Modular March EP 5: Value Accrual in The Modular Thesis with Tarun Chitra

Transcript

welcome to modular March I'm Andy this is Robbie and we are the rollup modular March is an educational experience about all things Moder [Music] enjoy [Music] welcome back to modular March I'm happy

to be speaking today with tun from Gauntlet tun welcome would love to just give a quick brief introduction about yourself um and then we'll Dive Right into some of your thoughts with regards to value approval in the modular Tech hey yeah um thanks for having me um so yam tun uh founder of Gauntlet and uh you know we've we do sort of risk modeling risk management throughout Defi and uh kind of the L2 ecosystem and we uh you know have been spending a lot of time thinking through both value Cru as well as risk properties within everything from reaking to uh data availability I've also written a lot of um sort of academic papers on things

kind of related to this so that's sort of the tldr yeah I mean I noticed when I asked kind of what you'd be interested in it was value cruel in the modular stack and I think there's like various thesises going you know being passed around right now um so maybe we try to set a foundation for kind of what your general perception is or or kind of overview for how value will acre in the stack yeah so I mean I think uh you know if if we look at Computing in general um there's kind of been this historical divide almost like two very separate camps that are is it sort of

reflects the current sort of modular versus monolithic uh kind of dichotomy which is you know do I optimize for really low latency but not high bandwidth or do I optimize for really high bandwidth but things are asynchronous like they're not happening at the same time I can't give you any guarantees about how time passes between the two objects I can only tell you some relative properties of them um and I think you know historically if we look at the internet the earliest parts of the internet and the earliest parts of uh you know compute and and where it intersected with Finance were really focused on this kind of low latency stuff right every everything was really

about like how do I uh make these Financial systems um very efficient how do I you know by reducing latency I can make price Discovery more efficient Etc and um you know there's a sense in which the modular thesis for normal Computing kind of only became interesting once we had applications that were kind of not working as well as they should right like when we had sort of the Twitter fail whale and like Facebook was going down all the time in like 2000 to 2010 and so there's kind of this interesting thing where like in the 990s a lot of the investment a lot of the interest was in the kind of uh low latency single

gigantic machine type of uh thing and then it kind of eventually moved to this more distributed thing now of course there's still tons of places where the lowest latency uh performance things are really important so lots of places in trading so like when I worked in algorithmic trading like that that's a place where you care so much more about latency than bandwidth but on the other hand for something like a Facebook a Google uh open AI you care a lot more about bandwidth because you just need to have enough throughput to handle a really high variety of different demand sources right like the person who's going into chat GPT getting help with co-pilot for coding is very

different than the person who's going to chat GPT to be like write my emails uh and training these models for those different use cases is a very sort of heterogeneous workload so why did I kind of give this sort of long backstory it's it's it's there's sort of the sense in which the highest value in these kind of modular asynchronous systems is derived when you have very heterogenous workflows like workflows with different demand characteristics like you know some types of things are like very regular transactions but maybe small value or small quantities some types of trans some types of things are you know there's you know basically no activity but then a ton of spike in

activity and then no activity again you know that's like an nft auction whereas the first thing is more like a defi protocol and being able to handle these like heterogenous demands uh safely and efficiently oftentimes works better in the modular World versus something that's homogeneous demand like there's a particular application that that specific application you're trying to tailor you know low Laten and performance for and so I think the modular stack within crypto still hasn't figured out if we're being completely honest value a c story because there's no applications that are taking that have these kind of heterogenous demand

behaviors like you know Facebook does or Instagram does right or like where like if I think about Instagram right like if I imagine I could make a map of all the different Instagram accounts and I could just show you like a heat map of like and imagine they're some some somehow locality sensitive so like people who are connected or clustered nearby just like a 2d map and imagine I could give you sort of like a heat map of hey which which uh accounts are having a lot of activity which ones are having low activity you'll see that like the demand distribution changes throughout the day the kind of like which which types of accounts are posting and generating a lot of uh you know load on the the

network at at Facebook will will depend on the day depend on events right like right around the election they're going to you'll see the whole map red but you know maybe on New Year maybe like right after New Year's Eve you actually see this huge demand depression because like you know people were posting a lot right before and then it kind of and this is what I mean by these kind of like heterogenous demand uh distributions now crypto apps right now are just very primitive because there there a lot of like finance and trading apps or nfts or or very minimal games and because of that we haven't had any apps that really have benefited intrinsically from the fact that you could have heterogenous demand

distribution so I think one really important thing to people building applications in the modul is to figure out if your application has this kind of property where where where sometimes you just have very different usage of the system and you want to Outsource different compon components of you know the high usage and low usage times to different uh modules effectively right that's when you get the best benefit and I think application developers World developers who want to think about when it's you know optimal and when you know value acral occurs it's in that part of the world and the reason for that is when you have these heterogeneous demand distributions you can specialize

components to handle different types of workloads different types of compute loads different types of uh transaction loads and you know when you have a a large amount of usage that you know you you kind of get this like economy of scale from that uh and and I just think we're just in this n stage where like there hasn't been like the Unis swap moment for modular stuff where like it's an application that takes advantage of this ability to specialize um and you know I think that's the thing that you know we're all on the looking out for so sorry if that was like a bit of a rant but yeah and I was on with uh brandan farmer of polygon and we kind of were talking about this thesis earlier about how what's GNA kind of happen is a lot of these Stacks are

almost like commoditized very down to low fees not very focused on rent extraction and things and and kind of what's going to happen is there's going to be a big hit that is built on these Stacks which is like oh the reason why this hit is possible this game this application is possible is because of the stack which it's it's built on and that's kind of like this Unis swap moment of adoption there and so I think that these St may not you know have individual value creation it might be hard to tell which one like da or execution Etc will be you know the most profitable per se but I think the applications which are built upon the stacks kind of speak volumes for the stack itself yeah I mean I think there's just kind of this thing where you really need to

um uh understand what types of applications need specialization and I think right now we are still in this part of the market where people are trying to take things that have worked in crypto and say hey look if you did it in the modular world it's better but that's not always true right like things that rely on liquidity really rely on you know the liquidity kind of being aggregated and obviously one of the kind of costs you pay for asynchronous communication is aggregation is more expensive than it is in the the monolithic world and so I think there's sort of this question of

like how do we get developers to start thinking about applications that take advantage of these things rather than like strictly trying to Port existing um applications that exist and I I think that's where I think things like gaming and like native onchain gaming are like more close to taking advantage of of the modular stack um than you know def for instance and what what specific parts of onchain gaming or other applications do you think are like more expressive which are are better suited for the modular stack well I think the way to think about um you know if you're if you've ever

written a game in like Unity or unreal is like the structure in which you write your code the way it gets executed is already inherently modular so a I have maybe say a and this is a very simplified version of what a game engine does but I think this gives you an idea of of why specialization helps so you know a I may have a set of modules for defining characters so that's like defining characters I can move the set of actions they can take the set of States they can take what sort of uh features they have like maybe they have a health feature maybe they have a hat feature maybe they have an we glasses feature then I have um states that are like what do the landscape uh

what items are allowed on the landscape what objects are allowed on the landscape then we have sort of this notion of um you know what actions can those two things do to interact and you could think of those as two separate modules in fact often times when you're playing a game they'll be running on separate threads or separate cores of like the characters are like assigned to this one Q so there's like One processor that's dealing with the the characters moving in the game and that needs to have high responsiveness uh but then there's one cue that's dealing with like the landscape and maybe that doesn't have to change as much or it outsources to the graphics processor and that very simple Dynamic is something where separating the components and having them run separately and independently is more

efficient than having them run as one full object because if you had to run them as one full object maybe there there's a ton of character updates and then all of a sudden a a landscap background update but the the background might take 100 times as long as a character so now the characters can't update their states like the user is just like hitting like move move move move move and instead it's like waiting for the like sun to turn into a moon and you know making up some kind of like example like that and so I think games inherently are built to this way because they have they already kind of have these components and taking advantage of those components uh in s different modules is a very natural thing to do the question

is you know within the crypto sack there need to be some like economic interactions between them so what what are the economic interactions between the you know the character chain and the landscape chain uh and you know that's the stuff that people have to figure out right like what the value has to come from like people want to pay to make their character change the landscape more or something you know like and and I think those are the types of mechanics that a lot of people working on onchain games uh you know are playing with right like everything from from Dark Forest onwards has has sort of and a lot of the autonomous World stuff is has plays with this idea that like these modules unlike a normal game now it costs something to

to for to to transfer between them and I think that's where we'll see a lot of things that are you know probably better for the modular architecture than say defi like like I think there will be stuff in defi but it's it's it's it it will look very different than what we see now got it yeah that's a it's a good example because it's it's something that I think is understandable for you know a lot of folk and and and generally breaking down these kind of individual components so I think like the key thing that we're getting to here is the value of will come from people and developers and Builders thinking about what Innovative Concepts can be built only

using these kind of specialized Solutions and then kind of how to kind of put them and go to market with them whereas now right we're still almost in this nent stage and I'm curious if you can draw any parallels to the to the current state of kind of modularity to any early emergence of tech crypto or non- crypto that you've kind of seen I know that that that we kind of touched on the uh on the Instagram and the AI stuff and I think there's probably some parallels there as well well I think the modular sack is like the equivalent of like moving from like on Prem servers to the cloud right because like once I'm willing to Outsource different components and not have to own the entire Hardware stack uh then it makes sense to separate the components that

I'm like I want to Outsource versus the stuff I want to do locally and inherently like the the entire data center industry and the entire Cloud industry was built around separating these components Outsourcing some comp some pieces of them and providing a so they could interact with each other and I think you know the the culmination of that was sort of this sort of microservice Boom of the 2010s but uh I think like people have a more negative view on what microservices are now but arguably a lot of the modu ACT looks kind of like microservices I think the difference is that in crypto you somehow need to have a way of there being real Financial interactions between these components um and like such that you

know maybe it's like hey by you know splitting my computation in two doing them in two different places and having different fee structures for the two things I actually make it cheaper for the end user right because like at the end of the day it's not that the cloud is more efficient than on Prem right I can just buy a much better server than the one I'm using Amazon they're also virtualizing so there's some overhead for the virtualization all all of that type of stuff but it's that I can do it for way cheaper and on demand and I think like those are the value props you need to get out of these systems for them to have a huge amount of usage versus kind of the the monolithic stuff where it's like all we care about is performance which to to be honest right

there are a lot of applications where that matters right so I think like there's a they both kind of are the yin and yang of this kind of world that points to thesis that Andy and I I think I think share which is that the value acrel ends up at the user and because it is more on demand and it's more convenient and it's more affordable for the end user to access this this like Cloud compute version of blockchain in the modular stack now they're able to benefit through cheaper costs and that's ultimately where value acrs because essentially all these pieces of infrastructure are commoditized fees are driven down through competition and then the end

user is offered the services at the best price because of the specialization at each part of the stack yeah is that a yeah yeah I I I do think the value acrel story is probably to the end user right like if we're if if we ignore token prices uh you know the value acrel of Celestia for Avo users who trade on ao's rollup it's it's not like Avo or Celestia is getting that much value out of the the migration in fact it's the Avo users who are getting the highest utility increase from like their cost going down right and so I kind of think modularity looks like the cloud in that way and that it's more about the costs

going down and also the cold start problem going down what do I mean by cold start problem um if I think about you know companies prior to having to get Cloud servers and I have I have a huge upfront fixed cost of building my own Data Center and then I have a huge upfront variable cost of maintaining the data center so you know I have to do a lot of calculations to amortise that cost to make it cheap versus in the case of the cloud I can aggregate many people's calculations and amortise that across the the the hardware spend I made I think modular applications inherently need to have the same kind of amortization uh type of property for for

it to to work and I think that that's why you you're likely going to see much more of the utility driven to the user not to the networks like I I I think like for a long time they're going to their tokens are basically going to compensate as a subsidy for the fact that like the networks are not going to generate much in fees until they've aggregated so much demand that it it's possible Right like a cloud provider probably you know in a lot of cloud economics and I think this is why we have ended up with this kind of oligopoly of of cloud providers there used to be a lot of small thepoke Cloud providers uh and and it's kind of they've kind of died is that a lot of cloud providers lose money

on a lot of API calls there's some API calls that they make a ton of money on but there's a lot that actually cost them money but they you they have to service them to retain the user and so it becomes this question of like how long can you subsidize that type of behavior and I think the tokens in these networks are effectively serving in the same way as Cloud credits uh in in the cloud so I I think there's a lot of similarity to the the cloud and data center World um and I think that's where we should draw the most historical lessons when thinking about this ecosystem might might kind of uh throw us for a whole whole different world here but what's the bearish case for modularity in crypto like why won't it

why won't it come to fruition why won't this Vision happen so historically Finance has done better in low latency and integrated environments um and part of that is because the ux of Finance is that like the longer I have a delay the more inaccurate a price or quantity or interest rate is right so I sort of can't tolerate asynchrony as much right versus like streaming I could totally tolerate asynchrony I I don't know if if if you remember but like there was a long time where you'd have to like you know you'd

stream a video and then there' be like hey it's buffering and it's like not loading and then it you know like that's the asynchrony right like you're waiting for the the the blocks of the video to to come to you they're coming out of order and like oh you maybe got too much from the end of the video and not enough where you are right now and so the buffering was just like waiting until you got those but you're fine with that right like it's not you didn't you didn't lose money you just you lost a little time but you're you're you you're much more sensitive to that and so I think the question is are there non are there Financial applications that benefit from from modularity and I think to me the most interesting place that will happen is probably

games um because like they inherently have this modularity and if you're adding an economics model to them that economics model has to interact with that somehow um between like the different characters between different items between the all of the different components um maybe it's not a very cons consensus view but I just think the the bearish case for modularity is that like everything looks like everything in finance we already have where it's like it's it's everything in finance we already have it's better because it's permissionless I can get whatever exposure I want that part is obviously just a better thing but that that's something that's not unique to the modular ecosystem versus montic I I I think I think it's it really needs to be

something that takes advantage of uh you know this idea that specialization at the cost of synchrony is you know buys you some utility gain for your average user are there uh are there compon are the components of Finance expressive enough to take advantage of the specialization like to to have one aspect of finance like a lending Market run on One Core another aspect like a pers you know deck you just and and have certain specializations of Finance run in parallel yeah so I think like one one way you kind of get that is you know I I

and I I think I think this is not to to to make da sound worse uh not like it's not doing anything but there's a sense in which da exists because of like inefficient resource pricing in ethereum right you kind of priced everything at the same time you're forcing everyone to compete for sort of a fixed resource uh and it's sort of like imagine imagine if you went to a grocery store and instead of just scanning the items you wanted you had to enter an auction where the auction was trying to sell you apples grapes and meat all in the same auction well yeah well sometimes there's going to be

a ton of people who want meat at the same time as you want apples and now your apples are priced at the price of their meat which is like not fair I guess to the the vegetarians uh and there's this resource pricing problem of like hey we have a single dimensional demand auction like like ethereum and the alternative Das are just basically saying hey actually if we separate out the this ephemeral data usage now there's a lot more to that right like you have to have sampling you have to have all of the Das all all the like client stuff for sure I'm just saying from an economic standpoint that doesn't matter right the economics are pricing that as an option of like if something went wrong I can

exercise this option to do something right so if I look at it from Pure economic lens you can kind of think of uh uh you know da is like just adding in a form of resource Dynamic resource pricing that like the main chain was unable to add to itself now obviously you know it's trying to EG like the the upgrade today um but this is one of these cases where specialization can be bolted on top from modularity so you can take something that's already working and like make it more efficient and I think that's another case where sort of bull case for modularity is that it can fix things that you know kind of work

uh later while they're live yeah it's an improvement and it's quite funny there was a panel in um Denver modic debate and the monolithic kind of Representatives generally agreed that they're going to scale in a modular fashion as they scale which was uh it was interesting to kind of hear I think last thing U before we wrap up here generally about like the current situation about block space I'm I'm a big fan of kind of expanding block space getting rid of this scarce block space problem but also like I feel like we're we could get to a point where too much block space and block space loses its inherent value as being block space on ethereum at some point whether they settled by an L3 and L2 app cfic chain Etc like what's your vision for how

block space expands and contracts in the development of kind of this modular expansion as we're calling it I think if this system is successful it will look like Cloud provisioning where you expand and contract in response to demand uh you're not you know like you could imagine ephemeral blockchains that exist only for a certain use case and then die and I kind of think one reason the modular world is better for that is that this idea of killing a blockchain is much easier to do in the modular world than it is in a monolithic world because the monolithic monolithic world

to kill it means you're killing the whole thing right like you can't kill some subsection very easily um and so I think this idea of you know one thing that I think is an interesting observation from from actually uh you know someone someone who I think is a sort of longtime crypto person was a Bitcoin Maxi until very recently uh for like a decade sort of uh pointed out to me is like he like you know when I first got into Bitcoin I I really valued this idea of immutability and the idea that you know this thing is has this form of object

permanence but in reality nowadays most of the things I value about blockchains is the verifiability component much more than the immutable component and the idea that users may have Dynamic preferences where that change over time for like maybe I care about immutability for this thing but then maybe later actually it turns out we don't need it but I really want to verify State because it's like it's money so I need to need to have the ability to verify things I think that inherently is is something that um you know is the world we're kind of moving into I think and I think the modular World allows users to choose the amount of immutability or the

amount of trust that they want now I would say Obviously the project that's the closest to allowing parameterization of that entire en High Spectrum is igen layer um I think you know I think people building stuff on top of Celestia could emulate a lot of the functionality but I think if the the the whole system for being able to sort of choose how much trust you want choose how much immutability you want choose a lifetime like em morality I think reaking is right now the only system that actually kind of control touches all three of those components and I think that's kind of the future we're moving to where like verifiability the core thing about blockchains and everything else uh you can pick and

choose which properties you want and and the mod the mon monolithic world makes that difficult this pick and choose component you have to take the whole basket or nothing yeah the big thing with I layer is of course not needing to secure your own economic basically validator set you can launch chains without even launching a coin right and you can kill chain you can kill the chains too yeah and and I think I think it is this is just a common thing I think not just in the infrastructure level but even in defi right you're seeing a lot of D5 protocols become much more modular uh and separate out their components whether it's unisoft V4 like liquidity

hooks which allow you to separate out the actual execution and liquidity from like some rules around how execution happens or whether you you know something like Moro or or or AA or Oiler some of the newer Lending markets separate out the different components of instead of like sort of pulling everything uh you you know like you still have pulled assets but you're you can construct some rules on your assets like oh I only want it to be used in this when a certain type of trans of lender I don't want to lend to dot dot dot right so I think this notion of the like I get to choose what subset of properties I want that is the next 10 years of blockchain and is is latency

not an issue so for Lending markets latency doesn't is is very minimally an issue it's only an issue for liquidations it's it's usually people are not borrowing and closing their loan it's not like a swap where it's like I might trade back and forth many times it's a slow interaction which is also why you have to be but it also is higher risk right like I'm keeping my Capital At Risk while while Ling so it sort of it is sort of different um yeah yeah well thank thank you tun for um coming on today and being a part of modular March feel like I could talk for uh another another three hours but very much appreciate some of your insights man definitely thanks cheers