have meaningful reason right um and that's that's when i think you start seeing it yeah yeah and i think if you think if you look at you know what happened in these past bear markets right you had you had a big crash in price and you also had the crash and i mean one was like confidence in the thing right there was uh many people who were like we think this is gonna happen and it was like oh i guess it's not gonna happen after all yeah so i think there was that and then i think there was also like a decrease in activity right because like funding dried up and activity dried up and a bunch of people went elsewhere a lot of funds just died yeah a lot of funds companies died right

so i guess i mean obviously the volatility is here and is here to stay at least for a while right so i think we we're gonna see the big crashes but i guess the question is does that actually slow down anything or is it still going to be that like in terms of you know people building applications and usage because right now there's so much capital here right so there was like a run right from 93 till 2001 on the internet yeah and this is where i think well this is the one it's gonna be eight years of craziness and then and then write exuberance and then so where do you think we are on this 93 to two thousand well if it's three if it's three million users then it's like

93.94 yeah yeah [Applause] yeah which is wild yeah cool yeah i don't know anything else you can talk about i think a lot of your listeners went through the bear market and we're building with us in the trenches um i think that's what actually will get you what will cause the super cycle yeah is the people building that have been through the other ones that are like okay well this is fun right i'm actually seeing traction i'm not gonna stop yeah i mean that's that's the other thing right if you think back to 2017-18 right there's all of this exuberance and there was a lot of things you feel like is it like too much exuberance or

something right and now there's there's definitely similarities right you can definitely see a lot of that stuff too but at the same time right if you think of like the that that transition to go from like three million users to like three billion users well there is going to be that exuberance right like that is just gonna be a part of it it doesn't mean it's like a bubble that's gonna it's only a thousand x [Laughter] but actually it's not a lot it's yeah if you look at how fast smartphones uh like grew from nothing to billions of users it took about a year yeah

lucky like baby two yeah yeah cool well thanks so much anatoly it was uh you're great catching up and uh yeah i mean i'm excited to like continue on this like learning with crypto and salon in particular like what's good cool all right thanks so much