Avail: Unifying and Scaling Modular Execution Layers | Prabal Banerjee, Avail CEO

Transcript

we will use an election algorithm called fragment election which has a property called Justified representation of Stak what it does is that no matter how much value you come with you will always be chosen in a manner which maximizes the decentralization of the network what's up everyone welcome back to another episode of zerx research March is just around the corner and I want to make sure to give you a quick reminder to not top tick your prices of your D London tickets if you use codes 0x10 at checkout you can lock in a 10% discount on your ticket don't miss out on your chance to get ahead of the curve I'll see you in London before we get into today's episode just a quick

disclaimer the views expressed on this podcast by either myself or any guests are their personal views and do not represent the views of any Associated organization nothing in this episode should be construed or relied upon as Financial technical tax legal or other advice all right let's jump into the episode welcome back to another episode of zerox research we have a great episode lined up with FAL co-founder at ail which is a robust based layer designed to meet the needs of Next Generation trust minimiz applications and Sovereign rollups today is Tuesday February 13th as a reminder be sure to use code 0x research1 at app. blockworks research.com for 10% off your annual subscription on to the interview with foral thank you so much for coming on

today thanks a lot for having me R and Sam yeah could you start us off by outlining what exactly is ail so at a we are trying to find the ways in which the rollups and the rollup Centric road map can play out and one of the basic building blocks over which rups strive is data availability so our main core offering is the data availability Focus blockchain that we have been building for some time now and using that we are trying to reimagine how rollups can scale blockchains gotcha appreciate the

succinct info um I was surprised to learn that ail uses babe and Grandpa for consensus which originated in the polka dot SDK could you walk through some of like the do considerations that went into that yeah absolutely I think you are one of the first ones to straight dive into B and Grandpa because uh uh it's one of the of the code design choices that over which we build and it's uh the the main thought process which went behind uh choosing babe and grandpa is because we definitely wanted a hybrid Ledger and what that means is that any there's a typical dilemma in consensus systems

where you cannot have liveness and safety both together at the same time and so many system systems they tend to prefer safety over liess for example if you use tendermint and instant finality chains which work on top of tender they choose safety overlin so every block needs to be finalized before the next block can be produced what that means is during periods of network partition or if there is a large amount of nodes which go down the chain might suffer and that's why you will see on the other hand other systems like the P chains

like Bitcoin and ethereum prev before the merge they used to you know typically favor liveness over finality so it was all a probabalistic finality the chain used to survive everything you can reduce the hash rate you can increase the number of nodes it survives everything right so we were thinking about how exactly to go go through this dilemma because we come we we started building this from polygon and in polygon there was this the same dilemma playing out day in day out there was a two- layered architecture one with bore and another with hindal where Bor

used uh like used to favor liveness so it was all probabilistic and then heimall used to give deterministic or probal finality and we wanted such a hybrid Ledger but with better security guarantees like that of a vrf with nominated proof of stake and uh a fragment election algorithm for more decentralization and such and then we really liked how Grandpa you know finalizes over multiple blocks at the same time chains rather than uh individual blocks and those are the things which you know attracted us towards this design and using babe and Grandpa as babe as the block production engine and then Grandpa as the final

Gadget much like ethereum has today with Gasper I wanted to follow up there quickly for a DA layer like if you had to choose between the two safety and liveness which one do you think is more important that's a really good question like you will have to I will have to essentially talk about both in separate orientations and so it can be a bit long- winded but I can give a c sponsor to want um if we talk about Sovereign uh rollups but rather than sovereignity it's about based rollups where users send transactions directly to the base layer in those base layer

constructions liveness will be very important because the users inclusion of transactions comes from the inclusion and the block production engine of the Bas layer and that is the times when liveness becomes extremely important so you need need to give at least a soft guarantee on The Ordering of the transactions otherwise the RO ofs on top cannot survive on these orientations like this becomes important but other than that safety is extremely important because finalization is over which the rollups who batch together through a sequencer will will Tri on right so for example if you

see um op stack or an R orbit chain or or polygon ZK VM St Madara zkn hyper chain all of these have a sequencer at least one or many sequencers on top which determine the ordering so you do not need ordering guarantees from the DNA and hence the DNA only is relied is relied upon for the finality guarantees and that's where finally it becomes important so you somehow need a best of both and even with those kind of systems there is

there is hybrid constructions for example in arbitrum there is an inbox smart contract on ethereum which can act as you know the source of force transactions and in other constructions as well they might want to choose the base layer or the DNA for Force inclusion or Force transaction acceptance and that's where liveness again becomes important so sensorship resistance comes from the base layer and there liess is important finality comes from the safety arenes of the base layer which can be abstracted upon by the rollups and again sorry again for the

longwinded answer no I think that was a as as one could answer the safety versus thess question I think another interesting design decision that a bill has opted for is nominated proof of stake which you don't normally see that much compared to just normal Pro stake blockchains or sort of pro stake blockchains of native delegation so could you talk about the benefits of a nominated proof of stake model and some of the tradeoffs potentially that come with it yeah absolutely so there are many ways to look at it but the main thing here is St distribution and centralization vectors so for example if you think about a typical

proof of Stak which is a delegated proof of Stak mechanism there as a delegator what you go ahead and you choose a validator against like for whom you delegate your stake to now while choosing the like this by choosing the validator for whom you delegate you will have to think about your return on investment right that's the most incentive compatible way of choosing whom you stick for and then for that you will always choose the highest amount of stick because they tend to gain the most number of slots and in turn the most number of

rewards and hence you will see in major dedicated proof of stake mechanisms there is a centralization risk the top 10 even top three sometimes control the chain in the sense that there was a recent cosmospace chain where a single validator held more than 33% of the stick and that stopped and hence the main it completely hunted and this is not to say that it's a problem with with the design or it's not a problem with the chain itself or the functioning of the chain it's about incentives right and that's where nominated proof of stti comes in and says that we will use an

election algorithm called fragment election which has a property called Justified representation of Stak what it does is that no matter how much value you come with you will always be chosen in a manner which maximizes the decentralization of the network of course I'm simplifying it a bit there are two or three parameters over which it optimizes it's a NP complete uh algorithm that's why there's some approximation and some bounds on the approximation over which it works but roughly speaking what it allows us is for every nominator to come in and give a prioritized list or validators over

which they want to distribute the state to and over that it can support a large number of validators a large number of nominators it can even support nomination pools and so on so there are many different you know Concepts which come together to make sure that the chain not only has a large and a validator set that's down to the consensus system the grandpa finality Gadget that we discussed before but also that a large set of people can come in and nominate in a fair and decentralized manner so that the decentralization of the system the objective functions like the Nakamoto coefficient and such remain

at the optimal levels throughout the design awesome that's super helpful thank you pral I appreciate that um in terms of uh how aail differs from other da Solutions so you've got igen da which is just not its own blockchain it's a committee based you know honest majority assumption and then on the Celestia side which is maybe more kin to aail they're using fraud proofs as you guys are using kzg commitments SL you know validity proofs can can you explain why you made that design choice and maybe the trade-offs and benefits that come with it yeah absolutely so again I will go down to history because that will kind of that has dictated our design

decisions for for a long time so from from very early on polygon had heavily beted on ZK rollups and validity proofs and when we were starting to design a from within polygon which we spun off last year so when we are designing it we didn't want to rely on fraud proofs because what that what that lets you do is that if you want to assertain the state of a rollup you need two sorts of guarantees one is a guarantee of the correctness of execution and the second is the guarantee of data availability these two

things together guarantee the correctness of the rollup now if you think about such a construction where you have a ZK rollup on a fraud proof based base layer then for the correctness of execution bit you have a validity proof or a ZK proof right but for the guarantee of data availability you are still relying on fraud proof and hence they will be a challenge period for which you have to wait and so on which makes no sense for creating a ZK roll up on top where you have already taken the design decision to avoid fraud proofs to avoid challenge periods to go for composability and so

on so that is from where we started and that's where we chose kerg political commitments and iner coding along with that to make sure that we can avoid fraud proofs so that's the FR proof aspect the second is that we thought about data aity committees but then we realized that you know a committee can never give the security of a base layer which can be as decentralized as we can be and again this boils back to the decentralization topic and the nominative proof of state that I was discussing before if you if you can make a decentralized base layer

which acts as a source of security of the ruup then you would want it to be as Sovereign as possible and as crypto economically guaranteed as possible Right without having biases towards one ecosystem or the other so one of the main reasons why we spun out of polygon was also that we didn't want to be you know biased towards any one of the rollup ecosystems we wanted to be you know credibly neutral we wanted to be completely independent and that is one of the reasons why we spun off and those same design principles and design choices also show up when we talk about why we are building a base layer rather than going for a committee and

then there are other terms there as well for example on a on a blockchain of your own not only there is cryptographic guarantees which we give through coding and validity proes through K commitments but there is cryptoeconomic guarantees which is what does it for you to corrupt 2x3 of the chain but then there are also social guarantees which means that even if the 2x3 plus of the chain chooses to get corrupted still there is a social layer which prevents the chain for going complet C right so that is one of the reasons also why it is important that a a base layer remains independent remains

credibly neutral and remains decentralized and that's why we are aiming for and then this is more of a quick hitter but I noticed that your block times are 20 seconds versus celestius 15 is that just because of proof generation times or why was the choice for 20 seconds made that is not nothing to do with the proof generation times and just for like reference we have public benchmarks but uh I think last I remember it took around less than 6 seconds to generate even 128 MB uh blocks and we are only doing you know 2 MB block sizes right now there are two parts to the story the

one part is why do we need or what is the relevance of the block time on the D as and this binds back to the questions that you guys were asking before is that what is the is liveness important or safety is important and the block time comes with the liveness guarantees so you would want the base ruup constructs to to get those soft confirmations very early and that's why a lower block time might be a bit important but if you see there are these you know you can always badge them up there are shared sequencers there are

in general sequencers and then there are you know of on top which are bundlers and such which always operate to give you those soft guarantees even before block times can appear and that's why it was not relevant for us to have you know super low block time like 2 second 5 second or something and I think the same things apply for every other DNA like they are not talking thinking about the shortening of the block times very aggressively because it gives them nothing extra the other part part of the story why we took 20 second itself and not something like 18 seconds or you know 10 seconds is the fact that we want a

peer-to-peer overlay network of light clients having all the cells available without Reliance on the rpcs now this boils back to the you know the decentralization aspect of it where we want our light client to be to be as compatible and as capable so as to have the data availability guarantees not only verified by themselves but also being able to contribute to the data vality by themselves so what they do is they not only sample this is a power of data sampling which I can come on later but they not only sample the data but they

also keep the samples available on this overly peer-to-peer Network now during our stress testing we we thought that under a sufficient amount of block size the current peer-to-peer status quo is not you know mature enough to populate that DHD within let's say 10 to 12 seconds of time when there are you know uh some hund like 200,000 cells that we have to populate so that's a lot of entries for the DHD and during our stress testing we realize that if we Hammer that too hard the knit

peerto peer and so on because they are alpha and beta level software for now we will not be able to give very high guarantees unless we bring in additional resources to the P like for example we can have not only our validators and like clients but also some other clients inside the peer-to-peer Network who can populate the DHD he can have specialized roles which optimize it and so on but in order to avoid any Reliance on beefed up resources or to avoid any Reliance on centralized entities we thought that it would be best to not only give those guarantees from the peer-to-peer n

itself but to have a very liberal block time so that we can you know populate that DHD without much stress gotta appreciate the in explanation there I wanted to move on to sort of trust and security for a d right um I think that's probably one area that not as many average P22 users understand so what type of trust or security assumptions are inherent to Avail and perhaps we can talk about it specific specifically in the context of a data withholding attack yeah absolutely so I think yeah data withholding attack is a very misunderstood or seldom well understood kind of a problem right so the attack

comes to the front where when there is a block producer or a publisher let me call about talk of them as Publishers when a publisher tries to withhold either the entirety or parts of the data in those kind of scenarios the rollups which are building on top of this data they tend to be insecure again insecurity or the security has to be defined so for ZK rollups it can mean like this because your you know accounts can be Frozen and on optimistic it can mean security because you cannot be able to generate fraud groups and such but

there are plenty of problems which appear when data gets withheld and in terms of what ders typically do and this goes Beyond just a right ders in general what they are meant to do is to make it very very hard to withhold data and make it very very easy to catch hold of withholding of data now let me try to talk about both of them separately what does it mean to make data hiding difficult what we do is we provide redundancy in the data but for example if you have compact discs I

guess we we all here have had the luck of you know checking out compact discs and such so if you take out a a CD and you had scratches on top you would still be able to play the audio or the video that was there and this was due to the fact that there was redundancy inside the encoding which made it uh made it okay for it to lose a few bites here and there the same principle we used in a reverse manner where we use the integer coding to make it hard for people to withhold parts of the data so to now to withhold even a small fraction you have to now withhold a large fraction of the data because there was

redundancy so this is how you make it hard for people to withhold and then the second is how to catch fold that someone with h data in typical blockchains all the full noes download the entirety of the data and then only when they have the full data they agree that this is a complete block the light clients they have no guarantees as such because they are not downloading the entire block and that's where they get left out and they have incomplete guarantees so you cannot take a blockchain which is not purpose built for da and you can assume that it has the same guarantees because because to

get the same guarant guarantees you have to Hast full nodes what we want to we wanted to do is to give the same guarantees but to light clients and what we do is there we use data availability sampling so what that means is the light clients they do not download the entire thing but they download partial samples and by using those partial samples they try to get a high enough confidence that the entire data is available because there is redundancy so that's the two parts of the story about data withholding attacks now coming to you know the trust assumptions and such with light clients with validity proofs we have

reduced the trust assumption from let's say an honest minority where which you need for fraud proofs and for a honest neighbor assumption which you need in peer-to-peer networks where FR proofs dissipate right so for example when you have a a fraud proof you need someone honest who will generate that fraud proof and then you need a honest peer to propagate that FR on the other hand because we use validity proofs we have we have been able to cut out the fact that we need someone honest to generate the proofs because everyone is able to download the proofs and verify that the only other thing which we cannot remove is the fact

that one single light client is never enough you need to have a good number of light clients because the probabilistic guarantees that the light clients have the sampling guarantees that the light clients have comes from the fact that there are many people like that who are doing the same amount of sampling right so there needs to be a critical mass of light clients which needs to do the same amount of sampling for individual light clients to have the same guarantee because it's very easy easy for an individual light client to be fooled again if it is a very long answer I can repeat bits and parts of no that was super helpful context I

appreciate it um from a higher level like I've been seeing a lot of people say that da layers are going to be commoditized would you agree or disagree with that statement I think I completely agree with that statement like the the way that things have been moving today um it is very evident that you know the da costs are not only getting subsidized but to the point that it is right now raised to the zero and you will see absurd numbers being fled around by different new entrance about how much cheap it is and how it's almost negligible and close to zero to be able to and and it's very clear that there

that is not a sustainable model right for a d to have and hence it will very quickly see that the DN will be commoditized and that is exactly why we think that it's not about the data availability cost which will actually drive adoption it's actually the guarantees that data avity gives you now why should someone use a rather than using then local storage why should someone use a wi rather than finding five different vendor getting into a contractual agreement with them and using them for da right those kind of questions need to be

answered for for people to realize the the power of Da and that's where you know the guarantees that da layers can give to rollups on top in order to interact with each other in order to interact with other ecosystems those are the things that will come to the fora and those are the things that will you know differentiate between the different dealers yeah I think we're all definitely in agreement with that then I think the next natural question to ask is what do you think comes next for da layers right do you think they eventually vertically integrate to some other part of the stack because the value of crew as you mentioned right has been relatively minimal and will probably decrease even more going forward right now D layers

today just do da right um but with the data posting cost being like either subsidized or so cheap these D layers don't make enough money um to be frank and chances are they're going to have to vertically integrate for example facilitate in operability somehow or launch some other like part of the modular blockchain stack or like a roll up as a service whatever you want to think of so what do you think these da layers like eventually expand to to hopefully increase their value of crew or the profitability no I think the the the real mod here is the ecosystem right you need to build an ecosystem on top in

order to in order to be able to you know Drive Traction in order to able to be Drive activity and that's where your real you know adoption will come in in the matter of whether they will do some vertical integration with other parts of the stack maybe they will right maybe there some of them will actually go ahead and make something which is very specialized but the fact is that there will always be two ways in which ders can be used one which is very unopinionated which is extremely uh you know simple that you P push and publish a blob of data and that blob of data will be kept

available that's the basic service that every D ler will have to provide and there will be almost no D layer which will say that no you cannot be this simple you have to also it's part of a package and you'll have to take XYZ out of of course there will be add-ons on top there will be you know favorites clear favorites in ecosystems and this has already shown up right for example if you if you are thinking of doing a swap in ethereum ecosystem you will find that you know Unis swap is the is where most of the liquidity for most of the asset classes on there right so that is

a natural choice and this is not like you know ethereum has you know somehow said that unisoft will be the only only place where swipes can happen but this is General because unisoft has been able to gain that traction has been able to get all that liquidity has been able to C cater to all that users and so on so that's you know some of the natural uh design choices and the parts of the stack which will you know Bubble Up in due course of time and we will keep on seeing that whether ders will endorse some of that or not I mean it's hard to

say maybe maybe that those are the differentiators that will come up as well like for example I can only talk about a wi and there we are deeply thinking about interoperability we are deep thinking about like making this fragmented world of you know thousands of Ro apps of a more coherent more composable world right because we are convinced that there will be tens and thousands of rups and if in that world we don't want to have the same user experience as a normal user today has across let's say stet uh po on

zkm uh ZK sync era Arbitron one and some of the Bas base chain right so we we don't want the similar fragmentation that we are seeing today to also happen to all the rollups which uh you know like stay on on top of a whe and that's why we are deeply thinking about it of course it will no matter what we come up with it will it will always be a choice for the rollups to either participate because sovereignity is something which is very close to our heart and I'm I'm very sure that none none of the rollups uh will like will want to give up their sovereignty um and

that's why it will be their choice at the end that's one of the main reasons why you would want to have a rollup of your own to be honest yeah that that that point you just brought up really Rings home to me because I've always thought of Da lers as kind of like like the value is in the ecosystem and the ecosystem is only as good as the Imp interoperability between all of the different chains utilizing the solution so I saw that Avil is working with uh succinct labs to build a two-way biral bridge between ethereum and Avail and Celestia seems to be trying to do something similar and they're also leaning into IBC a little bit so I guess the question here is how is that going and why maybe did you choose ethereum I assume because that's

where you know most the assets and liquidity are as like kind of your starting point but yeah if you could just expand on that that'd be great yeah absolutely I think the choice of ethereum was I mean a no-brainer right I mean if you want to be able to use a settlement layer you will want to use ethereum like full stop right that there cannot be any other debate about a more potent and a more valuable settlement layer that you can have for any rollup today and rollups are rollups would want to settle on ethereum because of the network effects that it has right because of the amount of liquidity the amount of

faction users everything that it has it's a noer for a RAB developer to be to to settle to ethereum and hence uh we wanted to you know have a bod directional bridge to ethereum as the beginning when we were developing we we thought that what can be the best way to build a bridge and mind you that bridges are almost always the the the weakest link of these ecosystems right we have seen that time and again and hence uh we wanted to have the strongest amount of guarantees that it can be be it can be having across two different blockchains and that's why we thought of having a

light Cline based verification on one chaining another and if you want to do it natively it will cost you a lot for example if you wanted to bridge ail to ethereum and ail has ed25519 signatures of thousand validators the verification itself cannot fit into a single block of ethereum right so it was almost impossible to do it that way and hence the the next best thing that you can have is a ZK proof that the validators did sign it correctly and that's what we are trying to do with Vector X which is the bridge that we are developing from ail to ethereum along with sassin glabs and sassin Labs

already had telepathy which was being used to bridge ethereum to other chains and hence we use telepathy for the direction of ethereum to aail so that's why we wanted to have this B directional Bridge with ethereum there will be other bridges of course as you know that it's not going to stop with one one connection right we don't want to have you know 100 different connections with 100 different chains that's not the business that we are in but at the same time because if you now have a plugin to ethereum you can almost do a hop to any other ecosystem right so if we can abstract it well enough we'll be able to

you know take information from any other chain in any other ecosystem to a in and vice versa and so that was the starting point from where we started what's up everyone March is approaching fast and I want to give you another reminder not to miss out on Das London it is coming it's right around the corner and it's in March from the 18th to the 20th we have three full days of content this is your chance to bump shoulders with some of the world's top Executives and have open dialogue with both attendees and speakers we're going to be focusing on a range of topics that I'll let Ren discuss for you first on the list we have Bitcoin Catalyst the Haring and spot ETF next we have a view from the buy side from investors on things like strategy portfolio allocation and more we also have a topic on Aras

tokenization and stable coins which I think we can all agree are going to pay a large roow in crypto's future we'll also talk about global regulatory Frameworks like compliance best practices and the evolution of global standards that are shaping the global investment landscape we also have someone from the institutional front to talk about infrastructure such as Banking and payments with financial giants like visa and JP Morgan and last on the list the macro case for digital assets so don't miss out on this Monumental event seats are limited so be sure to register today by hitting the link in the description and using the promo code 0x10 to save 10% on tickets see you in London I wanted to ask a question about the block building process for da layers I don't think a lot is understood in terms like how it actually works so I guess two questions

there the first is like what does a block building process look like for a d layer or specifically for a V and the second question is you think the block building process will ultimately become centralized in some sense because you know on ethereum the the block building is like relatively centralized between five Builders or are those Dynamics just not at play for a DA layer I would answer the second question first so it is not a choice for the DNA layer to become centralized in blog building because if you think about it all the me of cross rollup interactions will remain with the blog Builder all the sensorship

resistance that you can give for any of the rollups on top will rely on the blog Builders so if your blog Builder is actually centralized then you are you are not giving any value prop that D should give so ders should not only give you know guarantees of da that's of course there but it should also give guarantees on ordering and for giving guarantees on ordering the force inclusion the cage resistance those kind of things become very very important that's one part and then let's come to the to the fact that why I think

that it will not be centralized right so what is the blog building process so today the block building process is actually more complicated than in needs to be that is because we were building on top of you know substrate or po SDK as it known as today and in in that you cannot we have changed it quite a lot to make sure our block building can be implemented but then you cannot rip it apart right so we still have to play by the rules of substrate which is that of re-execution so what it happens today is a Blog Builder comes in it takes in the data blobs that were submitted from the

transaction pool it creates a commitment on top of the data blobs and it produces a header it propagates that header and the block itself other validators who who have to sign on top they have to take the same data blobs re recompute the commitment and then check equality now now this is suboptimal because in an ideal world you would not need to recompute commitments but you should be able to sample them as well because once you committed to a data even the other validator should be able to sample it but we were not able to

implement it at this particular phase but that's part of our future road map those are the kind of improvements that we want to do so that the block production engine becomes extremely thin so that anyone can become a block producer because today kg polom commitment generation is extremely cheap and extremely efficient it can it will take maybe you know 900 millisecond to generate around 4 MB worth of commitments on in special orientations that is so uh that's why we have to that's why I think that the block production will not get centralized and it doesn't it has to be ensured that they are not centralized

because that's one of the value propositions that any D will give okay that makes sense and we had Brandon Farmer on from polygon the other day actually and uh I found his thesis of you need to be able to scale liquidity in a singular environment like as kind of compelling because there is like a lot of liquidity fragmentation amongst rollups so obviously you mentioned or alluded to Dimensions launch earlier in the conversation and they're aiming to have kind of like a settlement layer with an shrin amm to like kind of unify liquidity how do you think this Dynamic looks in a world of a thousand rollups that you clearly have a lot of conviction on like where is it fragmented liquidity on every Roll Up is there going to be a settlement layer

where most of the liquidity lives and maybe that's ethereum and that's why you're building the bridge but just more thoughts on your thesis around how that plays out would be great so one of the main shifting thought process I will I will answer to that specific query of yours but before even going there I I want to clearly state that liquidity might be a flaw metric in this rup Centric world the fact is that I feel there will be more application specific rollups I'm not here to say that all these thousand rollups will all be AVM all be general purpose all have Smart contract support that's not how it's going to scale because by power law

there will only be a handful which will remain which will have all the tractions and all the others will die out so the all the in the application specific rollup world is where we see thousands of rollups and in that world liquidity is not that important for example if you think of web to world and of course we we want to create a better than web to uh world and that's why we are here but there are definite you know learnings that we can get from the web to world if you think of how web2 is able to scale all this is what they do is they have made microservices right so they have made

specialized thin services and middle Wares which take a lot of like API calls from one another in order to make a seamless experience on top so when you buy a product on like say Amazon it gets gives a call like when you go a payment it maybe gives a call call to stripe stripe doesn't have all the information of your credit card right so it gives a call to let's say visa and Visa doesn't know your bank balance right so it gives a it gives an API call to your bank and at only at that point the bank has your account details or may or may not have your account details and that's where it

either optimistically or checks it and pessimistically deducts the balance and then returns that promise or resolves that promise to Visa Visa resolves that promise to stripe stripe resolves that promise to Amazon and Amazon is like yes I trust stripe I don't need to know what your bank is I don't need to know what your payment aggregator is I just uh rely on stripe and that API call and I will go ahead and give you the the thing that you wanted in the web three analogy we want to create a world where all of this happens but in a trust minimized man so you don't need to rely your trust on stripe you don't need to rely your trust on on Visa or similar analogous services that will be trust minimized

but individually they will be specialized because if you want if you think of a general purpose chain its state is generic its transaction structure is generic its scalability has limits it's it has a lot of State bloat it has a lot of you know structural um you know nuances which were taken because it has to be generic whereas if you think of the app specific world you can have very very custom execution environments and those are the places where we will see a lot of new novelty of constructions coming up so I mean again web2 has seen similar Trends already when a new if you want to launch

your new website you will not create a new framework of your own right you will use something off the shelf you will create a you will create a website and as you get more and more users you would want to you would kind of think that okay this is suboptimal what can I do to make it custom built for my application and that is how all the major Frameworks today have been built all the major Protocols of the web to have been built right some person realize that this might be ideal for my use case then open sourced it other people built on top of it and that's how it become a standard right not sure if that answers the fully but I can come to brendon's comment in a

bit it does no and I I really like the trust minimize micro Services analogy right um in the web two world it's pretty fragmented there's probably a decent amount of like value extraction and definitely not trust miniz because you're depending on someone to maintain that API Ure it has chronical liveness and also has chronical like the correct execution of the API so I do think that Vision you have for like a world of like a thousand rups makes a lot of sense I think the next question I wanted to ask is braiding more from the BD perspective right I think Celestia has managed to successfully craft its own narrative and one could say that it has executed a

pretty good Golden Market strategy what are ail's plans for Golden Market strategy and capturing da layer mind share especially noting that I think we're still in a relatively early days of ders taking off yeah know that's a that's a very good question like first of all I want to say like all credit to Celestia because we have to explain the problem of Da a lot less than we had to do back in 2020 even within the or and so on right so there was a lot of you know lot of mental boundaries that cesa guys pushed and all credit to them for able to create a niche for themselves and to be able to you know help us making sure that we

don't have to explain it in too depth the second part of that is of course they have done a brilliant job in the narrative and such but at the same time the market is still very nent so if you think of one production grid ruup which has the guarantees of the base layer using Celestia it will be hard to find again I don't know exact maybe that is there maybe one or two are there I don't know of them but all the big ones which I hear of either do not use the bridge or they do not have short

proofs or they do not have the da garantees all set inside the protocols the sent doesn't even know or they are being subsidized so again I don't exactly know whether we have seen even the Genesis of this entire explosion that we are going to see all the production grade rollups which inherit the security of the Bas lay are right now ethereum so that's why we will there is a lot of scope for for us to come in and show what we are capable of again on the other hand if we compare the Mind share that they have been able to take versus us it's again Apple Stores right because they are the only production da

layer that is out there of course we will soon change that but credit variants due they are the only ones and that's why they are the only option if you want to build a production rollup right so as we go in production we will have our launch Partners we will see all the all the announcements that we do in the coming days so so that it becomes extremely clear that there are many more uh rollups that are going to get built there are many more production grade systems that are going to use a and a is not just a d that's just the start of our journe a little bit more of a thematic

question here for you what excites you the most uh in terms of like what could be built on an old da solution like a veil um could it be like a web 3 native app maybe like some decks or a D5 primitive that we haven't seen or are you like more excited about a real company in in the world maybe integrating blockchain rails through something along the lines of aail I know that's a little bit of a left fielder but curious to get your take on where the real adoption occurs I think again I I want to be here neutral and say like a both because who are we to pick and choose as I said we want to be neutral and that's not just by you know by not

leaning into one or the other but because we don't have a choice when someone submits it's a data blob we see it as a as as a binary blob of data we don't have a choice to know whether it's coming from Dy whether it's coming from institutional whether it's coming from a competitor I don't know right so that's why there is no choice for us other than to be very neutral that's to be honest but where I see actually the the new things happening is when institutional ones come up and use blockchain rails but in a new way so all the existing like it's it's it doesn't matter what

exactly gets built what matters is what are we doing to make sure that the cost of experimentation is low because if we do not make the cost of experimentation low enough then new things will not get built new developers will not come into the space and if new things and new developers don't come then we will keep on you know grinding the same Cycles over and over again the same uh applications in new rappers with newer incentives incentives will dry down users will clock out and those are the things that we will keep getting stuck at so we definitely want newer use cases to come in and for newer use cases

I don't know the answer first because I don't care that's true but the second is also because we are hent on making Ed that the cost of explation is low enough so that new use cases can be built whether that comes from institutional whether that comes from you know existing designs I don't know I wanted to follow up there I have a super random question that might not make any sense but I'm just going to shoot it out for example um you mentioned how institutions will build maybe some like Innovative solution right but say tornado cach right where the Builder rollup and this tornado cach rollup decides to Post Its da to Avail

celes whatever da layer is there a way and assuming like the SEC says okay this da layer is now like complicit in like uh supporting like olda transactions is there a way for a d to like censor a rollup from posting data to like it's D or it's not possible at all there are two ways to think of it the two layer at which this interaction will happen the first layer is how is the transactions getting batched if there is a sequencer one or more sequencers they are the ones who decide which transactions come in and which transactions don't so they will be

first of all hopefully liable for for transaction inclusion right so that's where the filtering should happen but at the same time time as I was mentioning before there can be you know Force transaction inclusions and such and those are the guarantees you might want to give in such crucial applications and in this regard the only way to you know prohibit someone from doing a first transaction is if validators choose to censor particular origin of the transactions so when you submit a transaction you submit it from an account address because someone needs to pay the fees and when that happens the validator

might choose to not include specific accounts because they are known to be whatever non-compliance that that they will argue upon right so that individual validators can do and that kind of thing already happens in the most sensor instent uh you know chain that you see today Bitcoin we have seen what governments can do when they CB you know mining activities on ethereum we we see that a large chunk of blocks are ofac compliant and you cannot avoid that I mean you cannot just make every validator comply to non ofac right because at the end of the day validators own the network they own the slot uh for

which they are being Chosen and in that they can do whatever they want so I've got a question on light clients bringing it back to the more technical side of things do you guys actually know like a quantifiable exact number of light clients that are sampling actively because that's directly related to the I guess the block size and how large you can scale I would assume based on what you described so I feel like being able to quantify that would be important so right now what we do is we do have a basic Telemetry kind of setup in which we have seen up to 5,000 light clients

operating um this is to just to say that we have also seen like 2.5 like 2500 full nodes also operating we have been operating you know 310 validators for our inent device test net and so on so the numbers are are of course large uh even without the light clients in the full node territory in the validated territory there has been huge amount of Interest so we have been you know lucky enough to be able to see those numbers but for now we have seen around 5,000 we are launching our light client incentivized challenge pretty soon maybe in the next one or two weeks maybe by

the time you guys publish so uh in that we will actually ask people to spin up light clients and during that we will do the stress testing that is when we will be able to know that you know the total am amount of light clients that we can support our peer-to-peer Network can support you will be doing a lot of stress testing there internally and with the you know peers that we have and the partners that we have we have seen around 5,000 today I think like clients because one of the questions that we've sort of like struggled with especially on the Celestia side of things is like no one has a straightforward answer when we ask them if like how do you know how many

like clients your network has and because the scale ability of the network is I kind of tied to that so I appreciate the clarity there I think the last question I wanted to ask is just what's your data availability Heart Take You Know What's like the spiciest opinion that you have that the general crypto me disagree with yeah I think uh one of the things which we tend to do and by we I also want to tell about myself is that the D that somehow we try to position ourselves as both a savior for high cost inv validium

constructions and having high guarantees of Da both punched together the fact is that in validum constructions the cryptographic guarantee that comes with data sampling goes away unless there are specific constructs in which you operating so in a normal sense there is no guarantee which data sampling gives in a validium unless your users are running the DS Services we are trying to do that but there is no no chain today which does that and that's why valm operate on cryptoeconomic guarantees and it's important to be able to call a SP a SPID in some sense

that the in that regard it just act as just another blockchain which keeps the the the data right behind the transactions and those are sometimes you know misrepresented the other thing which we tend to do is whenever we cannot guarantee something by the protocol we choose to fall to social consensus we choose to say that you know what yeah we cannot ensure this but the social layer will ensure this there will be enough people woke enough to realize that this is a problem with the network or the protocol just completely failed and that's where they will all you know roll

up their sleeves and start coding into their uh you know into their clients and go through go to a minority Fork because they want to do a social fork and social slashing and so on in theory yes but in practice it has been seen only a few number of times I mean there's only a handful amount of times where there is enough at stake when the social slashing has happened and then that's why we need to make sure that most of our guarantees come from the protocol itself and all the extra protocol guarantees can be left to the social layer but we should not rely upon it too much not sure if these were hot enough

for your liking no that's good I saw that uh thread that you had actually on Twitter between a couple different people um so yeah I I thought that's what you were going to say to that answer U but one last question I got to ask is this is a crypto podcast um you mentioned incentivized uh running of light clients on testnet here in the coming weeks so is there going to be a token and if so what role does that token play in the Avil ecosystem there will be a token we will come up with the details maybe later because we are still figuring those things out um the role of the token will be very simple right so it any network which is completely independent L1 it

needs a token to incentivize the validators and to be able to you know do DOS product prevention and such right so people need to pay sound piece and the validators need to be rewarded using some toen so that's majorly the role that it will play in the beginning of course as we you know talk about the different things that we will come up after the DFAS we will talk about the extended role of the token that it will have in the ecosystem so as to mainly Drive the ecosystem towards creating more value that's the you know hypothetical answer to be honest at this point because uh in reality today you

can use it or we use it on the test net to uh fake incentivize the validators and you know drop through faucet so as to do a little bit of spam prevention on the Block site but yeah we are not very successful at the Dos prevention maybe because we are seeing millions of transactions per week which is very good for us because we are going to a stress testing and that's very good awesome well we will have to bring you back on in uh six to 12 months and get some more details on on how things have been going but thank you so much for coming on is there any last uh second thoughts you want to share with the listeners maybe where to find you

learn more about a veil yeah so just get and get and see apr.org find us on Twitter uh follow us there I am probal allergy on Twitter I I have put my handle here um we have a clash of nodes inant device testet program where anyone can participate be a casual user who just want to try out the chain be a security researcher who want to find a vulnerability we have B bounties there and elsewhere we we right now are going through our partner challenges so if you're are a rollup developer you can today also run Dimension chains and Madara chains on on

us there is a zkm uh for change that is running on on us light Clan challenges are coming so yeah stay tuned and there will be many other challenges which we you know come up sh awesome thank you so much we'll catch you later hey everyone thanks for watching today's zero X research episode I wanted to take a second and remind you about an upcoming 2024 digital assets Summit in London this March seats are limited so hit the link in the description and use the promo code 0x10 to save 10% on tickets see you in [Music] London