Crypto's Future Powered by AI and ZK with Scott Dykstra

Transcript

okay so this is going to get real trippy then in our case in the case of ZK Co processing there isn't another option how do I make my AI agents interoperable and those agents will lie they're going rogue yeah great question I'm glad you brought that up U we're huge fans of ZK sync 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 roll up.co Ventures sit back relax enjoy today's episode after a quick word from some of our awesome Partners Redstone is def's modular Oracle supporting over 100 clients on over 50 chains offering 1,000 assets Redstone is the leader for Bitcoin

staking and yield bearing assets they offer the most robust customizable price feeds available in any model join the Redstone exhibition today [Music] what's going on guys welcome back to the rollup for today's episode we have Scott calling in from the shiny State of Florida along with Robbie we used to have a saying it's another beautiful day in Florida Robby kind of stopped saying that nowadays I don't know what happened to that saying it is it's another beautiful day in Florida it's finally it was cold there for a while so it didn't feel like Florida anymore Scott knows what I'm talking about it got a little too cold for Florida days but we're back

we're so back back it's a beautiful day back to uh perfect 75 and Sunny hell yeah yes SC thanks for thanks for joining us I uh been following space and time for a while primarily because of the interesting name space and time I feel like it's particularly relevant in today's world there's a lot of space and time things happening so we we'll talk about that we'll talk about ZK we'll talk about AI talk about some of the real world stuff that you guys are building out but but before we get into any of that um Scott what's uh what's going on man what's who's Scott and uh why do you care about crypto why are you building in this space and um you know what does ZK mean to you yeah great great to be here I've been following the

rollup for a long time so exciting to finally meet you guys on stream and be here um yeah so I I got into the space after a decade building Enterprise data warehouses like building these petabyte scale databases for Southwest Airlines and Walmart and the big Fortune 500 companies in America and like uh I was Co during Co you know trading crypto prior to co but during Co looking at smart contracts and was kind of shocked that smart contracts have no access to data like coming from a database background I was like oh this is you know back in 20120 we all thought like smart contracts were going to completely you know debank the the the bank and we we all thought this was the future of

permissionless Finance um and I was just shocked that the future of permissionless Finance had like no good data tools so that was kind of the beginning of space and time and exciting to be here Andy it's funny you mentioned the space and time name I'll sh I'll share like a really quick funny story about that one of our very first VC meetings with like a big name brand VC when we first started space and time in the early days on the call I'm very nervous it's our first big meeting we have like the senior you know very tenured venture capitalist probably in his 50s and then the young like 25-year-old old uh associate and at the end of the call it went great they ended up investing and the the the Elder

gentleman was like but you got to change the name just I'm not vibing with the name like you got to pick a more mature name or something and was very adamant and then the young guy was like oh no the name's sick the name stays that's kind of how I feel the name's sick and and and the name stays and and and I I mean I agree on the smart contract front um it feels like we're very much in the Elementary School area or era of of smart contracts where um like the these ZK code processors that we're building which I think we should also explain what that is Scott and explain how they work I also think that that that name is very bad it's a tough I think there's a very there's a much better way to explain it maybe you

can explain what ZK Co processer is and maybe if you have a preferred name because I think co-processor is kind of hard to understand yeah I'm actually chat GP teing right now what's a better name for ZK co-processor because I agree like I think it's a it's a name that scares normies for obvious reasons all it means is it's a service sitting next to the chain next to ethereum next to a rollup that does some compute some work some you know some kind of computational process it does some work on behalf of smart contracts on that chain and then proves to the smart contracts that it did that work that's it now the reason

we need to bring some computation off chain and do some work in a ZK code processor and prove it back to the chain it's because sometimes that work is way too big for what smart contracts can handle and a great example is processing chain data or like processing the history of ethereum if if I want to ask a question from a smart contract as basic as show me the total volume of transactions that Andy's wallet did in December a question that simple we got to bring some processing offchain run a bunch of computations and then prove it back to the chain because it's just too much data processing to do in a smart contract hence the name

co-processor it's great so what did chat GPT say yeah let's let's see did it does it have any ideas what is a better name for Norman and so like for for a audence um you know we've spoken to a couple teams who do co-processing brevis and lrange we had them on for our igen layer series if you guys tune into to that um and saw some examples and I think one big example guys uh for those listening just as more of a primer is is is um deck based competitions and and like contests and discounts and rebates and so being able to actually offer certain uh onchain addresses and participants rebates discs

things like things like credit card points and airplane miles and these types of things on chain truly are riant on these on this idea of a co-processor but did does does our AI friend have any have any tips or is are they all kind of mid they're mid as hell proof Eng secure math trip secure math chip I'm like BR these are worse okay um but that you mentioned the the the good example of like onchain loyalty yeah I think that's a great example so like if I wanted to say hey all LPS that have deposited at least uh $100,000 into my protocol um you know look up how long they've been LPS and reward them an airdrop based on how long

they've been an LP that processing that question answering that question of like looking back across chain history and seeing how long they were in LP that's just a little too complicated for what a smart contract can handle thus we bring it offchain to a ZK co-processor we do that processing there and we prove it back to the smart contract when we hand them the answer here's how long um you know Scott's wallet has stake this token or how long Scott has been an LP into this protocol cool and so you know a lot of that has to do with kind of like the the load that is put on something that is decentralized and so we've talking a lot about like the nature of ethereum and of of all these blockchains with

their their validator sets re-executing all of these transactions um and for a very large data set there processes a lot of data um to re-execute that and and have it you know one node do that and then broadcast that to the other nodes and then other nodes need to process that there's just so much so much redundancy that's a little bit unnecessary and so with ZK we can prove it once offchain bring it onchain and all those verifies just verify the correctness of that of that compute um and so we we've talked a lot about kind of like the transition from uh applications and chains that are doing a lot of this onchain and re-executing all of that to this this direction towards

ZK based applications um and so yeah maybe you could speak a little bit to like the the vision for uh the way you intend uh applications to use your Cod processor kind of in this new paradig yeah the first thing and the most important is that all ZK code processors and really all ZK systems in general are designed with like really really lightweight fast verification put all the onus all the work on the prer the ZK code processor itself have it do all the heavy lifting usually on a GPU but could be CPU and then verify super quick super fast almost instantly on the verifier side and that verifier side is intended to be

smart contracts now because of that I see this Paradigm Shift happening over the next two years like maybe five um where less and less computation is done at the Smart contract layer more computations and data processing and like actual like transaction uh defi calculations if you will all the calculations all the work done to to to build um like like to get you a loan on chain or get you an airdrop on chain that'll move off to ZK coprocessors and really the only thing we'll need blockchain for like we need bft consensus of like you said Robbie a set of validators coming to consensus

and redundantly executing work the only thing we'll need them for in in a few years is just verifying proofs like like ethereum's already become like slowly morphing into this player for like rollup proofs right that was my next question is the snark ification of ethereum and beam chain like we're already that's a concrete example of ethereum moving a lot of this computation off chain and then some ZK VM that is processing all these transactions and verifying all of these proofs on ethereum yeah and and the reason is like because the verification step is so light we don't mind doing it a 100 times redundantly or a thousand times or you know across all you know

active validators set right now but the um but so now can I can I can I ask you one more question like we have this offchain uh computation that's being done right and then we're going to go and we're generating a proof to the chain and then the chain one proof one computation done once on one node with one proof but you could verify it quick 100 across the entire validator set now we've I like also had a lot of conversations with ZK uh proof markets and like proof marketplaces and pro pro networks is another thing that comes up a lot how's a ZK Cod processor different right because you you do one proof one computation generate it once how is that

different from like aover Network or aover Market yeah so very interesting question and uh you know to be completely just hot take I'm a little bearish on proven markets and I'll explain why um so you've got one server doing the computation once and in the case of ZK provs whether it's a whether it's proving a rollup or proving some data processing I.E space and time lrange Etc uh both of the like both of those Pro you you just need one prover and you need liveness you need to keep that prover alive you need to guarantee that no matter no matter freaking what uh it returns this proof up to the parent or to the client that requested or no matter what happens when a asks

space and time to give them some Wallet information during a lending transaction we have to return that answer on chain there has to be this like guarantee that occurs and so like the way we solve that guarantee is just with proof markets where we have this sort of like decentralized almost a dpin network of nodes that are raising their hand and saying hey I'll run the proof I'll guarantee it's always on me and all my buddies all these node operators will all redundantly keep an eye on these client requests and will fulfill them and we'll guarantee that um there's always someone out there in the world ready to go ready to run this proof now if you had a decent like team that was good at running infrastructure they could also just spin up like 30 nodes

and yes it's all run by the same team and so you could argue that it's centralized but each node is cryptographically proving all of its work it's proving what it does so it doesn't really matter who runs the infrastructure whether it's a third party decentralized Pro cloud like you said Robbie or the team that that's actually like deploying these provs long answer just to Simply say like all we need is high availability we need liveness how we get there doesn't really matter because all the nodes are proving what they're doing whether you run the node I run the node doesn't really matter so I'm like like do we need these whole prod like we need this whole decentralized dpin networks that are like onboarding commodity Hardware to offer like decentralized proving I'm a little bearish interesting

okay so couple things first a general theme on our show here is that um we tend to believe in the like the tradeoff space of making applications faster more performant better ux and just at least a broad step in the direction of just increased throughput is typically worth it in terms of offloading a little bit of censorship resistance or adding a little bit more of a validator set or offloading a tiny bit of decentralization or for giving a little bit of synchronous composability like we're we're kind of of the mind that like you know for example ethereum L1 is not going to be verifying all the proofs in the future I mean that you know this

beam chain idea like Rob said or some of these other ones it may happen in through some sort of snar ification you may get to this point where there's at least some verification happening but you know there is verification layers um that that will typically like be used and same kind of thing with uh with ZK code processors where it's like again there is a little bit of trust involved and you just explain this guarantee process as to where this trust comes in and how you guys effectively maintain that trust through your work with um uh with the proof generators um however uh I think that one theme on our show and like that we can continue to kind of harp on is that like these trade-offs are are usually worth it and um as long as you're intellectually honest about

them and you don't try to act like um they aren't uh there isn't a trade-off then it's typically a net positive for the space so maybe you could in light of what you just said understanding the guarantee maybe that is the trade-off or maybe like if not what is the trade-off of using space and time versus the more secure option here well that's the that's the important thing to explain is like there is no other secure there in in our case in the case of ZK Cod processing there isn't another option it's bringing a new capability to the space that we didn't have prior you started this this this question Andy by saying like hey we think about the developer experience and

the user experience in crypto what we're not doing is we're not improving the user experience for something that already exists we're not making it better we're bringing a new capability in that didn't exist prior to ZK co-processors if I wanted to ask a question as basic as show me the total volume of transactions in Scott's wallet a question that simple couldn't be answered natively on chain like you can't answer that in a smart contract so what people were doing was like bringing data off chain and decentralized databases using a centralized Oracle to then query that data from their smart contract out of a centralized database or like using like the Dune API or using the graph which is not verified it's not ZK proven they're not cryptographically

proving their quer results or proving that the data is not not um yeah so so all we're doing we're allowing smart contracts to ask questions of onchain data cryptographic proofs which didn't which is a new capability yeah I just to just to break that down for listeners you know you could argue like well what's wrong with that why why why is it a problem to use Dune or use the graph or use chain link to you know to service some of this well you know chain link and and some of these companies may be too big to fail and they might not you know take the reputation risk of of doing something you know uh wonky but you know someone else that's going to go and broadcast uh data back to the chain or S serve up this uh this computation they bring it offchain into a

centralized thing they can do whatever they want over there right so if we're relying on this for like a liquidation price and I I you know cook the numbers over here a little bit and then I put that back to the chain and now now the chain is using that for my liquidation price all of a sudden you know I may have profited because I run a liquidation bot but you didn't actually need to get liquidated I did it for my own self-interest and so now with ZK coprocessors well we have a a ZK proof that verifies no th this computation was done correctly and completely and there was no funny business going on offchain so you're telling me that that space and time is the solution to Arthur hay is nuking his own

users yeah totally that's exactly what I'm saying but yeah Robbie I think you're you got it right like um bring these computations off chain when they're just either can't be done at all like there's really no way to do them on chain or they're just too massive to be done redundantly 100 times right like yeah like if I'm going to sort through like a lending example let's say a wants to give people a better borrow rate if they've already paid off loans then we got to go through the chain and we got to find all the times youve paid off a loan quickly and prove it back to a that you've already paid off a certain amount of loans while you take out your loan like during that transaction and you don't want to wait very long you're

certainly not going to wait 10 minutes and a lot of these proof systems are very slow so we're focused on two things like one make it crazy fast like return back queer results data back to the client that asked for it like next block as fast as possible subc if ideally and then the second thing we're focused on is just like ensuring that high availability ensuring that when a asks for the total volume of loans that Robbie's wallet has paid off we get that we we actually return that answer right like 100% guarantee now you know Andy asked about trade-offs which I I think is a good question because you know the the idea of ZK co-processors in general is a zero to one it it's unlocking new use cases that you know were not possible before um however there are you

know trade-offs and Design there is a design space for how the the ZK coprocessor is built you know we have lrange we've you know we're homies with with ishill over there and he's using igen layer to uh leverage cryptoeconomic security to to ensure liveness basically make sure that you know these operators are always online and and that is the proof availability uh so that lrange is always spinning up these proofs and giving it back to the people that request them um space and time chose the elastic chain and you guys have main net coming up very very soon so I'm curious why did you guys choose the elastic chain to ensure liveness and guarantee proof availability for your users yeah great question I'm glad you brought that up U we're huge fans of ZK sync and

really the whole ZK end game the ZK road map in general is exciting and I think they've done a really good job of putting together not just a very performant rollup but a a a good ecosystem of like tools for developers to connect a bunch of rollups together like they're doing a good job of connecting their whole rollup ecosystem across the elastic chain which we really like and you know our business is delivering ZK proven SQL query results basically delivering ZK proven indexed data to Smart contracts across these different l2s and so like having like the whole elastic chain ecosystem of all these different l2s that we can deliver data to is exciting now the reason we deployed our own rollup is because space and time is this network of nodes that

are essentially like database nodes that collect the data and secure it and come to consensus on it and then like load it into our approvers and so you can think of you could almost think of space and time as a permissionless L1 app chain specific to running the the data like you can think of it almost like R weave or filecoin in the sense that it's a very data specific application specific blockchain it's not a smart contract layer there's no smart contract execution on Space in Time all it does is runs queries gives you data back that's what our chain does so all of our staking all of our liveness and availability all of our approver slashing and uh you staking and slashing all of the smart contracts that like

onboard node operators and slash them when they do something wrong or like slash them when they don't deliver a proof on time that all needs to live in the evm so we chose ZK sync to be kind of Hosting all those smart contracts for staking for payments for node operator rewards for slashing Etc almost building like our own little um almost like a Min igen layer on ZK sync that sounds very modular um so another theme of the show it tends to be as you would know as being big Believers in purpose-built Chains very much um with the exception of a several high quality extremely high TPS performance general purpose chains

which I have to put that disclaimer in or else I get yelled at by teams that we've invested in but my thesis has changed over time um H we're generally way more excited about teams that have a specific focus and that have a a a customized build to adapt and adhere to that specific Focus it's a very it's very clear to us that you know let's say you have um you know if it's like you go to the buffet and you you or you have a set menu of what you have to eat that day at a dinner or you can pick and choose it's very clear to us that the buffet style with an an idea in mind will lead to a better meal than just being like here's your alicart menu build a chain with it um and so you

know I I I think that's generally something that that we believe in so would be curious to understand that a bit deeper on the technical side um when you mean when you say that there's no smart contracts in execution what comes to mind is like something like a Celestia an a veil like just like a data availability layer or like you said something like a um something like a file twin right so how exactly what exactly is the level of execution um and then uh why why is the or what what is the design aiming to produce like the outcome like what is technical design aiming to produce the best result for um and we can just start there I'd like to

get a bit deeper into into the tech of the actual chain yeah I love that you said per purpose buil or purpose specific chains because that's exactly what space and time is so a a bft consensus blockchain that only does two things collects data from other chains and seres ZK proven queries against that data meaning serves that data to clients that need it that's it like you can send space and time a SQL statement and it'll send you back data and you you pay for that in gas just like you pay for uh transactions on ethereum in gas the gas you pay for is not to write a transaction to The Ledger that is space and time but rather to query space and time and get back data same with chain link by the way like we we were kind of

incubated by chain link we were one of the first projects in chain link build and like very inspired by the chain link design of um this network of servers offchain that are listening for client requests on chain and returning back data on chain and getting paid per job for that data essentially in chain link gas um it's no surprise you got the Sergey beard and now now I need like a a light blue uh flannel um but yeah so short answer is like purpose specific blockchain for delivering data ZK proven as fast as possible subsec on on massive tables of index data you know so you can get back your query result next block and your

smart now you as a as a smart contract developer can finally ask complicated questions about onchain activity about your own users your own wallets the people using your defi protocol you can ask questions about them and get back answers next block but but because we're so purpose specific we don't have Smart contracts we still need to deploy all of our star contracts for staking and payments and slashing and staking on the evm on on the elastic chain so is that on zyns L2 zync era or like where is that being built yeah so we're we're doing a lot of work on our own rollup and then we're putting staking specifically on era because we're sketched out about like putting you know hundreds of million dollars of staked SXT on like a brand new rooll up that

we're not like you know um ready to scale that it's not optimized for that it's not meant for that yeah so so it's very much like a is is a good mental model think of this is like u space and time chain is very much a B2B chain it's very much meant for interactions with other blockchains not necessarily human users yes wow that's a great way to put it space and time be yeah okay so this is gonna get real trippy then how have you thought about AI agents interacting with space and time chain like uh is which like we can kind of open up that Pandora's Box unless Rob has any other questions by the technical design no I

hear Scott's doing some cool things with AI so I'm I'm very curious about this wait wait also Scott are you guys using um I I should have done a bit more deta are you guys using eth blobs for your da because you mentioned it's an L1 but it's also elastic chain so you're not to ail shouts out aail da there yeah got it awesome so it's not an L1 you're an L2 so so like space and time chain is just a purpose specific blockchain got it okay we let's not get let's not get sidet I'll answer the question it's a good question I'll answer it Bas in time is is a purpose specific chain that serves queries in addition we need an L2 to deploy all of our smart contracts for

staking and note operations that's on zkc but the main chain the main operation chain is this is this L is this is this purpose built chain specifically like this to B chain understood yeah l1's and l2s is also a thing in the past by the way we we leaving that in 2024 yeah I think the definitions will start blurring especially with Sovereign verus non Sovereign or like rollups that have their own bft consensus that's yeah exactly like celestian and a veil can literally build their own rollups without needing to settle to ethereum so like then what is that so it's it's it's all over the place um Okay cool so uh okay so we have this B2B AI CH or we have this B be uh zika coprocessor very

effective it's also an AI chain somehow like what is how has your mental model of the space broadly changed in the last three months with the introduction of AI agents and AI in general whether applicable to space and time or just in general as a developer yeah I I would I wouldn't say space and time is an AI chain rather we serve data to AI agents and like we indexed I don't know 100 terabytes of chain data from Bitcoin ethereum era polygon Basu you name it every major chain and and then like a bunch of offchain data from trafi markets as well so you've got this big database that can either be ZK proven or just like non ZK proven you just handing data to clients

that need it and so if you're building like a sophisticated AI agent that wants to transact on chain you need context about what's happening on chain maybe like you know you to provide your agent a real-time list of mcoin activity for it to decide on which mem coins to purchase on your behalf maybe you need to provide your AI agents uh information about yields across like 30 different defi yield indices to have your agent decide like where to move your money around to to to find you the best yields takes a lot of data and like will be a top quality data provider for the agentic future and sometimes rarely that data will need to be ZK proven when a lot of money is moving hands and uh we are we are building out

a uh an agent well shout out hyperbolic that is building it out on our behalf um we will have that soon I'm good there's been a lot of shout outs today in's podcast I kind of like this style just shout everybody out we're gonna start shouting out space and time after this hell yeah please do yep and so we'll have an agent very soon it's the name is roly poy for fans of the show may have seen this roly poy out there um in in the on the TL and so um you know initially this agent will be used for mostly offchain things but eventually potentially onchain things um and

so you know like like you said this agent is going to need access to a lot of data and potentially a lot of computational work that has to be done in order for uh the AI to understand what the state of the chain is and then to go and Inter and then it can go and interact with it but from from you know let's take away even the crypto like the financial side of that let's just say you know we have this Ai and you know we're training it on the content of all of our podcasts and it's become very well educated right on on you know this certain database of knowledge you know and and suppose you know there's another AI That's out there

that's also very purpose-built and it's very specialized on another data set of knowledge just say you know not crypto or or technology perhaps but you know what you know let's say it's very educated on coffee right how can these AI agents start to learn from one another maybe th those two things should be a little bit more related but you know without even getting into like you know making transactions on chain is there is there a a ra a reason that you know this agent that's very specialized might need to tap on another agent that's more specialized in a particular domain and space and time can serve a ZK proof that says hey here's this bit of Knowledge from this other agent you you

wanted to consume that you had to request it from another agent how do I how do I make my AI agents interoperable imagine if the world's GDP moved on chain guys say GM to safet the newest protocol by safe isn't a blockchain and it's not a new layer to it's a transaction processing Network built to send money anywhere and swap anything on thousands of layer twos from a single Universal balance in sub 500 millisecs it's fast it's smooth and it's simple and it's built with safe multisig level security in mind safet unlock smart accounts across multiple chains allowing users to supercharge security scalability and speed so get started

today at net. safe. Global and remember to follow at safe onx is z is ZK the right technology for AI agent interoperability interesting very interesting question I say no not at all like I would say that ZK is a way to secure data when there's Bad actors that are trying to manipulate like like if an AI agent um needs to prove to another AI agent that the data that it's providing it is not tampered not manipulated not falsified sure so I'll revise my answer hell yeah if agent a is proving to Agent

B that they're not lying to Agent B like a here let me give you an example agent a is like hey was there a hurricane in Florida today and it's asking the iot weather specialist agent that has access to terabytes of iot sensor data of wind speeds in Florida that agent says yes there a hurricane today wind speeds exceeded 100 miles per hour agent a would say Okay prove it because I've got $300 million of parametric Insurance payouts that I have to pay out on Shan I need to be damn sure that that there really was average wind speeds greater than 100 miles per hour but then you got to ask yourself is that really an agent use case or is that just like a data provider proving to a Smart contract

that manages parametric insurance that it pro like where do Agents come into this right maybe there's an agent adjusting the insurance maybe there's an a an automated adjuster okay now an insurance agent sure even better literally an insurance agent but um the reason I initially said no is because I think about ZK uh more just like providing data and I think about agents as needing to transact financially with each other phase one of crypto agents were chat Bots phase two of crypto agents crypto and AI was uh Twitter Bots like like you know virtuals and Ai 16z and axt AO and Eliza like automating uh Twitter Bots the next stage will be

agents that uh transact on chain and like manage their own wallet and manage their own private key that's going to be all the rage for a couple months starting soon of like a million teams building agents that manage their own that that financially transact on chain which I think is is going to be really important and that will be like the killer use case then finally the fourth and Final Phase after that will be agents that build their own protocols agents that build defi apps agents that actually generate working D apps that have a smart contract and a front end and a little bit of a database back end and that's where we're playing we've done some co-engineering with Microsoft using like some of Microsoft's Coen capabilities and like Microsoft's one of our biggest backers one of M M12 is one of our investors so we do a lot of like

AI co-engineering with them in their AI lab we've been working on a um very mature AI studio for for creators to like generate their own D apps that are very data driven like since we have all this data since we have this big database that's CK proven let's just help people build like data driven applications and let's like help users without writing any code generate like a smart contract that's data driven generate a front end that's data driven and basically launch their own D apps I think that'll be like the last phase and where ZK pays into that is just like securing that securing those apps like making sure that all the data processing is trustless yeah um selfishly for us in our audience could you tell us a bit more about what's going on with

Microsoft's in their AI lab my uncle works there um he told me that they just invested 80 billion into their AI research so what's been your experience working with Microsoft personally and with space and time and what you know what what's some of the things that you guys have been working on yeah it's been awesome they they've been a really good supporter not just on the business side and like introducing us to like major American Banks and like other Enterprises that want to use our data but they've also been helpful on the engineering side so first starting in like early 2023 we co-engineered with their ai ai lab like a a tech to SQL prompt to SQL system to like get really good at generating SQL queries and this was kind of before everyone had this

like we were like maybe six months ahead and then like everyone in their mother built like a text to SQL system um so that was a good like first start a good like first chance for us to work together and that evolved that that engagement that relationship evolved into building this AI studio for non-technical creators to actually generate entire apps we thought okay we can generate SQL queries we can generate little dashboards now let's generate entire crypto apps like think like um you know like a a little boyco you know something like that like a little mini like little mini apps that have a smart contract on the back end that uh handle very like specific onchain use cases which are data driven very cool and when when you originally approached

Microsoft's investment arm M12 you mentioned is that what it is I'm not sure the exact relationship there were they um open to crypto invest think were they open to the fact that you guys were building this technology for crypto use cases yeah they came to us and now I'll say like I I think you know that mindset has probably shifted a little bit like I I think they gone all in on AI and they're not doing crypto investing anymore we kind of like we're one of their only significant investments in the web 3 space but they came to us and and they were like looking to make an investment in the data in crypto space in 2022 and they were looking for like high quality teams that um were Enterprise grade data providers in the crypto space I think

like I think that was you know early 22 when the mindset was different and I think like as you guys know better than anyone it's a roller coaster of emotion in people's relationship with crypto and I think it's swinging back positive right now I think right now everyone's coming back around in a positive way yeah that's good yeah and 22 I mean real Builders you know come out in those times and so I good on good on Microsoft and their timing because you know that that's when if you're getting into crypto in 2022 like you got a real reason for doing the work that you're doing you know we we come across a lot of bull market Warriors that are quite a

bit op opportunistic um and they you know are are here for a few months to take take uh an opportunity and Leverage The the positive Financial conditions that are happening in the market but um you know when it comes to real pure technology plays uh th those are built those tend to be built by uh you know the survivors and the ones who are really keen on building something uh during th those swings yeah I I want to loop back to the the tech conversation because I do have one more on the ZK side um and you know you kind of laid out this agentic future and how you see that playing out and you also describe how these agents could leverage uh ZK proofs for verifiability and essentially prove that an agent is

not lying to them for security reasons um I want to I want to compound the the security of verifiable AI with some of the benefits of of uh ZK for interoperability because you know we see super chain building out their ZK arm and essenti I my understanding is that on the super chain the goal is that they use ZK you know you have chain a and chain B let's put names let's use B and ink right so you have B on the super chain you have ink on the super chain if you want to do some interoperability between bass and Inc bass is going to uh have a state and it's going to send a ZK proof

of its state to Inc and Inc can verify okay that's the state on base and I can interact I can interoperate and I can send a proof back and I know what the state of base is because I have a ZK proof that I can verify to the state of Base um and this illustrates the pro the interoperability properties of ZK but you're not you're not really looking at ZK for its interop capabilities when it comes to agents proving the state of itself back to one another you're more so looking at the verifiability of it maybe these start to go one one in the same H how are you kind of maybe you can kind of like fill in some of these gaps and and just you know round out the picture of how you see uh AI serving ZK

proofs between agents yeah I think you're on I think you're I think you you you articulated it well that um there's always a party a that needs to prove to party B that the sensitive and valuable data they're providing to party B wasn't tampered by party a like that's always going to be the case whether that's base proving something to arbitrum or that's my weather specialist a proving to your insurance agent that there was a hurricane like that's always going to be the case and I think that like when we get to more Sovereign more autonomous agents that are actually making their own decisions and managing their own capital and transacting on chain in a sovereign way without like me

uh micromanaging every little detail of what the AI agent does then that'll probably be more important like it'll probably be more important as time goes on that a sovereign agent can prove that the data that it's providing is accurate the key is it needs consensus among a set of nodes like whether it's Bas proving to arbit from the state of base or whether it's proving their weather in Florida you need a source of truth that's consensus driven or that's like you know chain driven I should say so like you could have a 100 iot weather sensors in Florida that are all redundantly collecting the wind speed and then a ZK proof that wraps up those 100 wind speed reports into one final

aggregate report provided to the client the insurance agent my point is whether it's agents or not ZK proofs need a foundation of like consensus among a set of parties that initially witnessed the data that can be a blockchain that could be a set of hurricane sensors point is in our case our whole business is providing ZK proven data and we need a set of nodes that redundantly witness the data initially like like redundantly data from Bitcoin and ethereum come to consensus and prep them to be ZK proven and so even in the case of AI in the agentic future you'll still need a set of nodes that can come to consensus that they all witnessed the

the data initially and signed it so that the agent can prove agent a can prove to Agent B that it's not freaking lying yeah those agents will lie they're going rogue they'll lie and they'll hinatea what you're saying is that we have all of these all of these parties that are attesting to a certain data set and then we can wrap all those attestations into a ZK proof and we can we can prove you know this the thing that they are attesting to I also I I love I I mean the use cases here like you know prediction markets settling prediction markets even using like data to to you know kind of prove the outcome of certain things um I

think that you know it's so far reaching uh or like Pro proving onchain activity like a proving that it did a certain set of actions on chain I proved that I deposited in Bara chain and then bridge to monad and then proved that I took a loan out on on monad that's a lot of chains you're using that aren't live yet sir this is in the agentic future we're talking about we're talking about the future right now talking about the Aged future right I must missed that part I think that's part of part of the value of ZK proofs is that you know all of these chains are going to have a ZK verifier like on day one so the ZK verifier is just a set of cryptographic curves that your chain needs to support today ethereum doesn't really support the curves we need so everyone's like

bending around ethereum like they're kind of dumbing down their proofs to work with ethereum so supporting ZK proofs verification is just having like the right set of curb the right set of like features built into your chain that make it easy to verify the cryptography and some chains have it some don't salana is working on it I've got got some more shout outs um what we've got aligned Lair and ZK verify that are kind of leading this like verification love RJ and team shouts out aligned yeah they're homies modular modular ZK verification so in the sense of like you know an ethereum that doesn't quite have the verification capabilities baked into its

chain Scott and space and time are going to serve the proofs we need to we need someone to come in and verify the proofs and then ultimately interact with the chain walk us through like you know how a modular verification layer leverages space and time and ethereum and ultimately P you know completes this picture also one thing uh Scott briefly what's the like what's the be case for these verification layers as well because from what I understand there's different ways that l1's can just kind of move around them as well yeah you knew I had to have a hot take you knew I'm ready to go with the bare case for anything um yeah I'll start with the bear case

and we'll go to the the bull case sorry Rob that's okay I think I think l1's baking in the verification is where Scott's going with this but maybe mean you you already know the answer right which is I can already verify space and time proofs on ethereum I'm not saying you can't verify proofs on ethereum you 100% can't you just have to it's just slow and inefficient you have to dumb down your proof to get it onto ethereum and it's expensive expensive and inefficient because ethereum doesn't have the right curves if they did we could be a sports car Ferrari just cruising on ethereum but that Arbitrage presents an opportunity for a line layer shouts out

RJ and team love y'all um they could offer a cheaper quicker you know don't pay ethereum gas instead just pay pay us a few cents we'll verify you we'll sign your proof and we'll relay it on chain for you and so your gas is pretty minimized bare case is for example salana realized by talking to builders like space and time they said hey what curves do you need we want to make salana crazy freaking fast for ZK verifying what curves do you need we answered that question and they're working on that so like I think ethereum will be slow and I think a line layer has a great opportunity in front of them for

supporting the ethereum ecosystem but I think l2s will be quicker to adopt the right curves and I think um you know chains like you know Salon will too and for any of those unlaunched chains that that you listed off what you know that may may be listening I no idea no but what what are the right curves like if you were to give them a recommendation of which curves to implement which ones are those BLS 12381 is a common one that a lot of teams need including space and time and BL s 12381 is not something that ethereum supports for the Nerds listening they know what I'm talking about they're probably begging for BLS 12381 right now in some Forum somewhere I am not begging for that

however Andy thanks cat appreciate that um so coming to a close here I think what what would be helpful for us um and for our audience is to understand the practical the Practical future of crypto as we know it on chain and interacting with smart contracts whether via natural language as running an agent that is kind of your agent assistant like some of this DF stuff or hey you pronounced it we now know how to pronounce it dude e f a i don't even know what's going on that Danny guy is in or Antarctica Shing DF DF whatever it is knows what's going on there a lot of lot

of short termism happening in the AI space these days um anyhow um right so like we moving forward and we have uh natural language you'll be able to implement and write what you're you know expressing intent almost uh which is kind of my current mental model of chain abstraction intents is very much like I'm G to enter my natural language into a query it's gonna operate it's going to send a command to an assistant agentic assistant right but the question here is like with that in mind whether agentic or human interaction when space time mayet is live and in the Years following like what is the material differences that

you expect to see in the space What are some of the zero to1 ideas that that you think are yet to be explored because the lack of infrastructure specifically around Z Cod processors and um what like do do you guys have anything up your sleeve for main net there's like four great questions in one I'll try my best to answer all four stuff up Our Sleeve for main net yes um we got some real tricks up our sleep especially around launching this AI Studio that we've been talking about on this on this podcast is like hey here's space and time chain now your smart now your dii contracts can query space and time and get back CK proven data and you as a dii developer can build uh more sophisticated data driven protocols oh

by the way here's a AI Studio to build those datadriven sophisticated protocols if you don't want to write code you can just generate that crap we're excited and that'll be all around late q1 um the agentic future that I see crypto enabling or sorry I should say the the exciting future that I see crypto enabling that like some zero to one is is that like crypto enables crowdfunding and uh permissionless Finance use cases that you can't do outside of crypto and like I I see some people on crypto Twitter complaining like oh just because like you you built a little Twitter bot and launched a token for it like that's like web web 3 is two years behind web 2 like web 2 has already way

better agents way more sophisticated Tech and I'm like you're missing the whole point we're not trying to outcompete web 2 we're not trying to play those games our what crypto enables is whole new games with a whole new financialization model with permissionless Finance where the chain becomes like a permissionless stripe becomes a a an Internet payment API for websites and web apps and agents online to like transact in a permissionless way just like they would stripe or PayPal but like without any accounts like any banks and so that plus the community aspect the crowdfunding like the community socialization aspect of Twitter will continue to spawn these new protocols that can only be enabled through crypto

and an example is like all the AI agents on crypto Twitter that like built their own communities autonomously with their own token that's something that we'll never see in web 2 that couldn't have been done in web 2 it just doesn't have the right mechanics so we'll take web2 Tech and we'll build these new protocols that are very Community oriented and require some kind of permissionless finance and that's going to continue for the next five six seven years beyond that it's Quantum and that's like that's Way Beyond any of us and we're all screwed then so let's just enjoy the next eight years love it yeah I I think it paints a very Vivid mental model of uh how we leverage ZK Ai and and more emerging Technologies

on chain in the next three to five years so Scott thank you so much for coming on man really appreciate this conversation learned a lot yeah Scott thank you did a great job of uh translating right curve topics into left curves thanks appreciate you guys having me on it's been a blast great to meet you both talk soon my Che thank you