Ethereum’s Native Rollup Roadmap with Justin Drake & Federico Carrone

Transcript

the big questions today if you want to launch an app is which L2 do you launch on you know it's a it's a difficult question B rollups from my point of view are even cooler than native rollups essentially like the ticker is eth but why like why does consuming all of these Services acre value to eth the asset and you basically just came out now and are saying that is the only thing that cannot be commoditized welcome back to the rollup your home for technical talks and open conversations that shape the module our future nothing on today's show is to be construed as Financial advice you can see our full disclosures list at tholl up.co ventures sit back relax and enjoy today's episode after a quick word from some of our awesome Partners Redstone is def's modular

Oracle supporting over 100 clients on over 50 chains offering 1,000 assets Redstone is the leader for Bitcoin staking and yield bearing assets they offer the most Robux customizable price feeds available in any model join the Redstone exhibition [Music] today back to the rollup today we have an episode about the native and base rollup design that Justin Drake here has been pushing forward towards the EF we're joined today with fed um fet is a outwardly spoken Argentinian man that likes to stir the pot and express his

opinion openly so glad to have him today fed it's your first time on the show Welcome in and um yeah man how you doing all good I'm here in Barcelona so I cannot complain great CD nice and Justin it's it's your second or third time back uh quite a few things have changed since our last conversation uh H how are you and uh how's your latest research bin yeah I'm doing great there's so much happening uh behind the scenes at thef and also so publicly um small side note my mom is also Argentinian so I do have a little bit of that uh you know blood in me no you can see his BL when he debates

with biton people the other day I saw the the debate on elus and it was pretty good I like it I gave a couple of uh eum talks in Spanish but uh yeah those were really hard my accent needs some work though um epic I mean it would be incredible if we could do this entire maybe our AI agent will translate this into Spanish for our uh Latin American friends um as Andy said we're going to be talking about uh Native rollups predominantly today also based rollups and just putting together

the entire ultrasound thesis um I listened into uh one of the first Native rollup calls uh I think it was maybe a combination of a SE sequencing call and and a native rollup call was a big topic and I was astounded at the level of social consensus that has already started to coales around this initiative I think we heard uh videos from Steven goldfeder from Jesse from base from Hayden Adams it just s it seemed like everyone that was running a a prominent rollup stack or app in the e system had had something good to say about Native rollups and describe their intentions to end up upgrading to this uh this new

design so I I I really enjoyed uh the support that is already garnered around this um I am curious kind of like what needs to happen from an implementation perspective uh like do we need a hard Fork do we need an EIP or an ERC maybe we'll get into that a little bit more but uh before we do Justin maybe you could lay the groundwork here and before we get into anything too deeply technical you could just speak at a high level um about base rollups native rollups the way that they work together to ultimately achieve the ultrasound thesis for e and perhaps Justin a bit on why this is an important change from the current setup I think would be helpful okay got it um so yeah I do

agree with you that um Native rollups which is completely different to base rollups is a an easy sell for people at least so far has been an easy sell uh just to contrast things uh I propos baseall laps you know a year and a half ago or so uh you know seriously trying to to advocate for it and at the beginning it was me versus the world like no one was bought in uh and like slowly slowly every week I'd have like one more supporter and kind of this up only kind of process of pilling people and now finally we're starting to to to reach a place where people start to appreciate it on the other hand native rollups you know like day one people like get really excited so um it's it's it's a rather easy sell now to to zoom

out the way uh that I think about uh the world computer as a as a service that that you know is provided to l2s is that it breaks down into four different subservices the the first one is settlement and that is the most basic one if you want to be an L2 you have to consume settlement and that means you know posting state roots and having some sort of of a bridge uh so that you can make use of all of the tvl uh and you know we're talking about hundreds of billions of dollars um of of tvl that that you can access by by being an L2 but then if you want to go above and beyond there's three optional services that are at your disposal the first one

is data availability and if you consume L1 data availability you're said to be a rollup and then the the two other optional services are number uh number one sequencing and if you do consume L1 sequencing me you have the L1 validators do the ordering of transactions uh for your L2 then you're set to be based and then the final one is execution and this is something that you can't do today and this is the whole proposal of of native rollups is basically to have l2s if they so wish have the execution be done by the L1 validators as opposed to having a custom form of execution in the form of a fraud proof game which is like custom

written or a a a SN proof um now specifically uh Native rollups what they're trying to do is is solve multiple problems uh one of the first problems that we have today with evm equivalent rollups is that they're actually not equivalent uh they have bugs they have vulnerabilities they're trying their best to be equivalent but really it's extremely hard to build a perfect copy of D evm in the form of a fraud prooof game or a a SN circuit um the second problem that native Roa is trying to solve is that of security councils so you know we have bugs and and therefore we have security

councils but there's a second reason that we have security councils which is that the L1 evm is itself a living thing right for every hard Fork that changes the evm whether it's a new OP codee a new pre-compile or a gas schedule change um if you want to maintain evm equivalence at L2 well you need to be able to to mutate and in order to mutate you need some form of governance whether that's a security Council or maybe maybe some sort of of token governance and the unfortunate reality is that this uh this form of of of possible mutation is itself an attack Vector above and beyond the vulnerabilities uh that may happen

accidentally um the third problem that we're trying to solve is single threading so the evm is kind of notoriously known and famous to be single threaded and really what we're trying to do here is to open an easy way to parallelize uh L1 evm execution and uh the fourth problem that I think is is related is just removing gas limits so the dream is that there just won't be any more uh gas limits the only meaningful limit to computation is data availability if you're talking about pure execution uh then then really you shouldn't be uh limited and that's you know partly because you have access to this parallelism but also uh to a large

extent because you we can make use of the the magic of of of SNS and then there's all sorts of other like ancillary benefits of of of native rollups that you know I I've kind of discovered over the the last few few months so one of them which I didn't mention in my right up is that it actually solves post Quantum security for us so today as as an L2 uh that's a ZK rollup for example I will post uh some snot proofs that are very small that are based on tic curves something like rough 16 or plunc and those are are not postquantum secure so the reason why the teams do that is to pay less gas but then you know that means that that they're not you know credibly going to

um to to survive if they were to be immutable for for the next few few decades and centuries whereas if you were to be a native rollup well as soon as the L1 is postquantum secure then the l2s also kind of automatically uh become postquantum secure and then there's some synergies around uh composability this idea that um you know you can the L1 has a few superpowers that LTS don't have uh one of them being that they have the option if the community chooses so to do what's called delayed execution where the state roots are kind of only exposed to dl1 with a one slot delay it's a bit of a technical kind of detail but it's

an important one and what that means is that we can have synchronous composability with next slot real time proving so the all of this is like maybe technical Mumble jumbo but it basically means that synchronous composability only requires this weak form of real-time proving where the proofs need to arrive within 12 seconds as opposed to maybe 100 milliseconds which is two ERS of magnitude less and and and so much harder and fed maybe you have some thoughts to jump in uh if if not Justin could you kind of like just round out the picture of combining each of these services that the the ethereum world computer offers

settlement da sequencing execution you let's say someone is entirely bought in they're they're you know bought in on being native for execution based for sequencing and a rollup for da and ultimately consuming settlement how does this essentially like the ticker is eth but why like why does consuming all of these Services acre value to e the asset yeah so this is a question that I get a lot like um you know two or three years ago I was known as Mr ultrasound money and uh people made the assumption that all of my research from that point onwards was all about maximizing the bird and and pumping Eve um and this is like far far from the truth uh at least

for uh for base rollups and and and Native rollups um if you look at Native rollups uh this is something that uh the L1 would provide for free effectively um there would be no value extraction it's it's something that's that's optional uh that can be that can be used use for the l2s if they so wish um and what we've seen is that um l2s that can make use of the superpower namely the evm equivalent rollups are super excited about making use of that for free now technically speaking you know if you read the E research post there is a gas limit and there is an EIP 1559 mechanism but again

um the bottleneck is not going to be execution the bottleneck is going to be data availability um so the only reason why we have a gas limit is to prevent a denal of service uh at the Prov level not at the verifier level and I don't expect meaningful amounts of of extraction to happen there now if we look at the base rollups um the story is a little bit more complicated and the reason is that yes when you opt in to the base sequencer by default if you do it naively then you will lose your me and that's kind of and that me will basically go to the R1 validators and so in some sense you're indirectly paying for that privilege but there's kind of two things that are happening in

parallel the first one is that the applications are starting to recapture the me themselves so that me do will no longer leak to the sequencer and therefore you're you're making a very very small uh uh Gambit I call it the meev Gambit where you you you give up you know maybe 1% of your Revenue but you you double you triple you you Quint Tuple uh your your traffic and you you increase your your congestion fees that do ACR to you as an L2 regardless of whatever sequencer you you you may use and then there's a second thread of research which is around something called execution tickets uh which is this idea that maybe the l2s can sell

ahead of time the right to uh to to make use of their l2s by the sequencers and basically recapture that me that that would have been leaked but my thesis is that in the long term we won't need all of this complexity with execution tickets we can just give it up and have a very simple design the first part native on base rollups and I can I think I can explain or give another perspective of why I believe that base rollups are more difficult to they're easier in some way to implement because we don't need any change in the uh evm we don't need a new pre-compile but they're more difficult due to from my point of view social um political reasons uh to to have them in production

uh as a default native rollups are uh something that everybody will want because you right now you have two ways of creating a rollup a rollup at the end is just the blockchain with a safe bridge to ethereum so to have a safe bre what you need to do is to verify that the second blockchain the rollup did things correctly and there's two ways of doing that either with ZK with mathematics uh I'm I'm I am on that team I have been working on building that for the last few years on with different customers on the other way is with fraud proofs originally everybody thought or I thought also that fraud proof would be easier right because mathematics cryptography is more difficult but what happened is that the way you do fraud

proofs or optimistic rollups is by being able to the rollup does everything and you suppose that everything went well but somebody can go and challenge what the rollup did you can go and say like no look from here to here you lied you did a mistake to be able to do that you need to implement in solidity in some way or form uh there's different ways of doing it but you need to reimplement the evm in solidity so you need to reimplement um uh a VM to be able to show that somebody lied logically in one step so to do that a it's complicated you have a lot of solidity there was if I'm not wrong uh solid uh arbitro I believe found a few months ago a buag in

the fraud proof system of optimism uh and they fix it on two it's complicated not only because you have to implement it in solidity but because your VM needs to be able to be open up to have like int introspection to be able to see the memory the call stack so that anybody can go and watch what the BM is doing to then be able to go and challenge it on Del one so what I'm trying to say is like it seems super simple uh theoretically super simple in engineering Wise It's so ult to get right um that's why until recently we didn't have permissionless fraud proof rollups arbitro Just Launch bold on you can anybody can go on that's an amazing

I I have huge respect for arbit engineering skills they implemented something where uh you anybody can challenge uh the steps done by the BM but that has been super there very smart people and I'm prettyy impressed by how they they did it because I believe like long term ZK is gonna solve that in an easier way so why native rollups are uh already sold to everyone it's because you don't need to do that what we will do is just have something inside the VM that something is called pre-compile um it's just a way to easily show uh that anybody can go on dispute and say like hey I want to make sure that the change from A to B went well so what you only do is you have a the default evm of

ethereum making sure that that change went correctly so from my point of view this will able uh as a ZK person I think this makes way um cheaper because you need to run uh a VM on top of the evm um written in solidity that is back prone uh difficult to do expensive slow this will make um optimistic rollups way more uh yeah the market should adopt them way more because they will just work they will be easy so I don't see any I don't see any criticism to Native rollups base rollups from my point of view are even cooler than native rollups I was on the team that publicly criticized Justin when they came because I was working

with all these teams that raise money and I couldn't like I didn't criticize it because I was against Justin or anything like that I didn't fully understand its power it was a little bit more complicated because I didn't fully understand the map Pipeline and I didn't fully understand the the full flow uh once I understood it I could see that there was no contradiction between the teams that raised money between the people that were the centralized sequencers on giving that uh to the to the L1 proposers on the L1 validator set however we can go into the details I believe that the problem is not M as sh as Shin said like it's not a technical problem I believe there's more political or social problem because if instead of having a centralized sequencer what you

do is you put this uh sequencing on the L1 you have a morphe market and there will be people that will be competing for that map more openly on that there's obviously teams that don't want that or would prefer not to do it like we can go into details and expl can explain it in detail that that's my theory I think technologically it's uh there's no way it has something bad there's some potential criticism but I think there's a more like economic type of debate where some people might not might want to delay delayed I think long term everybody will do it anything to add there Justin no yeah I I I do want to highlight something that that F said that uh one of the

stting points of native rollups is Simplicity right you're able to take potentially thousands of lines of code that describe a SN circuit or fra prooof game and replace that with basically one line of code that calls this execute precompile that's what it's called um I would also very much agree with fed that uh the native rups is in some sense you know 90% a technical thing and maybe 10% social whereas it's flipped with the with the the base roll ups it's 80% um you know incentive alignment and social alignment and uh we don't even need a hard Fork to unlock base rollups

we can we can start doing them today and you know the 20% technology is all around unlocking pre-confirmation uh which is something that that we should see uh uh you know in in production in a big way uh this year awesome so uh before we get deeper into the technicals throw a curveball here I put out a tweet the other day that really riffled some feathers and it's because the first part of the Tweet said ethereum needs to tax the l2s and people said many things to me about this uh from a multitude of what are you GNA do it's it's impossible to tax the l2s um you can't raise uh da prices uh you can't enforce attack but what I was

trying to to really get to is that uh there's there's very little rent being paid to ethereum and to ethereum holders and stakers from l2s today and centralized sequencers of l2s uh namely the larger ones are are making a lot of eth and then not actually using it back in defi or it's just kind of sitting idle um and not really contributing to the broader flow of the L1 stakers or validators um and so I I'm curious Justin if from your perspective while I'm sure you disagree with with this idea of taxing because why are we going to sit here and try to fight for scraps when we're still so early right the entire game for ethereum is to operate uh at in at the economy of scale um but

would you say that native rollups are a shift in the direction of and you kind of explained it earlier but I just want to get further confirmation of returning value back to ethereum L1 returning value back to eth the asset like is that the main prerogative that you have for this or is it is it that plus the technicals plus the other factors here um because part of my thing was like we really need to figure out a way to get these centralized sequencers basically returning their like some of their profits and some capacity back to stakers or L1 or validators and so I'm not sure that that's the exact right way to do it and I frankly I don't think that me using the word tax was the right nomenclature but I think that that's

kind of the general thought that I had so curious if if you if you see native rollups as kind of that directionally in the same kind of way yeah so from an enforcement perspective like you need to be able to to enforce the tax and I I I agree with the people who said that you know it's it's very difficult to enforce because you know the L1 doesn't really know what is an L2 doesn't have S such such a notion um one of the mental models that I have is to just look at the the fundamental computational resources that are at Play with with a blockchain uh we have things like bandwidth compute uh and and and dis IO and storage uh and like my thesis is

that um we're going to commoditize almost all of these resources so with with SNS we can commoditize execution uh we can commoditize dis iio um and the the one thing that is basically uh you know impossible in my opinion to commoditize is is is the bandwidth component which comes through through data availability um and uh I think that is the the one fundamental bottleneck that that we have in blockchains and that is the only place where we can you know meaningfully ACR value now um in some sense the native rollups do a couple things on the one hand they will try and

commoditize the complement of data availability uh so they will help help commoditize execution right right now um well if you if we go if we roll back the clock 5 years ago um you know execution was a differentiating factor we had l2s you know raise hundreds of millions of dollars to build this technical mode around who could be the first one to have a fraud prooof game or snar circuit and now you know ironically we're seeing a lot of the most sophisticated teams in the world uh L2 teams I mean uh that are just ditching all of their technology like scroll just ditch that technology Z Cas saying ditching that technology and there is basically just using

commoditized components uh you know ZK VMS from vendors like like R zero and sp1 and and and others and in some sense native rollups is the natural conclusion of this commoditization uh effort now the other thing that native RPS is doing and this is a little bit of a of a subtle point is that it is in some sense a little bit pushing the needle towards rollups versus non rollups so basically it's trying to encourage l2s uh in some way to make use of of da um now it's not to say that you can't have um Native l2s that are that are not rolled UPS you can have uh what I called

optimums where basically you you you post the the data uh offchain uh and then there's a a a fraud prooof game that can be brought on chain um but if you want specifically uh rollups that have the synchronous composability um and the the security properties of of uh of of native execution you're going to need to consume uh da ethereum da and so in in that sense it's not only commoditizing the compliment but it's also to an extent uh you know increasing the demand for for da uh but like my thesis is that any anyone who who wants

this highgrade security this high grade composability is going to want L1 da regardless of the of the Native execution so I don't think it's it's that much of a of of of a shift yeah yep yeah if if you introduce Alda you introduce a different sequencer or different validator set you lose some of the compos ility properties um and that's important for this entire design to work also um what's funny Justin is that on Twitter uh there's been a war being fought that da is a commodity for a while and you basically just came out now and are saying that is the only thing that cannot be commoditized right now I think data availability committees

are can are Commodities right those are those are not those are Commodities because they don't have the same network effects that you get from this native rollup design the interrup all all all the security properties um but I guess then I mean before we get deeper into technicals my last one here is base is saying okay look we're going to push to 250 megag gas per second we're just up in the gas limit right now we're at like 20 I think 25 is maybe the highest that they got recently around there uh that's 30 to 40% of ethereum blob space currently um so if if Bas so Bas is basically saying to everybody else like we are willing to go and push the limits

and eat all of the blob space and everybody else is going to have to kind of figure it out as they go in terms of their ability to to increase the gas limits so with your vision of this kind of uh infinite gas limit per se or this ability to remove the gas limits that is also very much reliant on the pectra hard fork and dank sharding and the ability for ethereum to scale its da component uh relatively quickly is that correct or is is there a better way to think about uh this thesis with regards to ethereum's road map what will happen over time uh is that yes we will increase da and I we will increase it I think even more aggressively Than People realize um so yes we have the 2x bump in

pectra yes we have you know dank sharding which might bring us to 128 blobs per per slot but I'm actually a believer of you know 1,24 blobs per slot uh and and Beyond and you know part of the reason is that we have neilson's law which is that every year on a compounding basis we have a 50% increase uh in uh in in bandwidth uh for for know internet connections around the world and there's no real reason uh to believe that that's going to stop I kind of went down this rabbit hole for for a whole week at one point and I was kind of asking myself okay how much throughput can we push through a single fiber uh of

uh of optic cable and um you know this is something ridiculous it's about the the width of a human hair um and you can push um you know orders and Orders of magnitude more than what we can today like way way way more than a pabet per per per second and so the the hard bit was kind of laying all of the fiber digging all of these holes and and connecting uh to to everyone's homes and now you know the only thing that we have to do to upgrade the internent infrastructure I'm simplifying a little little bit is to to update the end points so all of the the routers and uh and the switches and and all of that uh infrastructure which is so so much uh easier to to up to upgrade than than the

the fiber which is a a one-time job so I am expecting Nelson's law to to kind of keep on continuing for a decade maybe maybe two decades and when you compound 50% per year over a decade that's 57x so like call it 100x like it's roughly two old of magnitude and you know um what I'm hoping we'll see is that within the next 10 years when you have neon's Law plus full dank Shing plus all of the amazing data compression that the rollups can also do on their end uh we're going to get to 10 million transactions per second which sounds completely insane but you know that that's that's where I think we're we're heading um and it's it's I think it's enough for for everyone because it's

it's roughly uh 100 transactions per day per human uh and you know every human will have all of these agents acting on their behalf and you know they're going to want to make you know roughly 100 transactions per day that seems like a a reasonable back of the envelope to me I know that you know people like don't believe that there will be that much demand but you know I I am I guess a believer in the in the um uh what is the term the induced demand that we will have uh from from being able to to lower the fees um but yeah at this point in time actually the the economics are are totally distorted um

we are very very far from this endgame one of the things that we're seeing is that the applications are still not um you know recapturing their me and so if you look at an actor like base they're making a ton of value through for me and that is just like temporary noise um another thing that we're seeing is that the vast majority of the burn uh is coming through L1 execution as opposed to the blobs and again this is just a a temporary anomaly just because the the L1 execution is so limited but as soon as we have these these snar verifiers and and and all of that magic then it will no long no longer be the the bottleneck that anything to add there

yeah quite a few things so the first one uh it's funny Justin that you were mentioning the fiber cables uh that connect all the world I used to work on that on the routers and I have helped scale those routers before crypto uh a lot of like physicist you're a physicist right Justin like physicist mechanical engineers because there's a lot of like things related to light backends on I was only on the router you're you're a physicist right or no mathematician sorry no mathematician uh danat is the physicist oh okay so yeah so related to that I have been working on the core infrastructure of Internet since 2005 professionally uh on since 2003 uh more on my uh yeah as a hiker or engineer and I have seen how everybody

thought that uh the demand won wouldn't be there on that uh there that you would hold some special property uh as an infrastructure provider that will have a motto that will block competitors I don't believe that exists like I agree with everything that Justin said I think that bandwidth is the and I think so people also say it is the is the end game but from my point of view I'm more like I'm closer to a taxi driver than to to J or dangr way of thinking on I think things from the customer I think the the interesting thing about Native RS on Bas rollups is that what you're doing is creating a very you will be able to have LSUS that are faster and cheaper than uh

solena like that's what you're going to have that have perfect inop with ethereum and I know from the outside it looks like okay this Mambo Jambo of AUM researchers again some new term that nobody understands but the what we're building now is just a solution to the problems that the eum um rollup Centric R map created etherum decided not to leave decentralization on verif verifiability many of the ones changed their consensus I believe the consensus is the most interesting part on the bean chain that Justin is working on of AUM decided not to change its properties actually it's making it more decentralized more postquantum secure uh

more safe more CK uh Native because it's going to choose CK hashes uh sorry hashes that are CK native and I think that's a great take because you can change you can choose performance over decentralization and very very fa ability the thing is that if you do that you can continue until having a Ser like 10 servers and it doesn't change anything from Web Two I have been building web two applications for my whole life and I love doing that but the core idea of blockchains is that they're verifiable and they're decentralized what I agree with you Andy on with many people is that ethereum at one point got too obsess with that they deliver in some areas or there are some issues and

I think these are fixes to those issues what are the issues a there's not good in trop if you show if you show to your mother or father like yeah this is an ALU and you have move assets there's some protocols like across that make it simple but there's no like it should be for the user AUM should be only one place and I want to have that and I believe that native rollups make it safer and Cheaper by removing fraud proofs or like by removing CK verifiers that can have bugs but base rollups enable interoperability between the the Del one and Del two you could have something on Del one on the same block being executed on Del two and then go back or switch assets between those two on this won't take a year like I cannot

disclose everything yet but we have quite a few friends uh working together on site on London uh and we're gonna publish something in the next few days that shows that pre-ms are interrupt are not so difficult to implement and we have them almost ready with real code like it's not going to take five years it's going to be ready in a few weeks it's going to work and we will be able to Showcase it with uh hundreds and thousands of transactions per second on what I believe is the m of AUM it's not going to be it's not that I don't believe in justtin research or the burn I think it's going to be this incredible community that has the best applications where everything works perfectly on we have the everybody is building something

wants to build first on ethereum so we need to keep that motto on I think that's what what's going to make things um long-term uh like it's what what's going to make a te the best on if we don't have base we don't have native you don't have that coefficient and you're breaking the you're not making the user feel comfortable so this puts a te al1 or Al two back on track to give the best experience to users I think that's what's going to make if one of the things that it's going to make is um um performance long term the usability rather than only technological research is the fact that applications work correctly fast chip on they're connected to the New York of D that is a t imagine if the world GDP moved on chain guys say

GM to safet the newest protocol by safe isn't a blockchain and it's not a new layer 2 it's a transaction processing Network built to send money anywhere and swap anything on thousands of layer twos from a single Universal balance in sub 500 milliseconds it's fast it's smooth and it's simple and it's built with safe multisig level security in mind safet unlocks smart accounts across multiple chains allowing users to supercharge security scalability and speed so get started today at net. safe. Global and remember to follow at safe on X yeah and my only point on that is just that I think that that there's been a lack of

prioritization on ux and uh a overemphasis on decentralization which is probably uh By Design um but I think it's it's time to kind of really prioritize what people are doing on chain and ux but then also as Justin said we have ai agents and these AI kind of operators I mean are we really even going to be taking our ledgers out and doing transactions on chain in the future is the question right so does it really matter or are we just going to be using natural language models to express what I consider now to be intense on chain so Robbie take us a little deeper into the technicals of native rollups will you and some of these pre-compiled State proof State access proofs things like

this yeah well I I'd like to maybe get into that and to set the scene for that I want to put a a mental model uh in front of you guys and kind of get your take on whether it's it's valid or not because Andy and I have been having lots of these conversations with uh ZK teams and uh also talking about this evolution of of scalability ultimately starting you know from the L1 but then going into you know the rollup Centric road map first optimistic Roll-Ups then the acceleration of ZK I think everyone's kind of on board like every rollup will be a ZK rollup once CK is accelerated But ultimately you know the the roll up L1 L2 landscape starts to get really really blurred and these things just

kind of like look like Blobs of state that exist in the void connected via ZK Justin I see you're kind of nodding your head could you kind of just validate that mental model for us and and maybe start to connect that to some of the the pieces that ethereum is working on Stark ification Etc yeah so we basically started with this execution environment where you know everything was luxurious we had um you know full synchronous composability we had very little friction um and then you know we had this this this problem of of scalability and and fees and so what we did is that we we kind of left

this very cozy execution model and kind of made sacrifices we made there were trade-offs made uh we basically introduced uh boundaries we have these silos uh we have these decentralized sequences and we have slow asynchronous communication be between those silos and really what I think will happen is that we're going to go full circle or we're going to go back to this luxurious world uh but the fees will be you know a 100 times lower or even less than than what they were previously um and in order to to get there we need to solve two fundamental challenges the first one is this like really low latency communication uh trustless communication

uh between uh between this you know these abstract pieces of State as you as you put it um you know previously if you wanted to communicate between two pieces of State for example a and Unis swap like you want to make a Liquidation on on Unis swap you just make a call to Unis Swap and you wouldn't think twice uh now today you know you you really have to think twice and because of these asynchronous boundaries you can't even do like liquidations properly because you don't have unified liquidity and things like that so uh what I think will happen is that yes we will solve the the low latency communication that's with the realtime proving and with Native rollups it's sufficient to have next

slot real time proving as opposed to same slot real time proving and then the second thing that needs to be solved is more of a social problem is like how do we all coordinate around a shared sequencer because if you don't have the shared sequencer then you don't have the the synchronous composability um and the way that I think about it is that um the only viable strategy for such a diverse ecosystem as as eum to agree on a a shed sequencer is to have something which is maximally credibly neutral and the only thing out there that is you know Maxim credibly neutral is eum itself because it doesn't introduce a new token a new brand or a new security assumption and I certainly

don't think that um if a specific cluster or or L2 player comes up with their own uh sequencer that they should expect that competitors to go to go and adopt it um I mean we do have a project like espresso uh that are trying to be uh you know more neutral because they're not you know associated with any one L2 uh but even even espresso you know has a new token a new brand and a new security assumption so you know some players do like espresso but not everyone and because it's it's a subset uh it's going to be is going to be an uphow ble on one of the issues is that none of these

clusters uh this happened before with uh like Hardware in the from the 70s to 2000 none of these clusters have incentives to prioritize the interoperability between the different clusters like you want to keep your users because that that's what you going to make money with the fees of like actually it's a tax and like of getting out of the CL cluster or in like that's the only way they can do it like they that's the way you can make Revenue long term I respect many of these teams I think these teams improve the team a lot but also at some point I don't think a development can be blocked by the business needs of different like many of my customers might get angry with me but

hey I'm an A Team I'm an A Team person and that's what I tell every everyone it's like I want the team to win and for a team to win we need the team infra to be top-notch on elus need to start using this infra I don't think there will be a lot of social coordination problem I think everybody The Experience that's what I was trying to say before I don't know if I explain myself because of not being an English native speaker the taxi driver thing is like what I'm trying to say is like the user if he's able if a taxi driver understands that he can move assets very cheaply and very fast between l2s on L one on the other one the other chain that doesn't do that doesn't have it he's going to choose the first one right that's what I think you

need to think from the customer backwards to the infront not the other way around so uh the social coordination problem will be solved by the incentives that you have of users being happy by having a cohesive um um flow on ethereum so I think it will be that that's the incentive you will have in place I think the demand will be there I'm speaking with countries with governments with big corporations many of them wants to launch rollups the demand will be there it's it's already there like in the last few months due to the changes in the US policy uh everybody is trying to do something new on there like I'm getting calls every every day by big players they want to use this so the demand will be there and I don't believe any

blockchain not Solana notu not abdos uh there's a new one called p network super interesting none of them will be able to have all the users that are needed every Ro every block will have rollups the difference is that the teum already has a me pipeline has a lot of research has a lot of teams that are working on that the only thing we need to do is think backward uh think backwards from the wallet to the infra align the me players on align the wallet to make an excellent interop interoperability uh between each part and whoever doesn't want to do it he will be left behind yeah and I think there's a couple Dynamics on the incentive alignment that I think will happen like one of them is

that if you are part of the long tail if you if you single-handedly have an M you know a large amounts of tvl while you have an incentive to go connect to to to some other source of of tvl and so what I think will happen is that the the longtail of l2s will have an incentive and might be some of the first movers to to become based um what I think would might also happen is uh is app chains uh right because what what one of the the big questions today if you want to launch an app is which L2 do you launch on you know it's a difficult question uh but in my opinion the obvious answer should be you just launch your own uh chain that is like the most credibly

neutral thing uh that is the the one that minimizes dependencies on other ecosystems you know governance structures and and and security assumptions and and branding and and all of these things uh but in order for app chains to be viable you need them to to connect to to to broader Network effect right because if I'm a tiny app chain and I'm a little island in the middle of the Pacific like I'm just going to die out um and so today you know apps have an incentive to launch on l2s because they could act to a broader ecosystem but once we we we basically uh realize this vision of L1 contracts with lower fees um then you

know you can just deploy as if you you did in 2017 as a smart contract also I think the long tail to support the long tail is going to help ethereum long term because Warren Buffett always says I don't invest in Tech because a kid in India or Argentina or anywhere from a garage can disrupt the market so that's in some way bad for uh compet competition because if you're there's no motto but that's good for users that's good that's one of the good things about capitalism anybody can innovate on disruptor market and provide the new thing for the users and I think that if we have Frameworks there's one that we're creating between many part is called fabric it's more like for now uh it's more like uh guidelines on specs I

hope at some point it's also code like G um you will be able to support uh people that didn't have tvl they didn't raise a lot of money and they can innovate try new ideas and if those ideas are cool we they can influence the full Market that's very difficult nowadays if you have a that's one of my criticism with crypto Dynamics in terms of like the social layer you need to you cannot create something new on influence everybody just by creating something that is good for the user you need to align a lot of people I think this will help uh just launch a new rollup that is native on base on if you have cool ideas on the users like it they will start using it I think that Dynamic is particularly important because if if you're one of those young people that's

wants wants to build on on ethereum and you know some you know some VC or some fund or somebody's taking a chance on you and you want to launch an app chain to to do it today is is to launch as an island in the Pacific Justin your your thesis here about based app chains is actually the most compelling that I've heard for any sort of app chain because I got so much push back all the time about this where it's like no man we don't need more chains we don't need app chains because uh there's no interrup you have to Bridge and there's all this all this uh concern but basically what you're saying Justin is that the new like like the the new launch your your dap on AR because it's defi or launch your um your game on

on optimism because it's it's a gaming chain or something uh that's more going to be like launch your app chain as a based app chain and then tap into all the other based chains and app chains that can coexist in silos but also have that that kind of sort of native synchronous composability which is pretty uh compelling for etherum I think yeah and just to like set the expectations it's it is going to take take time in order to truly reduce the friction close to close to zero uh because we we need to simulate the equivalent of these synchronous calls we know we need to do things like you know blob packing and blob sharing and shed proofs and we need to do things like

like AG layer where we have uh cross roller bundle safety and uh you know there is a a bunch to be built so it will take time um one of the things that I think will happen the shorter term and we're already starting to see it is um projects choose baseb UPS just because they're they're more simple and more credibly neutral like one great example here is is name chain right ens uh has only a single source of Truth you can't launch ens chain simultaneously on arbitrum and and optimism there needs to be a root of of of of trust and um the only natural credibly neutral thing to do is to launch your own chain and the only

you know appropriate sequencer for for name chain is is in my opinion the the L1 sequencer and it's no surprise that you know they've announced that they're going to become based and in that vein what we're going to see I think is um similar similar entities so FID talked about nation states like put yourself in in the shoes of a nation state do I want to launch on optimism or obit no it no right like I I I need this maximal credible neutrality um I need to be as close to the metal as possible put yourself in the shoes of a very very large institution like do I want to launch on top of arbitrum or optimism no like I I want to be do you

want to launch on Solana they don't want to launch on a chain owned by somebody else they want to own their own chain I'm working for example in a concrete um um land registry uh it's one of the ones I can talk about on they literally told me like we need to launch our own blockchain and they wanted to launch something with customers and I explained them why that was a bad idea uh for too many reasons and I told I explained them until base rollups I didn't have a proper solution like I started with ZK rollups but to be fair there wasn't an easy ZK rollup to launch so then like I didn't want to do the infra like I sell infra but I didn't want to sell them the infra I just wanted AUM to be tapped uh on real use cases on base rollups are an

excellent example of that because they need to run their chain it needs to be proven it needs to be verifiable and they also win all the good things that team has the already the TBL the user base so you have from my point of view it's the in some way the end goal of what blockchains uh wanted to do verifiable computers that all can work together between each other I have a quick clarification because you know we're talking about launching app chains which I fully agree with and and retaining sovereignty um and ownership over that block space um and you know we're talking about synchronous composability and I love the example of you know going and executing a Liquidation on Unis swap via Liquidation on a I think that's you know one of the most practical

examples but we're talking about it almost entirely in the in the context of the sequencer and being based and that is how you tap in you you know you can be a based app chain retain your sovereignty and tap into the synchronous composability but you know we started out talking about Native rollups and how if you use EA and you use uh the native pre-compile you also get synchronous composability so is it the case is that still the case like do you do you necessarily need to be based can I just can I be a native rollup that uses EA but I run a single sequencer can I still get uh synchronous composability so the synchronous composability technically only requires shed sequencing uh and one example of a shed

sequencer would be the base sequencer but it doesn't have to be uh you don't need to be native so you could imagine a non-native virtual machine maybe something like stocket which Cyro that could be synchronously composable uh with uh you know with evm State uh you know there needs to be a layer of translation and the good news is that we have several actors that are specialized in this translation we have things like like Searchers and builders that that uh you know will will will Bridge the the Gap I mean you can think of it as being a very similar problem to Bridging the the language barrier between two dexes uh you know let's say uh you know Unis Swap and and Susi swap or something um they're you know they have a different

notion of of state and how they how they handle the the the order book and um and the the arbitragers will just specialize at at Bridging the Gap and then what I think we'll see as well as you said Andy is is AI you know just just building that Gap so the fact that you're you're not native um is is not a fundamental limitation the one thing that native does unlock is this this fact that you don't have to worry about the realtime proving because that's done by the L1 and the L1 has this superpower that it's allowed one extra slot uh which the applications don't don't have that that kind of luxury got it I'd like to to go through some of the push back that you

received on the form Justin um specifically from some of the larger prominent uh L2 Builders and professors of sorts one being Ed Felton uh one of my favorite people in the space great smart guy um has a lot of built a lot of really great tech for arbitman offchain labs um one of his pushbacks Justin was that uh basically rollups within uh rollups need more expressivity uh basically that R rollups on top of orbit which can be l2s or l3s or stylus which can effectively have a multi-m setup cannot express themselves within this native pre-compile um and I don't know if you have that in front of you or if there's any more details that I was missing out

there Rob but uh yeah what are your what's kind of your rebuttal there or thoughts about expressivity of rollups and I I think you kind of just touched on it in terms of the different VMS but uh yeah maybe just clarify yeah so at first approximation uh there's the evm equivalent rollups and then there's there's the rest um and um in my opinion if you want to do something other than evm equivalence you need a really really good reason to do so uh so Cairo is maybe a great example right because they're an order of magnitude maybe two orders of magnitude more snar friendly by just completely changing the the programming model arbitrum is maybe another great example

right with stylus they've unlocked wasm and now they can tap into you know millions and of developers that can now develop for arbitrum versus the whatever it is you know 100,000 or less developers that that that know solidity and so for them you know it's a massive you know 10x unlock and you you can go down the list of of virtual machines so one of them may be Mega uh like they've decided to change the state Tre and that gives them 10x faster uh right bandwidth um you know that is another 10x optimization but if um if you're going to depart uh you know from the evm um and and you only making you know 10%

optimization then you know is that really worth it um and I think the the native uh rle that pre-compile the executive precompile will basically um create kind of reinforce the gap between these two categories we're going to see things that are evm equivalent and things that are you know very diverse and Innovative and and get a huge benefit by by doing so now one of the questions is can we relax evm equivalence so that it's not like super rigid uh and allows for some some amount of innovation I think the answer is a big yes so one idea here uh is to actually expose um a lower level virtual

machine like Risk 5 within the evm and the reason why risk 5 might be an especially appropriate choice is that within the industry of ZK VMS more than half of the vendors are building risk five zkm it seems to be the de facto industry uh standard and so if we're going to be implementing and enforcing this precompile using zkv Ms in the in the back end and their RIS five zkm then we might as well make use of this expressivity this is basically the same idea as stylus right because the way that the arbitrum fra prooof game is implemented is with wasm and so they were like hey let's just expose it we

have it for free uh and you know for arbitrum it was a great unlock and I think for E could also be a great unlock there and so if you have a custom pre compile for example um then you could implement it in Risk 5 and that would kind of be a simulation of that pre-compile and You' get similar benefits you'd get maybe uh a more optimal implementations because you have access to this uh lower level um virtual machine now the other thing that you could uh imagine doing is basically having uh the execute precompile um basically stop and pause every time it hits an instruction that

it does not understand and it will basically revert back to some sort of external State transition function that can be handled by the application so there could be a custom foolproof game or a custom snar circuit for these State transitions that are not understood and so if you imagine a long string of computation you know there might be interleaving parts of the computation which are not natively understood by DVM but that's fine you just have checkpoints in and out every time you you kind of inter leave and then you fret the needle by using this glue State transition function that will that will piece everything together into one continuous uh execution so in theory

even with the execu pr compile you can do arbitrary complex uh computation just by Fring Fring in and out um and then there's there's all sorts of you know uh other things that were like considering in the design space so one of them could be um you know what if we could programmatically uh change the state tree so as a as a as an application developer instead of using whatever is available on the F you know right now which might be a Patricia mercel tree or a binary tree or whatever it is um you can specify your own thing and so in the case of Mega where they are like 100% evm equivalent with this one exception where they've changed out the tree well

now they can make use of of that pre-compile and I think what will happen is that um there's going to be this design exploration phase uh where we kind of try and understand what makes sense to to to support in terms of flexibility there's going to be something things that are very low effort and very high value because you know it means that more people can come under the fold of of native rollups and some things that are going to be uh you know extremely complicated and might only uh benefit you know one L2 and I think maybe those things uh you know we'll naturally have to wait appreciate the indepth answer there um fed anything to add as well yeah I

was gonna focus on the risk five part like for me e ethereum is first a set of values uh on a community of people of like that come from different parts of the world with different uh objectives but they coordinate around one idea to create a world computer the safest uh fastest cheapest world computing computer on I have never been a big fan of solidity or the evm I I develop in many languages and I have for different reasons I I could go into details I'm not a fan of of solidity more than the evm I don't have a huge issue with the evm I have a bigger issue with the solidity um um I think if we're going to

become a world computer risk 5 is the world is the is the way to go like for me AUM is the consensus the social layer plus the consensus like the consensus has no mat somebody the other day said that uh Celestia is today what the dream is going to be in 2028 or something like that no like a the consensus is uh favor availability uh on liveness over safety you have 4,000 real world notes running uh tender can only run with 100 noes uh there's too many things I can would go over it those things like don't I love Celestia and Mustafa I think they created something very good I think it's going to also actually help a TB in many

cases because the D layer might not probably scale to the degree we want on I actually helped many of these um clusters of rollups to use Celestia as a as um as a daer don't tell Justin that don't don't tell Justin don't tell Justin he's going Los no no no I think what I disagree I think Justin explains very well that the team provides different type of services on you can choose different type of services and you can tap into the truth of ethereum that is the best trust available and you can rent or you can use which part you want for some use cases celest is okay for others cases it's not so I think in ethereum we're not I don't see Justin or vitalic or dangra Fanatics I think

they're actually what the one thing that shows this is like the other day also somebody said people are saying that base r rers the way to go because of alignment uh if you check my Twitter I criticize thef I support thef in many ways they have the level of openness from Justin vitalic and many others to new People Like Us is incredible in comparison to the rest of the industry I'm super proud of being part of eum and working with thef but they also criticize it that shows that it's a open um organization on it's not because of alignment like I I am I literally critizing solidity on evm in a podcast of AUM or talking about AUM with one of the top researchers of AUM and the reason why I was criticizing it is because for me the world computer needs to go towards the ISA the um the CPU

that everybody's going to use in the world your MacBook uh it's armm similar to armm that is similar to risk 5 uh but everybody's gonna use r five or ERM on the hardware that's a competitor to the previous esis that were x86 Intel IMD and I believe ethereum needs to go towards that and this is also an opportunity for iium to support the most important Isa used everywhere so we will be able to like you your evm is different okay you can run it on top of risk five as Justin said you want to support move an amazing language created by Libra Facebook researchers I think it's an amazing um VM on programming language perfect eum will support it like ethereum will be able to run any

type of computation that we want with the properties I said before it's going to be super cheap thanks to uh centralized map players running base rollups is going to be safe because we don't have any more um uh um um solidity contract with the multic to verify uh the either the FR proof or the ZK than to Native so I believe that rck 5 in in some way is not going to only enable the current players but it's going to enable a lot of innovation in in in other ways on eum is that it's a decentralized computer like from me for me I'm not saying trting or thef or any other for me I'm interesting a because of that not because of the like those are languages I I like a lot of languages I don't like

that one I like the protocol of AUM that enables to run anything on this will enable more things yeah um I I realized we're drawing down on time here so I I think just maybe one last one uh from the comments that came to the native rollups piece uh and we're Outsourcing our comments today to Ed Felton because he's smarter than than Andy and I Mak Our Lives easier so uh for this case um we're talking about kind of in like you know risk five and embedding this deeper into the L1 to kind of like increase the expressivity of some of these State

transitions um Ed pushes back and basically says in practice a chain needs to have a single state transition function uh because it's a core principle for these contracts for any contract to call any other contract and that the these calls can Nest within one another and so we can't have a separate evm execution State transition function and then and you know another state transition function to extend these features to another VM um we basically need to have one single state transition function in order for these contracts to call one another and for them to Nest so does this contradict the risk five idea no fed shaking his head Justin's shaking his head so what is h how can we

reconcile these two ideas go check the code base of r55 it's not production ready but there's different ways of doing it you just need an AI or some type of way of communicating the two things like that's an engineering problem it can be solved yeah so r55 is this um initiative I believe from eaka the team behind ref so George and Company uh I haven't looked into it myself but it actually you know ironically a very similar uh thing to to stylus uh where you have this separate virtual machine and you can fret in and out and you know as as as Ed said you know you need to have like you know this this nesting and whatnot and this goes back

to the comment that I had where you have a long piece of computation you you you may start within DVM but you might want to go out in and out arbitrarily well you just have checkpoints and then you kind of patch the gaps with your custom State transition function if that was correct you couldn't call C code from rust like literally like you can call it because there's what you're doing is agreeing on an interface between the things that's what eum calls ABI uh well actually I think it's everywhere I don't know like English sorry dyslexia but it's AI it's just a way of like agreeing on how you call things between different things uh so you just need to agree on that I didn't even check the r55 code base I Know It uh but I'm sure they solved it and if

they didn't solve it we can solve it like it's not a impossible problem to solve I guess Ed does make a good point that if you take what was written in the E research post literally um then yes like as currently you know suggested then then then you have you know these engineering difficulties uh but like what was presented first of all is not meant to be the the N or be all like there is a whole design space uh uh exploration and also even if we were to go with that as as a first step there is room to to further generalize the pre-compile down the line got it Justin and fed both one thing that you guys mentioned with regards to

base roll specifically fed you mentioned how it was more of a social and political coordination problem regarding giving up uh economic value basically um and Justin on on one of our last podcasts I think it was like the first thing that we did we talked about the incentive mechanisms for rollups as they are today especially those with large user bases tvl and distribution to become based and you suggested some some ideas uh but now we're thinking about Native rollups more so um is the incentive question the same still or perhaps better said how are you thinking about teams that you've been speaking to and have received an astounding U

recognition for this idea and their acceptance of it how are you thinking about their transfer from their current setup which many of them may use use two or three parts of this four-part stack to a full native rollup and what is the incentives required for them to do so and is it just a game of like somebody does it and then the dominoes fall yeah so um you know there is going to to be some work like first of all we need a hard fork and the good news here is that um if we really want this pre-compile I believe we can have this hot fork in 2026 and the reason is that

even though the end game is to have this pre-compile be enforced by Sno we can have it be enforced through naive re execution and so in some sense it's just a copy paste job uh for for the for the execution clients where they they reuse you know the evm that they already implemented it and and now they just reexpose it through through a slightly different API um now in in terms of you know what needs to be done on the rollups um for the incumbent rollups what they'd have to do is take their current state transition function and decouple it into two parts one is the evm state transition part and then what some some L2 ecosystems call the derivation

function so the Der deration function is everything around the evm and it handles things like deposits and withdrawals it handles things like uh authenticating the sequencer special system transaction you know maybe the security Council making an emergency stop or maybe some some some governance uh transactions um and once you know these things are cleanly decoupled you can take the evm part and basically just replace it with this this one line of code um you know for some rabs that might be easy for other rabs it might be complicated it depends on on how how it was uh how it was implemented uh for for

new rollups um it you know the barrier to entry might be significantly lower and if you're designing to be native from day one uh then you might have like very clean abstractions and it might be relatively easy now in terms of the the incentives for for becoming native um you know I've talked about removing the bugs and the security councils but one thing I did talk about uh and this is something that teams might not you know be willing to share publicly but they basically don't want to go to prison um and I don't these are literally the words that that people tell me um they want to basically reduce the surface area of of of of legal attack um and whether that's uh you know a security

Council or or whether that's a governance token having control over a sequencer whatever it is they they just want to minimize minimize that you know these are sometimes entrepreneurs who have left the USA and want to go back home uh these are people who are not sleeping well at night or you know just like for for selfish mental health purposes um they they like that direction um but yeah that that goes you know above and beyond all of the the other incentives that that I've already mentioned two things the first one is I fully agree uh most people don't know it but one of I have been in crypto since 2015 working full-time since 2017 on my full team joining uh because most of my

team was in web 2 2020 but we became public after after we got a call related to tornado cach uh the pre like we were helping on Research uh a paper called toel you can check it out and then we wanted we were entering we got proposed by some people to enter the multi and other things on the day before um everything happened I got a call saying like hey you can have to stop working on privacy and you have to stop doing this and that and you might how uh like tomorrow some people might not knock knock your door and that's when I decided to become ful public like before I was Anonymous and one of the reasons was because I didn't want to go to jail

and recently I raised money I helped not me I helped raise one of my co-founders uhj from Aline um you cannot imagine the amount of time and energy that is invested in working with lawyers we have legal firms in Argentina that where we have the uguay where the labs uh is created uh the US uh kaiman BBI like the amount of energy on time due to regulation put on Founders to not go to jail for only doing the good thing that is creating good infrastructure is incredible and one of the reasons is because the definition of decentralization is not very clear and

if you're a good player it's very difficult to know how to protect yourself from these things so working on privacy all ing a token it's until recently was a huge issue I believe that in 50 years we're going to laugh about that as we laugh about people working on encrytion professors in the 90s uh that want that wanted to have like people did tattoos of RSA like it was a crime to to teach or or write books about RSA uh on the funny part is that they didn't prosecute the bad players they prosecute many of the good Founders on the people doing real business I don't know they would did the loss against consensus so I think that Justin uh she said something very important like I literally became public just because I didn't want to go to jail and I if not I would probably stay anonymous because I

wanted people to know me on yes like you want to be able to lunch infrastructure easily not to have to think a lot about the legal system on I think in in this way B robs help awesome guys well this is a conversation that took many turns some curveballs some laughs uh really in insightful messages across the board as well so guys thank you so much for joining us today always a pleasure and fed you're always welcome on back as long as you're staying out of jail can't have you in there and Justin same to you so guys thank you so much thanks Robie on Andy on one last thing I wanted to mention I it's the

first time somebody invites me to a podcast so I I am super thankful to you guys I think I I I think I'm pretty vocal on Twitter I think you guys are doing a great work at opening debates something that sometimes uh don't happen so much in in in crypto and second of all I'm pretty surprised uh I don't know if you guys are Engineers or what but I'm pretty surprised by the level of deepness on the technical things that you understand like having following your Twitter and it's pretty surprising how you follow debates that are in telegrams in calls like so congrats it's a great work yeah thank you appreciate that it's been a lot of years of podcasting and reading so uh what Justin says sometimes it just kind of flies over but you know you you live and you

learn right so that's how it goes thanks guys thanks guys