Rollup-Boost, Unichain, And TEEs With Hasu & Robert Miller

Transcript

and I said it was a nap chain and I got yelled at hammered the base sequencer 1.5 million times those transactions and they landed 10 of those successfully unrealized versus realized volatility like if it goes up and back down but you don't realize that change in volatility then it's not good for LPS The Operators don't want to trust the user to run it locally because they want to don't want to give away the training weights right which is really the secret Source why don't we consider 100 millisecond block times or did we and was 250 a better choice they call that God mode welcome back to the rollup your home for technical talks and open conversations that shape the modular 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 chain offering 1,000 assets Redstone is the leader for Bitcoin staking and yield bearing assets they offer the most robust customizable price feeds available in any model join the Redstone exhibition [Music] today what's going on guys welcome back to the rollup for today we have Robert and hu from flash Bots Robert hassu how

you guys doing today welcome on good thank you for having us on yeah doing great thanks thanks for having us and and Robbie absolutely guys thanks for coming uh I've been following your work for a very long time and uh appreciate the the uh contributions that you guys have made all across the space in ethereum land rollup land uh the meev space as well we're going to talk all about it today wanted to start with uh the work that you guys have done on uni chain I think that was one of the most relevant app chains that have gone live as of recently you guys were very fundamental in the white paper uh it's not an app chain bro it's a defi Focus chain come on all right okay it's a def5 focus

chain I got yelled at on Twitter for that that's all I'm saying I said it was an NA chain and I got yelled at so all right so maybe we can clarify this today is uni chain an app chain and uh we'll we'll get into a lot more around uni chain today too it's an interesting question I'm not exactly sure what the the definition of an chain is these days so that's one place to start the way that I think we see it at flash Bots and you'd have to ask our friends at Uni chain also how they see it the way that we see it is is that uni chain is purpose built so it is purpose built for their specific use case we have added all sorts of interesting things that make it better for their use case and really the use cases is how do you make

onchain trading as good as possible and if you think about that you come up with the the design for uni chain so you make block times as fast as possible to reduce uh the lvr on um onchain lping uh as well as giving users the best user experience possible so you you get to kind of 250 millisecond blck times um you know how else do you improve the user experience you add on uh free reverse which is another part of the the flashbox feature that that we have and how do you create the best environment possible for applications well you enable me internalization through priority fee ordering and if you kind of take that a few more steps you need to

make this that credible somehow and uh we've done that by layering on a trusted execution environment and so we we kind of backed out the design for uni chain from this purpose of how do you make it the best place uh uh within crypto for onchain trading um with that being said it is not an app chain in in my mind because it's not a single app it is still the evm it is still turn comp complete you can deploy a smart contract that does whatever you want and Unis swap with its uniswap V4 uh design is actually pursuing a platform strategy where anyone can deploy uh Unis swap pools with arbitrary logic instead of just having you know a single monolithic

application on uh on on the Chain so it's this interesting design which is not entirely you know generalized copy paste evm um and not entirely an app chain but somewhere in the middle where it's built for this very specific purpose and tuned for the very specific purpose of you know making great apps and making great onchain trading experience um so we think of it as purpose builds and we think that's going to be a a trend that you will see across many other rollups too anything to add hosu oh yeah yeah uh I mean so first of all agre with all of that I think personally that this delineation between app and chain I think it's increasingly just BEC not very useful as a mental model to make sense of the space so I mean we were starting from this idea

there's one chain and many apps right and then you know some we had started to have like more chains uh you know but like many apps but also more chains and then some apps started to be on many chains but they are kind of the same right so un is already on many chains right same as a and so on what does that mean um but then some some chains basically decided to some apps decided to leave and make their own chain um and you know as Robert is saying they basically better customize what the chain is doing right but now all of a sudden you are uh starting to see on one chain apps that are effectively saying we are kind of upstreaming or like

unbundling things that the chain has been doing we're doing ourselves so I'm talking about app specific sequencing for example where you're going to have a unisoft pool I mean let's talk about C actually because they have been live for 3 years so C is AB specific sequencing just nobody was talking about it why because they are I mean they are they are determining their own transaction ordering for their for their bch auction that they are running so they doing this pre-processing stage of transactions right and so they already you know in a sense they already there on chain right they they are just as much their own chain as a layer two is their own chain because cop is you know in a sense it's a layer two like the layer two is just

like ordering transactions of chain and then bundling them and posting them to the layer One Cop is also ordering that transactions off chain bundling them and posting them to the layer one right and so you know all of a sudden you can see what there's actually no difference between C and the layer two and so I think you know for for all of those reasons I think this you know this like mental model where you have separate chains and you have se you have apps and you can really like map them to each other I think that's increasingly like no longer useful you just got to ask what's best for the specific business what's best to the specific use case like what what are the things that I want to do what are things that I want to Outsource first of all what do I need

for my users and then where can I get it right and I think with unisoft specifically um there just wasn't any alternative right for what they wanted right like if you want to trade on chain I mean uh if if you if you want to provide the best Trading experience on evm you can look at uh centralized exchanges and they are pretty much the Benchmark for this right they have like very fast transaction execution you have like no front running as far as we know although you don't know right on centralized exchanges are like are there people who have like some kind of privileged access to uh like your order flow you know they can front run you the match matching engine on FTX

for sure it was happening you know at least we have heard people talk about it right um they call it God Mode God mode yeah yeah and I mean on binance also it's like pretty well known that they sell like I mean every every exchange sets cool location rights to market makers right so already like people are not trading on the same terms right this is not a service that's that's open if if you're like more of a retail uh you know Trader then you don't have that same access right and so you know if if you want to really recreate something like this in one evm those no alternative right and so yeah we just talked to them about what is what if you can have anything you want what is the

best trading experience that you what that you want to give your users right and then we try to build it all right lots to unpack there and I think that like the the core Point Robert made first which I uh wholeheartedly agree with and hu kind of really added on there is that I think our mental model of of this idea of chains is going to be changed right where we've been in in the space you know doing onchain transactions for what five years now and so we have this this this way that we believe the space has has you know developed and that influences our thesis going forward for where it's going to go um but the reality is I think new users just come in through apps like they are not really caring about the chain

especially if this whole account abstraction and chain abstraction Concepts come to fruition the reality is users come in through apps um and from from our perspective we tend to think they're going to be purpose-built app chains purpose built chains whatever you want to call them and so I think that notion is something that we that a lot of Builders and people in the space should be thinking about rather than this kind of chain Centric future it's more of like this app Centric future and then you can even spin that all the way to being like this intent Centric future where rather than users wanted to use apps they just want to express an intent and have that action uh uh kind of operate on chain um and so that I think that we at the RO have been pushing that

both of those ideas pretty strongly and it's good to see that uni chain is like a validation of that of this kind of like modular thesis of this apps Centric thesis as well as like the the the Practical benefits of um the actual user experience now to kind of like dive into a bit more of the technical side of the user experience there there's a lot of of technical terms here that I think that we want to uh unpack for our community and so the first one being uh what Robert was mentioning with regards to the block times so um we did a podcast with Alex nees buan and Yuki about lvr and uh ass app app specific sequencing and I would like to kind of further uh understand uh more deeply how

the idea of uh 250 250 millisecond block times on unit chain reduces lvr and increases the the experience for LPS and then how this broader idea of of of of increasing the experience for LPS or making the experience better then helps users for uni chain so if you guys could just break that down for us a bit that would be great and kind of how that ties into to the broader idea with uh tees as well you want to shoot hu uh sure so um I think the theory behind uh why lower block time lowest lvr uh it's just that

uh you know basically the more you chop up uh you know time um the more trades can happen in the same time period like the more profitable Arbitrage trades there are and the more fees dip can thus kind of internalized from that so back when unisoft launched I think there was a lot of discussion around uh the idea whether you know just like how much APS are getting cleared in these amm and if it you know how should how we should even think about like the idea of kind of impermanent loss and K put out this very good paper on their research blog and um you know it was basically making the point that you know um very um unintuitively like the lower you make

the piece um you know very often it actually reduces AVR um for a p so like you basically decrease like you you decrease the tick size like as low as possible why because then basically as the price fluctuates you do like you have more of this volatility harvesting effect right because as an IP if the price goes up and then goes back down again that's actually extremely good so so like volatility per se is not is like volatility per say is good for you it's it's but there's a lot of path dependence like you know in in there I mean not path dependence that's the wrong word because amms are fundamentally half independent you mean like unrealized versus realized

volatility like if it goes up and down but you don't realize that change in volatility then it's not good for LPS yeah I mean for for basically I mean you should probably talk to like one of the I experts um on this but but yeah basically you know for it it basically I mean if if the if the price ends somewhere where you didn't start that's bad for you and then volatility is bad but if you end near where you started the volatility has actually been good for you and this is exactly the scenario under which you want the block time to be lower and you want like the fees to be lower because then actually more um yeah basic can generate more fees from the same volatility so you can like volatility Harvest better yeah well then hzu why didn't you guys why wasn't

that I mean I know that wasn't your choice right but why why didn't this why didn't we consider 100 millisecond block times or did we and was 250 a better choice we did consider 100 millisecond block times uh really the reason why we chose 250 milliseconds is that is about someone's action time so it it provides an instant basically instant experience for users when you're transacting you probably can't really feel the difference between 100 milliseconds and 250 milliseconds I mean I guess maybe if you're like a CSO Pro you you might be able to if you have like a phenomenal Rea reaction time but for most people 250 milliseconds is basically instant and we wanted to start there um you know

because this is Cutting Edge super performance software and we may end up reducing it from 250 milliseconds to to to even lower um from there uh another sort of related idea that that we've kicked around with the unisoft folks and some other people is the idea that um maybe we can have uh Unis pools that in times of high volatility they end up settling off chain at a really fast pace basically just for Bots or market makers you on your L2 you have a pool that you if the price on binance is a lot of volatility or something like that it turns into an L3 with one millisecond block times and you can trade super fast

against that and it collapses back down when when there's lower lower volatility and I really like that idea which is yet another data point to the model of like chains mapping to apps and vice versa like breaking down because you know you can have one one app that's going to utilize many chains like at the limit every single unit of pool should be a layer three or a layer four whatever you want to think about it right and one reason why you might want that separation where you have a pool like that has super fast block times and then a um near the L2 which has fast block times but not Turbo fast is because uh we're sensitive to the eventual decentralization of the sequencer of uni change and other chains

and the faster that you're going 100 millisecond block times 50 millisecond block times whatever it is the faster you're going the more difficult it is to decentralize so we um you know we feel good about having some level of decentralization with 250 millisecond block times and we are um trying to take it slowly so we don't build something that is impossible to do in the future I I can build on that as well so I would add that another reason why we didn't go below 250 milliseconds so far uh you know actually has to do with the idea that we still running an me auction it's just a really fast M auction that's

happening inside inside this you know inside these flash blocks as we're calling them right and so maybe to zoom out like what what flash pods is doing overall is we're running these me auctions which is this these like these explicit auctions uh basically for uh you know for me for for basic you know getting your transaction uh placed in specific orders under spe specific conditions in the block and so we're really doing a very specific like we're doing an explicit auction for transaction execution and so the idea behind this is that there are in in any

Financial system in any blockchain um there are there are basically always incentives to cause like a specific State change to happen because that that that like one state of the world is more valuable for you than another I to to give you kind of a very basic example if I can be the first person to Arbitrage this unisa pool against this balancer pool then I'm making more money and I prefer this version of the world to the one where you are the first person to Arbitrage that pool right and so we always you know we have people who want to get specific transactions mind and who have a financial in to actually you know uh you go and cause

that change somehow right and so as chain designers also increasing the app designers is and this I keep coming back to this because it's so important that this line will keep blurring uh like we have an incentive to our our goal should be to to make this auction happen as efficiently and kind of as you know less damaging uh as possible right and so this is what we have been referring to for a long time as the me trilemma which is just the idea that like your me will be extracted one way or the other you can choose how right if you you can do an explicit

auction and then people are bidding uh you know with G they are bidding basically with with like in the gas price right so they they are bidding to you know the the the bundle that's paying the most you know they get exactly their their wish is executed but other chains are making other priorities and and often like implicitly right because they saying we don't have M we don't care we don't want to like acknowledge this or Embrace this and then what happens is of course not that they don't have any M but rather that the m is extracted in other forms to Laten latency auction or typically a Spam auction and you see it on uh you know for example Bas you see

it on salana and these are of chains that have very low transaction fees um and what happens there is basically people they do what we call onchain searching so they basically write whatever their bundle would do but they write it in a smart contract and they just Hammer that smart contract every single block uh so they ping it repeatedly and every time the smart contract tries to execute basic let's say you know specific Arbitrage um and 99% of the time it fail and then you pay like a little bit of gas fees uh but 1% of the time it may succeed right making a lot of money you know and this is profitable for the Searchers that's why they're doing it but it's

terrible it's terrible for the blockchains it's terrible for the users because at sometimes you see like 30 40% of the overall gas of the chain is used up by these you know reverted transactions right and so you know if you have these non explicit forms of auction you know it can look good for you for a time in a debate on social media but it will always backfire kind of in the real world to give some data on this a couple weeks ago on Bas uh there was a a single Searcher that hammered the base sequencer 1.5 million times those transactions and they landed 10 of those

successfully they had a one in 150,000 uh rate of success on their transactions to try to extract me wow that's where the side of the of of rollup boost um comes in where basically there's is it is it completely free transaction reverts is is that the right so if you send a transaction as a Searcher you're doing your M searching on chain blah blah blah and it's not successful then your transaction is not included on chain and this is not just good for me but it is actually scaling infrastructure too because out of those

1.5 million failed transactions they're using gas in the block this bot got up to 20% of the blocks within base for an extended period of time and that 20% is gas that could have been used by users making normal trades could have been used by you know games whatever else people want to do on on base but instead it's just this spot that is turning through failed arbitrages space basically and it's even worse uh when you think about the fact that in perpetuity for the rest of History every single full node is going to need to store that state and repeat all those arbitrages you know the 1.5 million just so the Searcher could get you know these 10 successful trades out of out of that 1.5 million but with the revert

production that we offer those failed trades they don't land on chain instead they're just set off to the side and only successful trades end up being executed and and landing on chain chain which frees up gas for everybody else to to transact instead and we think this is necessary to scale these chains because you know maybe we'll get to a giga gas per second a billion gas per second but uh 990 million of it is actually just used up by me Searchers on chain trying to find arbs or or or whatnot yeah not exactly the Giga gas era that we envisioned and I think part of the the importance of this is the reality that users on l2s I think you can clearly make the argument objectively that

they're more likely to be uninformed flow than users on an L1 specifically ethereum whether it because of the size of of the trades or because of the type of user where they come from think about if you on board your friend through coinbase wallet onto base things like this so I think there's a real call to action and need to protect users on l2s uh going forward and so that is like the that is the core important that I see of rollup boost as a as a whole and so I think all these technical design ideas and and philosophy are basically oriented around that reality um is that like the right way to think about it or

oh 100% I mean this is kind of the other thing that we can connect it to right which is just like if you're onboarding to a layer two then you've only ever seen this side kind of of the like user experience of crypto right you you have much faster blocks um you never had to worry much about Meb because you have a centralized sequencer it tends to execute transactions in order at least it's not showing these transactions to anyone else as far as we know but of course this is happening in you know is this happening by Outsourcing all of this to a trusted third party right and I mean this has brought a lot of

newers into crypto and has really helped um like blockchains you know catch up to centralized exchanges terms of trading experience and so on and like app app developer experience but it does come at the cost like if you have this much better ux system how do you decentralize it right we have made these choices on the layer one for a reason right if we knew how to build a decentralized system that has one second block time or even less and that doesn't suffer from M and all of those nice things then you know we would have buil did a long time ago right and so it has been the case that you were able to make these ux improvements for users by sacrificing the decentralization we have on the

layer one and by instead putting a centralized sequence just a third party into this place and that's not long-term compatible with what we're trying to do here and so what do rups do right they have been put into this position where they are competing on user experience like once the first rollup started to lower the block time and put it put a centralized sequence in place it was like Off to the Races right like if you didn't if you didn't follow suit then users would just migrate to the like faster cheaper Layer Two that was quote unquote more centralized right and you know for a really long time I think these rups didn't really know they

wanted to decentralize but they were also forced by the competitive pressure in their respective part of the market to make these centralizing choices and I think this is where roller boost comes in which is I think it's for the first time giving you at least giving you a vision for how this stuff can be decentralized long term without sacrificing the ux that we have gotten these users used to I would argue making the ux even better yeah from the from the user perspective it's almost like the the boy who cried wolf like you know we keep going to where the best ux is we keep going to where the best ux is but then we only care about decentralization once it matters right once the wolf has

gotten to us um and so you know we keep like salana has been able to achieve fast ux 400 millisecond block times because they centralized their nodes in a in a geographical region same thing with with rollups we we go there because they have fast ux but they have centralized sequencers and so you guys have been able to to essentially put together this this this Paradigm where we can have the fast ux but it's not centralized it actually maintains decentralization um tees are like a very a very crucial part of that we haven't talked a lot about te so that's maybe a little bit more unfamiliar to our audience could you guys provide kind of a kind of high level explanation of tees

in general and then talk about the role that tees play in rollup boost and why it helps to maintain rollup decentralization while also providing fast ux happy to yeah so tees are trusted execution environments that's where the acronym comes from and they're really interesting because it gives us the opportunity to escape this kind of trade-off dilemma that we have between ux and decentralization by introducing a new point on the tradeoff curve that literally did not exist before there's no other way to get this other than using a te if you want to have decentralization performance and privacy

there's no other technology that can get you this other than tees and I'll tell you why um tees give you two really interesting guarantees one is on Integrity with a te you can actually have a guarantee on exactly down to the hash of the the code of what is being executed inside of a te you can verify that a certain program is running and as such you know how your information is being processed you know exactly how this uh tve is going to handle your information or your code or interact with with you um which gives you some really interesting properties in the context of of of crypto as an example

you know that this rollup is going to sequence your transactions in a certain way or it's not going to share your transaction before it includes it on change that's the first property that's interesting second property that's interesting is privacy so te uh they have a level of privacy even from the party that is operating the te if you if you have the physical te itself you you can't uh see inside of the te and take the um information that's inside of it uh outside unless it is programmed specifically to to do that and what this gives us is the ability to decentralize things that otherwise previously were um uh relying on some form of social trust to maintain privacy

that really matters in the context of an L2 because users want these private mols where no one else can see their trade before it lands on chain uh as such no one can confront run them uh the final thing that I'll I'll add here before P when I got to toasu is that tees uh do all of this with a minimal amount of performance overhead so there are other ways that you can get you know some level of decentralization guarantees and and integrity in privacy but they're pretty slow like uh ZK proofs as an example or fully homor encryption um but te by uh uh by using this form special Hardware you get close to no performance overhead so we've been

running main mainnet eel1 block Builders which is a really performance-minded use case in tees for some time and and in practice we're seeing something like 5% performance overhead from from doing so uh in return for these integrity privacy guarantees that that you you get um you have anything to to add to that imagine if the world's 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 to 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 multi-sig 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 welcome aboard on our journey into the singularity this is the endgame road map for fra and fraxel the new L2 out of the fra ecosystem that is totally based on fractal scaling fractal is a modular L2 blockchain built by fra with modular and fractal scaling in mind fractal scaling is based on the

mandal BR set of fractal images and it explains how l2s become l3s and the entire app chain life cycle goes through its Evolution the frax ecosystem is made of some of the top builders in the space and has stemmed from its OG days of the frax stable coin it is now in the LST space the LRT space and this frax L2 joins the existing Suite of fra products including fra Swap and frax lend you can see frax faery the bridge to go from any chain arbitrum optimism or main net into

the fra main net Bridge your tokens and bring your FRA assets into the fractal ecosystem join today to get the best out of fraxel you are still early join today take part in our Expedition into the singularity no I think uh there was spot on on terms of the the te I I think we can connect it to the idea of an M auction right so what what do te is provide it's integrity and privacy um and the ability for you like run code like untrust run untrusted code on like someone else's machine right which is extremely powerful what do me auctions need they also need integrity and

privacy Integrity why because a lot of auctions lot of auction types require you to trust the auctioner uh why and what does it mean it means that the auctioneer is basically not inserting their own orders or like reading into the orders of the participants or censoring any orders and that they're picking the right winner they're like Distributing the the good according to like a predetermined algorithm that's fair and that's known to all parties the other thing is is the Privacy right is that the idea that you know I mean that the orders that you submit the bids that you submit to the auction you know not only private from the operator but also private from the other participants right because you know otherwise you might bid for

something and you know somebody else reads the order and that slightly out bits you right like you for a good auction you need these two properties and the third one that we've already talked about is latency right so we know how to do Integrity in crypto we know how to do privacy we don't we but we don't have no idea how to put them together for what it's worth because uh you know privacy is very hard uh privacy over shared state is even harder privacy under low latency is a completely UNS solve problem in crypto because all of the techniques that we do know either not decentralized and they take forever like FH um so yeah I think this is why Really T are this perfect match for you know I was

not only perfect match it's really the only match it's really this like there's no alternative to addressing these problems without the idea of basically going to justed Hardware just to put a a final t on this uh my highle mental model is that because they have integrity in privacy guarantees tees allow you to Outsource computation to untrusted parties which in turn gives us the ability to decentralize components of infrastructure that previously required these centralized trusted actors like centralized sequencers got it um just gonna throw a throw a

curveball here um I was at um uh e was that Devcon and obviously uni chain was like a Hot Topic and I believe I was talking to one a rollup founder um ZK rollup founder um and they were particularly concerned about the trade-off spectrum of using te's to the extent of the reality of there being and I might I might butcher this you might have to help me there being a secret key that is stored in a private place that if the information is leaked and and you have some sort of use of force to obtain this key without

others knowing then you have this massive advantage of transaction ordering or you have visibility into what is happening into the te so I don't know if I got that right or if I'm not trying to incite anybody to search for the uni chain secret key or whatever it is but like could you maybe explain the scenario of um the security and the trade-offs of using the tees because I understand that that uh like the benefits now but I think what that comes with these benefits also comes with some different levels of trade-offs that I think we we should be transparent about happy to yeah so I think that's that's

absolutely a concern um tees are trusted execution environments and Trust comes from needing to trust the manufacturer of the hardware so in the case of using something like Intel sgx for Intel TDX is what we we primarily develop on top of there is a secret key that Intel is injected as a part of the manufacturing process and it does require some degree of trust in Intel um that is a tradeoff that in the short term we think is acceptable to make because of what it it buys you so today if you want to run a centralized sequencer in an untrusting

in environment uh and do things like MV auctions there's no alternative to tees you can't use zero knowledge proofs for that you can't use fully homomorphic encryption or or multi-party computation something like that there's no pure cryptography solution right now that gives you the level of performance that that you need and in in practice a party like Intel is sufficiently disinterested in the Ned Market or in the L2 Market that think it's it's not a credible threat today especially because Intel wants to sell these things right and so they would be more than happy to if uh crypto was driving the adoption of their hardware and they would even be aligned

with it not wanting to you know screw around with transaction ordering to extract $100,000 of Meed and blow up their next Golden Goose for this trusted Hardware um so there there is really this uh trust guarantee but we think in the short term it's something that is an okay tradeoff because it in turn buys us the ability to decentralize this stuff that is centralized to to today um the final thing that I would add here is is that we uh at flash Poots are trying to layer on defense in depth and so to the extent that we can we want to use both tees and cryptography to provide the maximal amount of decentralization that

we can in these situations and there's a lot of contexts in which you uh even if you can use MPC under your performance constraints or something like that you're going to want to constrain parties within MPC or zerge proof something like that uh with tees because it just increases the cost of attacking a system um in the abstract and so we're not purists we don't think tees are are sort of the only solution we do think they're the best in the short term uh and in the long term we think defense in depth you know is the best thing to pursue anyway yeah I would have three things to add to that so first of all using it like using a te for muv

related things is right let's take a step back there's different things you can do with a te that have different levels of risk so if you store your private key in a te and then the you know then the te gets broken into your private key can be abstracted and your money will be gone okay so in theory I mean what is a rollup doing right I mean Ola is basically like processing transactions of chain and then it's posting the data to the layout it has a bridge contract that is storing funds into it it's making a proof that the transactions were basically processed correctly and it's posting the proof to the lay one to convince the lay on bridge to unlock the funds when people want to withdraw like you could have a completely te based

roller right where effectively you're making a prove in a te and the te basically says to the L one I attest to you that that these transactions were processed correctly and this can be this has all these nice properties because it can be completely real time like you can really like Justin Drake is often talking about the idea of we can have synchronous compossibility if only we have two things sh sequencing and same block uh you know ZK proving right so you can actually convince the lay one bridge in real time that what happened on the layer two is actually valid from the perspective of the layer one with the te you can do that today like you can you know you can convince the the layer bridge that the the layer 2 transaction

was valid but at this point the whole bridge is basically custodia inside the te like functionally speaking right and so if the te gets broken into you that would be very bad and all of those user funds would be lost and so I would strongly strongly kind of discourage putting like a lamb on bridge effectively into a te today however when we talk about me then what are we actually protecting here inside the te it's not actually any user balances it is information right because a te you know like an like an M option cannot make a transaction that was invalid valid or transac was valid invalid or something like that right all

it can do is leak information about what people plan to do in the next in terms of ethereum 12 seconds in terms of un unit chain in the next 250 Mill milliseconds so what I am and other people are you know seeking to do in the next 250 milliseconds I would that is not a particularly I mean that is that has value that's why we're doing it but that is not disastrous if that gets broken okay and so um that's not that's like a really like if you want to use tees for everything this is a perfect use case to do it because the risk of failure is not particularly high so second to the you were talking to this rollup operator and they were like expressing concern about using a te

for security in sequencer right but what are they doing today it's all about comparing different options with each other like is it better than the status quo so what that I assure you what this rup operator is doing today is they are running their own sequencer how do I know that because every rollup is running their own sequen right yes and that's correct so if you if you trust yourself you know then using a te is always better than not using a te okay because it's way easier to break into a centralized server then into a te okay and with a te you can actually attest to your users that you are not cheating those users unless you are also in cahoots with int and that is a very strong promise to give it's certainly

much stronger than just the users having to trust you p and so so yeah okay I have one more on open Hardware if you don't like remind me to talk about that later but go ahead okay we we will remind you about open Hardware later um my question is is how how do we move okay so we have centralized sequencers now I think it's objectively uh an improvement to the status quo to run these centralized sequencers inside of a te um how do we go from centralized sequencer inside te to decentralized sequencer I think it it kind of depends on what context that you want um and what exactly you want to decentralize so in the case of uh a

chain with flash blocks 250 millisecond block times what we've thought about is reusing this set of uh te operators that we have on L1 through something called Builder net which we haven't talked about but we can if you're all are interesting interested in it uh but basically we we have a set of people who are running tees on L1 all all around the world and we thought about reusing that to have these parties rotate who is sequencing blocks instead of a te over time over different geographies in the same sort of way that that Solano does actually inspired by that so you'd have one operator they're um you know running the te sequencer for a period of some number of of minutes and then it sequ

you know the leader changes to a different operator who's running the same thing in a te just from a different geography uh with with a different entity and and that's sort of I think the very next step which gets you you know many different parties that are sequencing the the the te but still you're able to operate uh uh you know with really high performance guarantees the 250 millisecond block times I think that's the very next incremental step I don't know if hu has any other ideas to to add to that though yeah I would also connect it to build anet I think I mean build onet is our answer to how we seek to decentralize block building in the long term and then when you talk about Layer Two then the answer is just use

build on it right like today we just put the centralized sequencer in a te and that's the starting point that's how we onboard those folks and give them more decentralization for the same or better user experience than they used to but yeah the longterm answer is just go to build on it because then you not only you get the same things but you also will get interoperability right you will get basically sh block building so if you have shared sequencing that is only half of the equation you need shared block building so that the blocks that you actually build that then get mind on these chains they actually aware of like they're actually contingent on on blocks that are built on other chains so you can have great properties like uh like you know Atomic cross domain transactions and so on and um I would say you know the the

key thing really like we have talked about tees in in how to how to decentralize block building right and T is allow you to run uh run code on on on a person's machine you don't trust and this is really this is the base primitive if you want to decentralize anything in me because me is about basically like low latency computation of private data and so if you don't if you if you can't run that code if you can't send your information to someone else and have them process it in a way that you can trust then you can for sure not decentralize a you can for

sure not decentralize M options that's why as I would say as the base primitive of Builder net is the ability to upload your code your programs into tees and have them run inside this network where they can talk to you know the code that other people have uploaded and they can all operate on the same shared pool of transactions so today if you look yeah or it's get three years ago there was no there was no private members right it was effectively there was this big public member everybody was reading from it today you have I think 40 50% transactions are in a in different private M say

increasingly very fragmented and if you want to be a block Builder um if you want to be Searcher increasingly like you need to you need to buy transactions right you need to basically go to these different wallets and apps and and so on and you need to make deals with them to get the order flows you can even operate in this market at all and we think that this is a really inefficient for the market really inefficient way of operating and so we would love all transactions to be in the same pool again but without any of the drawbacks that the the public M Pool had uh so we want we want to have a big shared m not just on one chain eventually on all the

chains but you transaction should be private and anybody should be able to compute on them and if they are able to make these transactions better and kind of increasingly form those transactions into basic bigger and bigger blocks same as you know cop is forming like transactions into a bundle and then you know C bundle is backround an M blocker and then that bundle is being included by a builder who may like yet again reorder it in some way and like this how like the overall block kind of incrementally gets gets better and better right and so this is this is the process that we're trying to recreate and the Primitive that you need for that is TE and not just tees but the ability to upload your code to the tees in a way that other people can trust I

know it's not base sequencing but I can't help but like be reminded by this idea of if we just have this Global encrypted mempool for all l2s then it's like it looks similar to like using the ethereum validator set or in this case Builder net uh which is like a more distributed set of nodes compared to a centralized sequencer to basically create this more maximally and sufficiently decentralized rollup world so um very very cool and we have quite a few more questions about the specific Tech side on the road map but ha I just want to have you fill in that point on open Hardware before it gets left behind as we kind of look forward yeah yeah and thanks for the reminder yeah I mean um we were talking about kind of what is

the endg game around tees and I think Robert made the point very well that t are already if they don't improve from today they already part of the end game why because they basically have no substitute and even if other technology matures you know if you have fhe then basically you still need to split the key so you need MPC okay but MPC has an honest majority assumption the majority of key holders in MPC can always decrypt the secret and so you need at very least to put the the the different key holders into te so you can trust this process at all under like what we're trying to do here right and so in the end game already T is a big part but at flashboards and I would say we are part

of a bigger initiative to create what what is what called open Hardware like like open tees and it's this idea of looking in a highly systematic way at all of the different things that currently can go wrong about the way that tees are designed and and operated on and trying to systematically improve them through Hardware Innovation like I mean it's like like going back to kind of your Theses of p and so on that like we're only innovating in the world of bits but not in the world of Adams right and this is like genuinely very very hard stuff this is operating very much in the world of Adams um and it's so Quint quintus kilborn one of our senior

researchers had a very good shout out qu had a very good podcast on this that we can plug here the podcast is called into the bite code and he's giving the full rundown on on what we are doing with trustless tees and and he's laying out that there's basically three attack vectors so te can be attacked um on the software level can be attacked by the manufacturer or it can be attacked by the supply chain and like you can really in a kind of systematic way look at um uh yeah basically look at the current failure modes and then try to incrementally make them better and yeah I really encourage anybody to check out the podcast because um it's the best resource uh there's out there right now and I would say

this is I mean this is an extremely ambitious project that would take you know decades you know to complete but there's a lot of things that we can do that are Rec Milestones on the path there every Milestone is making the idea of tees kind of better and more secure and more available and more diverse and there's a lot that we can do as an industry so I keep referencing theidea of the ZK moment so ZK you know zero knowledge cryptography was invented in like the 1970s or ' 80s right it's this very old kind of form of cryptography SL encryption and uh you know it basically because there was no real industry usage

for it it hasn't really caught on it hasn't really improved uh for a very long time and then Here Comes crypto and all of a sudden there's this industry usage for this old technology that is almost forgotten right and of a sudden so much funding like so much intellectual Capital so much Financial Capital flows into uh you know Academia and and and kind of research Labs industry labs around the idea of ZK and all of a sudden we're making these crazy breakthroughs that are just blowing our minds about how much we can improve ZK how much we can improve the hardware also around ZK how much I mean it's now not crazy that in five years we're going

to actually have real time ZK proving in crypto which is a huge GameChanger and it keeps getting cheaper and cheaper and cheaper probably by like an order of magnitude every two years don't quote me on that but um and yeah I think what we need is basically we need to build enough momentum enough intellectual capital and enough financial Capital behind te to give tees their own ZK moment so we'll have a te moment and I think that is going to be such a huge level up for crypto um and I mean fortunately it's not only crypto right because tees have already gotten a lot of momentum from AI right I me there's so much money in Ai and AI

really needs tees because you know basically you know people don't want to trust uh you know the T oper sorry the the AI operator with their data with with their queries basically but then the operators don't want to trust um the user to run it Loc because they want to don't want to give away the training weights right which is really the secret Source in a lot of the AI and so you have this huge coordination problem and there's no way to solve it actually without tees very interesting yes so they are yeah listen to that podcast because it's fantastic and um yeah and so this for them also they are looking at te as the only

solution to this they have no other solution so all of a sudden tens you know hundreds of millions of dollars are flooding I think I mean already Nvidia has announced that they are working on their own te chip specifically for AI and so you know hundreds like of millions billions of dollars are going to flow into tees all with the incentive making them better of course crypto I think and me specifically may have slightly kind of different things that we want out of this and so I think it's going to be very important to align the different people who all want to make te's better uh and kind of you know align also around like what we need to make sure that all of the capital that's

gushing in from Ai and so on is put to the best possible use it's making te better are for everyone what's funny is that tees aren't new per se like they've been around but I think flash Bots kind of brought this Resurgence of tees back into like the Mind share per se so um props to you guys for kind of pushing that forward Rob I think you had some questions about the road map specifically some of like the the co-processing things the multi-m it feels like what this is doing is like the rollup boost road map almost is like going in parallel to like how teams should be like rollup operators should be thinking about their own road map so um yeah Keen to understand some of that yeah and one more one more question on Builder net and just kind of how you guys are thinking about the

decentralization of block building before we get into multi-m and some of the roadmap components there was a Divergence that I noticed between two very smart people vitalik and and I think hu you put it out but on behalf of flash Bots and how vitalik put out his endgame piece and it talked about how block building could be very centralized and that's okay as long as block validation was very decentralized but you guys didn't see eye to eye because from your perspective from what I gather uh not only does block validation have to be decentralized for a decentralized system so does the block production and so you know when vitalic put out endgame I don't think tees were as ubiquitous as

as they are now now but how how do you see things like how do you reconcile that and and generally could you kind of like walk us through why you believe that that block production must be decentralized yeah so uh first of all I think so balic has let's connect this to another post that balic has put out it's called on decentralization I think it's from 2018 19 it's pretty old but it basically breaks down uh you know decentralization into into logical logical centralization and decentralization and Architectural centralization and decentralization and so he's given an example of a system that is logically centralized but architecturally decentralized and that

is ethereum itself right there's a single piece of state that is replicated across all of these machines but everybody's working on the same piece of state right and I think you could argue with like rollups it's logically decentralizing but this is what is causing the mation also in user experience right so you can you can this is such a good example because you can see the benefit of logical centralization which like Solana is all about logical centralization you get these great like benefits in terms of the ux for the for the app developer and the user but you run into like these scaling limits right and then when you have AR when you have logical decentralization system can scale you know way better you get other benefits also like ethereum has in terms of like

Innovation like every rollup is innovating every rollup is Raising on money andir on users every rup is different from every other rup right so it's like Justin Drake's Justin Drake calls it like ethereum is like BD Department right and I think he's not wrong about that I that is definitely the bull case for ethereum kind of rollup uh Central growth and so when vitalic talks about the end game he's basically laying out two things so he's saying that block building is very logically centralizing and so that means that at the end game like over time block building is bound to happen in one logically centralized

entity and so there you know why why is it because for a lot of reasons like having orderflow B having more order flow there's economies of scale there's cross domain MV like sequencing all of these chains in the same auction uh the barriers to end just get higher and higher and um and this like we completely agree with this flb has put out many posts arguing that block building is very centralizing and we need to act uh in order to protect you know to to take certain actions in in the supply chain to keep it more decentralized and work towards this like norstar making block building decens right so we're completely seeing eye to eye with them on that the what he's saying in his post

is even if even if this happens and uh block Bing basically is architecturally centralized also etherum is not because that that is what we this is the Baseline assumption like block building is going to be logically centralized but if it's also architecturally centralized etherum is still not doomed is what he's saying right it's not great but there's things we can do we con constrain this entity with things like inclusion lists and very cheap validation and and stuff like that right and completely agree with them I think we are more bullish than he was at the time right at least um that we can also decentralize block

building architecturally and yeah we're starting to have pretty good arguments I think around that um I think you know the idea of Builder net and the rise of tees and uh yeah just the idea that you can have a kind of Market driven system where people are incentivized to upload their Innovations their Alpha uh in order to search on a pool of shared order flow I think this is a unique Insight that you know just didn't exist uh a couple years ago and the technology to do it I mean it it you know it sounded more like magic to be able to do that right you just tell me everybody can upload their own source code and they're not giving away their intellectual property and

they can all search on the same orderflow three years you know we always thought you know that's what we need and the only way to do it is TE so like flashboard strategy in that sense is like very simp you know like make te better and then like build the system where this kind of decentralized block building can happen right um but yeah this is how I would you know kind of consolidate these different views one thing that I would follow up and and add to that and connect it to some other topics within the space right now is is that basically everybody whether that's falic or or us or um even folks like M Max tnik uh uh seem to agree that you want to involve as many parties as possible within block production and

that's kind of a overall umbrella that unites a lot of these different directions and so in the end game italic uh puts forward the idea that okay maybe block production is centralized in these big Builders or or sequencers but then we constrain them by having the proposers play a role too through things like incl inclusion lists and inclusion lists are just getting a new actor involved in building a block um by having them propose a portion of what ends up being included by so you go from having this kind of monolith where it's only the builder getting a block to now it's it's two entities Al be at the proposers are doing a small part of it and we think we're we're even taking

that idea to its logical extreme where it's not just uh the proposer and the block Builder but it's many many many many different parties as a part of the block building process contributing to how block is built and that's where you both get the most efficiency and the most interesting features the the most value back to users and proposers but also where censorship resistance will will come from in the end and uh that's all United by this idea let's get as many people as possible involved or at least more people uh involved in the process of creating a block I don't mean to just completely play devil's advocate here and you know we we we've only got about 10 minutes I'd like to talk about

the road map but couldn't it also be argued that by adding more players into the block building process you just consistently are adding more centralization vectors throughout that process while you're adding more entities to the production of a block every additional entity comes with its own set of trade-offs or its own set of vectors of centralization like wouldn't it maybe be better to think backwards on that or what what would be your kind of rebuttal to that claim What would thinking backwards mean maybe before we like basically instead of trying to add more entities to the block building process why don't we go back to less entities and deal with the trade-off set at that at that stage

versus trying to add more and then dealing with either more centralization vectors and trying to patch those or even having unknown ones that we we can't foresee because of the additional players in the block building process yeah uh I think it's a good question right so I mean I think what you described is like you have this one you just like this one entity that's basically building blocks right that's uh you know it has its algorithm and as long as it has you know enough orders basically then you know it's going to build a it's going to build a block that maybe people are like somewhat happy with right and if you have some if you can actually set your mind on something that you think is simple enough like that hits a good balance between like

simple and efficient it's much easier to decentralize them and so we earlier mentioned you know Max Resnik so I mean the idea of this like multiple concurrent proposers you have these different proposers they all run basically Forks of etherum like Shadow chains and they all basically uh run like a very kind of um a deterministic block building algorithm I would say that can be very simple it basically just or transactions by by gas price and as long as there's you know one I don't know how I don't works exactly but it's like creating censorship resistance as long as like some number of parties has basically is like engaging in like the

the honest ordering of these transactions um so this would be an example of like almost quote unquote like enshrining proping right we talking about enshrining like PBS and you can go a step further you can even Shrine like if you have censorship resistance so if you can start from if you can establish a shared list of transactions that everybody's starting with every everybody seen the same transactions then you can force the node to construct the same block from that by using the same algorithm because everybody can like rerun the algorithm and see if it gets the same outcome and if it doesn't then you know the block is invalid right and so that would be like an example of doing that and so this idea that you're proposing it exists it

has intellectual Capital behind it right maybe less so now that you know Ma is working Solana but uh like in the that is so but the idea is is fair right like people are like seriously like considering it so the reason why I would say we find ourselves on the other side of the spectrum is that when you are giving by making the auction less efficient for the participants you are giving actually you're giving up more than you think so this goes back to the very beginning of the conversation which is if you don't if you're not running an explicit M option you're going to run an implicit one right and so you go back to this very

simple transaction ordering algorithm and while you technically think it's fine because it does arrive at some moding and it has some nice properties for apps and so on but it gives sufficient room still for Searchers to express their views about the state of the next block through uh latency auction spam auction and you can be sure they will use these things and they will drive centralization in ethereal and they will cause negative exality in ethereum this is one side of the argument and then the other problem that you have is when you have a very rigid system the M Market is changing all the time it has changed many times like even

in the three and a half years that I've been working at flbs and um and so you know like the system needs to evolve it needs to be Market driven it needs to be able to incorporate new information new techniques of better serving users better serving validators and if you can't do that then it will be very hard to build a system that can be robust to disruption um and so I think those two are kind of the starting point for why we think yeah like the system needs to be inclusive of many different parties it is to prevent you know like it it needs to be compx enough to support an M auction but it needs to be Market driven

enough to basically survive for like multiple rounds of the meta game but do you have anything to add I would just add that uh I think we can do this in a way that doesn't add a bunch of new uh roles or actors in The MV supply chain I think Andy what you're touching on in part is that the MV supply chain just seems to get endlessly more complex with like different types of actors boxes that are stacked before each other and what I would say I was on a I was hanging out with a a crypto engineer girl and we we having some drinks and she was questioning my technicality about whether or not I'm a good podcast host and she was like describe to me the me supply chain and I I was struggling a little bit I know she

was laughing at me so this is this is where the question stems from i' would say yeah yeah yeah you got to study M before you uh go get a drink with that crypto engineer girl next time sounds like you might have embarrassed yourself a little bit [Laughter] anyway continue there yeah I mean uh so where where introducing these entities come from is because people don't trust each other in the M Supply Chain um so block abiler will have a really valuable block but they don't know who a random proposer that's spun up their validator last week is um so how do these two parties get to work together well they use the intermediary of a relay who acts

as trusted third party that mediates between the two but if we have some form of trustless collaboration between parties then we don't need these intermediaries and if we have some level of programmability uh uh an open competition then you don't need to just keep on layering on new actors in The M Supply Chain so we're really excited uh about this idea of letting people upload their private code inside of a t and being able to give some kind of trust guarantees um to enable collaboration between untrusted parties right now the primary way to do that is by introducing new boxes that then get you embarrassed

when when you go and uh have a drink with a girl at some point um so we're we're not introducing more more boxes in the M Supply Chain we're we're trying to enable uh more open permissionless Innovation within the existing Builder role to collapse all the boxes down into one uh while still letting untrusted parties collaborate together uh because of the guarantees that you get from T um you may end up just with one big box it's a decentralized network one big box of many small boxes all running on the same decentralized substrate for private Cod execution something like that yeah that's that's what my answer is going to be next time ay thank you for that guys

I I I mean we got just a couple minutes left and we wanted to talk a bit about the road map for roll up boost and where you guys are headed um you know there's two boxes that are checked here the 250 millisecond uh flash blocks uh and the second box is the verifiable priority ordering and then there's a lot of other boxes here one of the that gets us excited it's multi-m rollups others like uh inclusion guarantees parallel execution are also pretty exciting maybe you could just you know touch on a couple of these that if you feel are most relevant or coming the most quickly or or the most important milestones in this road map why don't I touch on te proofs and you do the mtiv them yeah so uh one of the things that you didn't mention that that I'm personally very

excited about is the idea of te proofs and I want to follow up on it because I was riffing on that a little bit earlier right I was using the you know te proofs as an example of something you should actually not do like put a bridge like put a secure rollup Bridge basically unlock it through like a te proof right because if you have that only then basically your entire Bridge deposits are secured by the te how ever if you kind of look at where the rollups are today rollups have you know layer to beat and balic have put out this this grading system for rollups right where if you want to actually go to to stage two right you need to have made your your proving system permissionless right

and um and I think you there's something about a security counsel I think also so the idea is basically that the the the the the Pro that unlocks the bridge it should be permissionless right it should not be should not be controlled byring te and turns out that's surprisingly hard to do uh and it's it's true for both actually the fraud proof and the ZK proof uh type of system for different reasons but um what kind of you know what they both have in common is it's very very new technology and it's very complex so the oldest fraud proving system avrum has only been live for one year uh the uh you know the oldest dek

system you know even less so and uh ZK map is very hard to audit and um you know we were quite far away from actually being comfortable uh you know locking like you a few billion dollars behind you know this code alone and so we just haven't seen enough edge cases we just haven't seen it break enough we don't have enough Lindy effect stacked on top of these solutions to really trust them that's why it probably take some time for these rollups to to to kind of fully mature and actually become fully non custodia from the perspective of the layer one and that is holding back adoption of these systems right because you when you're depositing These Chains you have to trust the operator

right that's like this is not what you want and so this is where te boost comes in sorry this is where um te proofs come in because the way that we suggest using it is not actually to replace your existing optimistic or ZK proving system but actually use what we call a multi-prover system and you can think of a multipro system is where the bridge is effectively not you know you have a you guys probably if you're doing a lot of onchain trading I hope you have you know a save or an Argent wallet um and the idea behind that is just that you know you want to have uh you know a tofa uh like a you know a two of

one so not two of one one of two or two of two or two of three you know some kind of multi signature setup right where you need at least two signatures uh to unlock a transfer and you can apply the very same concept to Bridges right now a bridge is a one like a bridge is a one one of one multi basically right but we can make it a two of two multi if you add another proof so you can you can use optimistic and Z proving and then you probably have the risk so uh you know you can design a system that the way that you need the the optimistic proof and the Z proof to actually agree for the bridge to be unlocked and the funds to be withdrawn

and if the proofs disagree then the governance can kick in uh but you can also use a te for the same purpose right so you can you can have like a combination of an op approver and the T appr or zik approver and a t appr and what you're what you're getting through that is you're getting the additional security you know it's it's purely additive because if the te proof is wrong it doesn't matter unless the op proof is also wrong and they're completely different implementation so there's like no nothing that can cause them to basically fail simultaneously that is where you can blame the te for that basically right and so it's a way of making these Bridges you know better and Elevate rollups to stage two faster

so we can you know they you know they can basically level up the security guarantees and get more users and get more TV and so this is something that yeah we are very excited about and yeah that we're working on with several Partners um and yeah um basically hope to uh have something here within you know the next one or two quns and then BT um what about the multiv BM multiv BM is is very interesting the notion is what if you can have sequencer in its te running an alternative VM uh kind of as a side car and then let users on chain interact with that VM um in the normal way that they do uh through smart

contracts that tell the sequencer how to interact with the alternative VM that may have sent it a little bit convoluted but as a user you interact with the special smart contract onchain and then that tells the sequencer hey run this program inside of the alternative VM that you have and that would allow us to uh that would allow rollups to give their developers new tools to build new types of applications like deploying your rust coat in a wasm VM that is running as a sidecar next to the the sequencer and and enabling a whole set of applications that right now you can't do because you're constrained by solidity um and the more general idea here is how do you use these tees and

the sequencer as a a co-processor do more interesting offchain computation to give developers superpowers that they don't have right now um one other idea that I will throw out there that that we're excited about is what if you could allow developers to call a special smart contract that makes an HTTP request uh and then the sequencer can return the result on chain and um in a smart contract so that it's available for developers you have a smart contract could even return it in the same in the in a transaction that you have yet to process right during the block building process right so yeah the way it could

work is you're making users making a transaction uh to a Smart contract that has this special you know HTTP uh request function that is called and the sequencer injects their own transaction before the users to provide the result of that HTTP request so then it's available for the transaction when when it executes and you can do a bunch of interesting stuff like that like you could have your smart contract calling out to chat TBT uh uh uh and then build apps on chain that have access to chat GPT um so we're super excited about this idea of how do you use a secretor as a processor to enable the next generation of apps I'm more of a claw guy myself but you know to each their own

sir yeah yeah we can uh make it drag and drop Claud chbt grock uh here at flash Bots we support all llms we're not waiting agnostic to the llms that's that's bullish flash Bots we're hedging our bats you know we don't want to play any favorites just in case one of them wins you know you never don't piss off the wrong ad yeah the one thing that I want to add I know that we're coming up on time but but on on the road map is that we're really interested in um pushing the boundaries of performance for rollups and making it really easy to be able to do so so the first thing that we're doing there is Flash blocks again this

is scaling infrastructure because you need to clean up the toxic waste that is me onchain spam or an onchain option but on the the road map I think you're hinting to stuff like parallel execution where we want to get even deeper inside of how the chain works and enable things like a parallel evm uh uh just through an extension um that you can access through flash blocks or things like on demand t3s that help with your chain congestion uh um you know as you need spin spin these things up up and down and is that the example that you were pointing to earlier with Unis swap collap like in times of high volatility spinning up this kind of like layer three environment and then collapsing back down is that an example of that yeah exactly because if you are on a rollup

today and you know there's a lot of volatility on a major pair e USC or something like that that you're going to have a ton of sexx Barb that's going to use up your block space and then suddenly people who are playing a game in some other completely unrelated part of your block space have higher fees and that's not a great user experience for them either so if you can have these on demand t3s we can both offer people uh who you know are targeting really congested blocks Bas super low fees you can also help to isolate fee markets uh so that you can maintain the user experience across these different things that ultimately don't really really matter um while keeping them composable in the times that you don't need these on on demand yeah and nft mint are another

great use case for that um exactly yeah cool yeah we keep coming back to I mean I was saying if if we if we can have the te moment for crypto it's going to change so many things terms of the security but also like the expressivity of the chain at user experience and the privacy and and those things and it's I this is what this is really I think what we're trying to give a glimpse of with the introduction of roller boost is that you can actually like if you are willing to look Beyond software Innovation if you if you're willing to leverage Hardware that like runs alongside you know your regular system that there's so many things that you can

do and um yeah I think this really can change a lot uh in in crypto crypto's te moment I think we've just found the thumbnail title thank you hu and thank you guys for joining today's podcast it's been an absolute pleasure lot a lot of technical topics I think that we did a really good job of explaining them uh because I understood about 95% of it so that that is a good sign um guys thank you so much and it's been a pleasure yeah all ours thank you for having us cheers