Is Global Consensus Overrated? Shresth Agrawal Explains

Transcript

If I pay you money, if I do a normal transaction, it doesn't matter if you can process a gazillion transaction in that network. What matters is how fast you receive the money. And that is what we call as time to finality. What is the trade-off that you guys have made in terms of application level composability within pod network? This demonstrates multi-chain in its logical next step. 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 the rollup.co/ventures. Sit back, relax, and enjoy today's episode after a quick word from some of our awesome partners. Redstone modular oracles are available on 50 plus chains, trusted by

over 100 clients, and over 6 billion in total value secured. Visit redstone.inance to find out more. All right, East Denver. Uh we're here with Shreth of Pod Network. Welcome to our studio. Welcome to Denver. Thank you. I'm very excited. It's uh it's a good day to talk about blockchain design. Um and I think we were just speaking about prior to jumping on air about consensuslessness and I think it'd be helpful for you to give us a rundown of your perception of blockchain design leading to your idea around the double spend problem and kind of why consensus has been important uh and then leading into kind of h how you guys think about it now. Yeah, definitely. So you know consensus have been almost analogous to

blockchains since I mean our first you know bitcoin era and the whole proof of work we've been like uh almost thought of blockchains and consensus as the same thing and then with all the proof of stake and the idea with with blockchains or consensuses currently they take transactions as input and outputs them in this totally ordered log. It's almost like processing the traffic of the whole world in a single threaded machine. And with pod what we're doing is we're proposing a new design where the the the decentralized system doesn't have to be the single threaded machine. It can be completely parallel or asynchronous. And u it seems that you know the double spending problem uh which is what you need you know which is what kind of

Bitcoin was selling for right that okay you you want to build this decentralized peer-to-peer payment network and to solve that you need to solve a double spending problem and for that you need proof of work but it seems that you don't need the full power of consensus. It's a bit too powerful and that's why it's a bit slower. While what you can do is you can still build it with something we call as local ordering and having the base layer as completely uh consensusless where uh transactions don't get a a fixed index assigned in the in in the system. Yeah. So that's kind of a high level but maybe I can go deeper. So I like think about it for like for for the smooth brains here. Um, it's like when I go to Starbucks to buy a coffee, the entire world doesn't need

to, all of the world's nodes don't need to redundantly rerun my transaction and agree upon it. What what is more impactful is that those who are a part of the the this transaction in this network they need to come to a a agreement on this on the on the changing of balances and that that just becomes much cheaper and it it's like local consensusoriented less so um global consensusoriented but you're actually taking it a step further and saying like you don't even need the the idea of local ordering necessarily as much as you do so I don't know is that the right way to think about like on the left curve side or yes yeah so I mean like f first step to think is like you know if

I pay you some money and if I pay somebody like somebody else pay somebody else money like those are two independent transactions right they should not be blocking each other uh the only case where these ordering matter is when I'm trying to double spend and that is enforceable without needing a global consensus um but most of the traffic of of blockchains are independent right like transactions do not depend on on each other and this is not something that is enforced on the consensus at the moment and everything is paying the cost of blocking the whole world computer uh which should not be the case. I I I think this demonstrates multi-chain in its logical next step like this there's

a logical progression where you know we had one blockchain Bitcoin and then all of a sudden there was ETH and you know they were arriving you know in processing blocks and then ultimately we we got more and more ETH killers or alternative layer ones added more chains to the ecosystem and progressively people came around to the idea that this world was going to be multi-chain take that all the way to its logical conclusion And you have every user has its own rollup. And that sounds like totally mindboggling because people are like, you know, why why would I run my own rollup? But that's just the idea that I have my state. I compute transactions over my state and then whoever I'm interacting with or or processing, you know, commerce with, I agree to some

sort of change in state state transition between our state. Um and so you know initially the like word that I would use to describe that agreement would be consensus between you know like between me and Starbucks right if I'm buying a coffee you take money out of my you put it in theirs we have to come to an agreement on the balance change of those two accounts would you would you describe that as consensus between those you know between Starbucks and I or is there another way that we should think about the relationship from a transaction point of view of of that that purchase Yeah. So consensus in in a simple term for me is like when two people have to like agree on something and if you look at even your Starbucks example actually

the one receiving the money doesn't have to agree on you know if the person sending the money is actually sending it. It just has to be convinced that the person spent the money right. So there is like it's still not a consensus between the two parties. I'm sending the money and if there's like multiple parties sending the money to the same same Starbucks account, there's no consensus on like which payment came first, right? Because as long as the balance is the sum, it holds. But the idea you mentioned about these like microchains or having everybody being their own kind of uh rollup is it's kind or like a their own chain is is very very interesting and I think that's one way to think about it. It's like everybody maintaining their own local order and somehow Yeah. like it's like

these microchs and still being able to use the base layer uh to do composibility and transact across each other. Yeah. And I I think people do sort of have to unlearn a lot of the way that we traditionally think about blockchains. You know, coming in it's like okay I make transactions on a blockchain but I think it we have to have a mindset shift from a chain ccentric view to more of like a quantum mindset where it's no longer just in parallel. It's everything unpacked and and modular to its fullest extent. And you can even think like let's call each of these little modules pods. Yeah.

Right. And then pods are interacting with other pods and they're they're communicating state to one another. Um and they don't need some sort of global ordering mechanism. They just need to acknowledge that there's a change in state. Right? So when you do have a change in state, how do you broadcast that change to the the parties that need to be aware of that change? It doesn't need to be everyone, but it needs to be a certain set of parties in that network. Yeah. So you know, as in any kind of like, you know, blockchain network, right, we also have validators, right? And these validators maintain some local state. And the the trick is when I'm when I want to send a transfer, let's say I'm going to transfer money to

you, I can basically just send my transaction to these validators and they just locally acknowledge like if if they have not seen a double spend and that is already sufficient and they just send you a quorum that okay there's enough acknowledgement um and this is significantly different from blockchains because in blockchains or in consensus protocols these validators have to talk to each other and require quadratic communication which is super slow and that's the reason of the bottleneck. Um so yeah it with with pod uh you can really achieve this one network roundtrip latency or finality which is extremely fast and and snappy like we are used to with web two applications today. I understand that the like the technical breakthrough here you know to kind of uh succinctly put it is that the

removal of transaction ordering then equals a much more performant network where there is much like just lower latency cheaper there's better overall but there are also scenarios where I need to have transaction ordering for me or for uh you know application specific sequence sequencing as being a large topic in the technical community and you mentioned earlier this idea about uh communive property right where 2 plus 3 is five or 3 plus 2 is five but there's there's there's situations where that is true and there's also situations where if I have my transaction before after somebody it impacts the result of my of

my um interaction on chain or on on pod. How do you reconcile the different types of transactions that that do require ordering within pods and then those that don't? Um and like how does that be like solved in your in your design? Yeah. So, uh thanks for bringing this up. Um again like you so some things are commutative and some things are not and it seems that a lot of applications that are currently built on blockchains are not from are not thought from the commutive perspective. uh but the commutive designs are actually quite interesting.

So for for example you have different dex designs. So you have unis swap right unis swap v2 the simple amm and that is actually not commutative right as you said there's an muv between like if I order the transactions but then if you look at cow swap which uses batch auctions now because it's a coincidence of once and it's using this batch auctions orders within the batch is actually commutative and similarly you can do like limit order based x's where certain operations now become commutative. Um so the idea is if you can push more and more operations to this commutative design they will leverage this like very fast execution but then what we can do is you can still provide ordering for the things that need ordering but everybody else will

not have to pay the cost for it. So it's like a sync weight like you know when you when you design if you're a programmer you you're using like you know parallel computers and you know different processes you use a sync a weight where you say I want to wait for this thing to happen I want to pay for the extra cost of ordering and you can apply ordering or consensus posterior so by default everything is parallel if I don't need ordering I go through commutative operation and whenever I touch something that needs ordering that is conflicting then I apply consensus and then there is you know we have some ideas where applications can build their own ordering gadgets. So I can choose to use a centralized sequencer or some kind of uh full consensus protocol or just a multisc whatever I want and this can

allow these applications to capture the me as they want. Um so if if they if they do operations which are non-commutative and they want to capture me they can decide how they want to do it. Yeah. Now a lot of the validators on a network proof ofstake network like Ethereum today MEV makes up a rather significant portion of the revenue that those validators bring in in pod network all of the MEV is captured at a layer above the validators. So how do you incentivize validators to validate transactions? Yeah, I mean for us the base layer so currently we feel that you know blockchains are just tackling such a small part of the internet traffic and they it's like a zero sum game of just

capturing each other's me but once you unlock this pod layer where it's like so you can you can get so much more traffic about even like web two traffic that that's out there on web 3 there's going to be much more potential to capture revenue for validators from the gas itself. But then applications will also have a way to capture their users MUV if they want to. Okay, let let's just you know thought experiment like you know suppose there's some you know non-financial type of application like a traditional web two maybe social media or something or you know there's there's a broader theme that a lot of founders that have been in the space for a long time you know would like to see this idea of userowned internet. Now in that world like you know the internet we send

packets to a server they send it back and we get our information. How would a userowned internet work in something like pod network? We actually want the experience of the userowned internet to be as close as the existing internet, right? And that's one of the reasons why we're not getting this global adoption is using blockchains are basically super difficult, right? And it's super slow. And this is one of the key things we're targeting is can we get the UX of web two like systems but with the userowned internet with the decentralized uh primitives and the key thing there is you want the the underlying layer to be Byzantine fall resilient. It has to be

open. It has to be censorship resistance sensorship resistant but still maintaining this like very fast very snappy uh latency uh for finality. And so one one of the properties of that makes blockchain extremely different from traditional finance and is one of the biggest selling points of blockchain is composability. Yeah. And right now we're seeing the explosion of a lot of different chains because of different requirements for customizations uh desires for builders to be sovereign in in their execution noisy neighbors etc. when you're thinking about composability within pod uh in terms of the applications built on pod because

it's not necessarily a blockchain itself what is the composability improvement from the design or what is the trade-off that you guys have made in terms of application level composability within pod network and then how does it also extend to other chains outside of pod that aren't necessarily following the same design as you so the way I mean I just Think about like the the current scaling solution and with all the roll-ups now is like you have this single threaded machine which is Ethereum and then you're sharding the state completely into like parallel computers which do not have a way to talk to each other which are like rollups. Like rollups exactly and that is breaking the composibility

completely. Right. And then you have different solutions right you have either something like espresso where you again put it back to a single consensus but that will have the same bottleneck as a base layer. um even like the the situations with the native and the base rollups. My my feeling is it was going to have the same bottlenecks as the base layer if you put everything back to the base layer. Um or you have like these bridges but they use some kind of like again like a middle consensus protocol uh that you have this additional trust assumption. Um with pod what we're doing is as we're discussing right the base layer is completely asynchronous right and then you have these applications which are providing their ordering gadgets and those are kind of the

analogous for like rollups here right and because the base layer could already have like transfers and a lot of commutative operations even when you have these ordering gadgets on top these commutative operations across these ordering gadgets still can use the base layer and can be very fast. So if I have let's say a uni chain sequencer and let's say another like optack based sequencer for some other DAP a transfer across them will use the base layer and will be commutative which means it's going to be very fast. Um so that's how we think about kind of scaling blockchains or scaling web 3 without breaking composibility. Innovating finance global institutional and onchain deep liquidity to drive capital

efficiency. Now maximize yields on institutional grade liquid staking with M. Unlock trillion dollars of capital via Fitbitcoin. All of this secured by Mantle's treasury of billions in stables and ETH capital pools. This entire ecosystem is coordinated by a vast network of seasoned VCs that are aiming to create the next generation of banking and financial technology. This is mantle. So the so the app layer within pod itself can still be composable regardless of if there's a application that requires transaction ordering to

some degree and one that doesn't. Exactly. And then externally outside of pod itself, you can still use that community of layer to interact with other chains which require transaction ordering because of their nature and they can still compose with pod. So I was I was talking more about the first part where interaction but you could essentially interacting with something which is outside part. Yeah cuz isn't that where the problems would persist though? Like if I want to bridge from a chain that is a blockchain itself to pod and bring assets to pod then then then yeah like how how does that work? Because if you guys Yeah. So that is a that is an easier problem to solve actually.

Um I mean I'm not talking about the ordering gadget state but I'm talking about the commutative state on the pod if you want to bring it to ethereum for example and this is what we are already doing with a lot of applications. A lot of applications we're seeing actually want to use pod as a very fast layer and do things and then settle on Ethereum right and this could be like you know pi square as we were talking earlier proof markets or other kind of batch auction protocols uh limit order based access and the the basic primitive is part provides the certificates that transaction is finalized which is the quorum certificate you send a thing and you get the certificate back and the certificate is very easily verifiable on ethereum so I can convince Ethereum that the transfer happened and vice versa. I

would have to probably have like a light client of Ethereum running on part to convince the state of Ethereum on part. Is it um almost like a you know you guys are providing this like assurance that transactions are finalized almost like a pre-confirmation and and then settling that to Ethereum and and so is it fair to think about pod network almost as like you know a network of pods or a network of rollups and then ultimately these roll-ups do settle on an L1. So the one of the fun fact is we we came up with pod while we were working with flashbots to kind of decentralize the swave layer to do this um PBS auction because there is no existing L1 L2 that can uh actually uh decentralize this auction protocol. So it is it was kind

of inspired from this preconfirmation um idea but then like the the key thing is you don't need to put it on a blockchain because if you put it on a blockchain you will lose all the benefits of like this low latency and low cost right if you at the end order everything you will lose that. So you want it to be unordered you want it to be very fast and you only want to put things in a chain or in in order when you need to. Now so um how do you guys how do you guys prevent against spam on pod? So we we'll have gas and we have some very interesting idea of building like streaming gas where it's like currently with with blocks you know the the gas is very discretized with pod because there is no blocks gas is much more smooth uh

and there's like ways to kind of use gas as a load balancing tool as well across validators. So we have some of those ideas and it's it's going to be similar to how we have existing networks. So we we uh we have this race to like infinite TPS happening right now where like you have one I think like somebody's going to say they have 100k and then they have 200 then a million like it's just going to it's it's not really a metric anymore of value like what is your northstar metric for your performance of pod like what is the yeah I I think TPS is a wrong metric okay GPS or whatever gig gas whatever because let let me let's take a step back right if I pay you money if I do a normal transaction It doesn't matter if you can process a

gazillion transaction in that network. What matters is how fast do you receive the money and that is what we call as time to finality. Somehow TPS become became this like big metric because blockchains were extremely narrow and there was this like kind of blocking process and queuing and that's when TPS makes sense because TPS allows the queueing to kind of DQ faster but then once the network is asynchronous it really the time to finality is is the critical thing because at the end the user only cares about how fast the transaction get confirmed and we achieve time to finality which is responsive to the network. So if your network has a gigabit level 200 millisecond throughput then you will get your transaction finalized in that network roundrip like

whatever your network supports. So is is your is your only limitation on um transactions which do not require local ordering on pod the speed of light then like that's why we call it optimal latency and optimal throughput. Okay. So it's it is literally bottlenecked by the speed of light and we have some tricks to like add additional assumption on the client and use like these tricks that web two services have to even like bypass those limits and that you know it can get back to it and maybe the ocean that are like this big but process the entire world's data. Yeah, it's kind of crazy. Um what are the net new ideas that are unlocked because of this? like um you know I find it great and fun and

I know Robbie enjoys to go into the weeds but you know if this doesn't bring us anything new to the space then we're just kind of building more in for the sake of it. So I was talking for example in the start about uh building fast marketplaces. So marketplace is a is a primitive or these auctions right you cannot do it in any existing L1 or L2 because they don't provide the relevant sensitive resistance guarantees and auctions actually come up or marketplaces come up everywhere. So you have proof marketplaces right a lot of them are coming up and then we have order flows right like cow swap limit orders intent layers but then we also have uh blockbased marketplaces and essentially if you look at web 2 all of the revenue comes from these fast

auctions um of like advertisements all of this marketplaces we don't have a primitive to do it and pod is this one layer which can which can enable this um but then I think a lot of applications that we couldn't do on top of blockchains just because of the throughput and latency limits. Like for example, we have a Minecraft running on pod like it's it's just an internal thing, but maybe we'll make it public sometime. And you cannot run Minecraft on Ethereum. Like can you build crazy stuff on that with pod? You can do it.

So you can build like fully onchain games that have the UX. You can build decentralized socials that have the same UX as the as like Twitter or Instagram. So yeah, these are some of the things that come to my mind. No, I was just I'm not sure we're comp I'm not sure we're competing with Ethereum L1 anymore in terms of our competition, but uh more like Salana or Bass or whomever, right? Yeah. I mean, we want to even expand the space, right? Like are we we we want to take the traffic from Google and Facebook and in like we don't want to be the the zero sum game of like who takes that 1 million users that we have, but can we get to the next batch of of users in the web 3? At that point, you know, you guys are um you know, looking at

centralized companies that are also like serving, you know, the these user uh intents essentially or requests like the idea that you know, no longer do do we have to reexecute all these transactions uh you know across the entire the entire all the nodes in the network, but there's still you know centralized alternatives, right, that operate very very quickly. So like are you guys ever going to be able to get to the point where you are faster than a Google or if not like how do you how do you com like what bar what do you optimize? How do you compete with with those people?

Yeah. So I mean you you know a centralized thing a decentralized thing would cannot be faster for sure uh than a decentralized by so I mean so what's the advantage but you can get very close in terms of latency you're only paying additional on terms of throughput and with additional like for example TE assumptions or additional security assumptions you can get very very close the advantages though are a lot because Now you can build global applications which are intercontinental. You can coordinate systems which cannot could not coordinate before. Right? I could pay from here to somebody in Nigeria with like 200 millisecond finality. Uh I

could include payments in API calls and a lot of things that existing web 2 could not do. You could now do it. Come on man. So like enough of it. traditionally like like like fintech, right? Like a cash app or a Venmo or Revolute like that would not make sense like that would not make sense to do on pod. Yeah, that would not make sense to be on the existing web too. Like it would be on pod for sure like everything outside like you know doing like a direct peer-to-peer communication would basically can should you be using some kind of decentralized internet? I for one am gigab bullish.

Thank you for putting up with our pointed questions. Uh let's just lighten it up a bit as we close out here. Um c can I have two predictions for crypto in 2025 and any alpha on pod um test net mainet how our community can get involved? Yeah. Um the predictions on general web three. Yeah. Just two or three predictions about crypto in 2025. I'm the guy on the spot here. What are you doing there? We need we need some normie friendly content. I want to be optimistic and say we will see some more real use cases of that adds value to people. um we have not seen them so much and I I think we'll see it in 2025 and maybe pod will make that happen. Um and for alpha and pod we have our uh devet uh if you are a dev and you want to try

it out uh reach out to me on Twitter and we will be releasing some very cool application fun things to just try out build and hack on it uh in the next months. So yeah, I got a prediction. I think we're going to see at least one consumerf facing marketplace on pod. The decentralized Uber, right? Every like every Uber around here in Denver cost me at least 10 bucks. But then I see it pop up on the on the screen when I'm driving or when I'm like, you know what, and it's it's like five bucks. You can't make it up. You can't get an Uber for less than five bucks. You they charge you 10, but then they give the Uber five. Who's getting the other five? We need to make that marketplace. I mean, it's not the Uber driver. Give him the other five bucks. Thanks for your time

and I hope you have a good time in East Denver. Appreciate you coming on. Thank you. It was a pleasure talking to both of you.