The Next Chapter of Modular Blockchains: Proof of SQL w/ Scott Dykstra (Space and Time)

Transcript

[Music] hello and welcome back to the strange water podcast thank you for joining us for another episode one of the most exciting parts of working in an industry so close to the bleeding edge is that we get to watch in real time as new research Finds Its purpose in the field in zero knowledge cryptography we're watching as this brand new research is expanding the purpose of cryptography from just encryption to the key enabling tool of decentralized computing ZK Roll-Ups are perhaps the most Salient example these days

computation inside of ethereum is what we call trustless but the same properties that make it trustless make computation very expensive and if we were to just naively upgrade ethereum with more computing power we would also lose everything that makes ethereum valuable but ZK cryptography gives us an almost magical solution we can do all the computations somewhere inconsequentially inexpensive and just record the result we found the bottleneck and ZK elimin ated it well execution isn't the only bottleneck on ethereum storing data is

prohibitively expensive so expensive that we basically Only Store strings and numbers and even that gets ridiculous I mean just look at inscriptions storage is so expensive that the idea of a smart contract having a large scale database is so outlandish that it just doesn't quite track right but that's the exciting part of working in an industry so close to the bleeding edge you get to watch in real time as research Finds Its purpose in the field you get to watch ZK transform ethereum into the world computer now today we have Scott dyra

the co-founder and CTO of space and time at its core space and time is a cryptography company founded to answer a very very specific question how can we use zero knowledge to prove SQL the result just like ZK lets us offload execution it will also let us offload storage a smart contract can encraft a SQL request send it off chain receive back the result and approve and once verified be assured that the result is as a as good as if it were in the trustless evm this conversation is an incredibly fun and wide ranging conversation on

everything from the interaction between Ai and crypto to the importance of capturing higher and higher levels of the value stack I learned so much from Scott and I know you will too one more thing before we begin please do not take Financial advice from this or any podcast ethereum will change the world one day but you can easily lose all of your money between now and then okay let's get started Scott thank you so much for joining us on the strange water podcast yeah great to be here Rex for sure so

before we get into space and time and like all the exciting things that you're bringing to I guess both the blockchain ecosystem and the regular developer ecosystem um I'm a huge believer that the most important part of every conversation are the people in it so um can we talk a little bit about who you are um how you like found programming in crypto and I guess why you didn't run away yeah um background in computer science did undergrad in compai and Masters in data science at Berkeley and then uh got into crypto kind of um in 2020 was starting to Goof Off with contracts but not really really more 21 when I got real obsessed with it like everyone I was buying crypto for a while

but not honestly didn't know much about defi uh until some close friends put me on to it in 21 2022 or sorry 2021 okay um then I spent uh spent a spent a decade in data warehousing at the Enterprise level like building Enterprise data warehouses for the Fortune 500 and it was a very eye openening learning experience kind of boring you know it's it's big database systems distributed scale 500 terabytes across these big clusters of compute in the cloud or on premise and data centers and uh a lot of SQL a lot of data analytics and it was all about taking the internet's data and basically siloing it off to these blackbox

analytics systems owned by Mega corporations that essentially steal users data and use it as a product and I contributed heavily to that you know and it's exciting to not be anymore and to be working on something that's sort of the opposite the antithesis of that sort of the uh decentralizing the world's data and working on a Sol ution that puts ownership back in the hands of the internet constituents with space and time yeah and I I want to Circle back to um just your entry into crypto and and Defi and realizing that there's an opportunity here but while we're talking about your history in um data warehousing so brief background um I used to work for anheiser Bush inbev and I was on the my first job was in Technology Innovation and one of the

like major projects that I contributed to was you unironically using micro strategy something no yeah I maybe at the time that that name wouldn't have meant anything and so I don't know but the big things that um you know we were tsed with which anheiser Bush at least in 2013 like was still using Magnetic Tape was still using like machines from the uh like 70s that like were just not working and so we were tasked with finding like a a good data warehousing solution yeah bi tools so you know we worked with sap business objects we work with click view we work with like herok or um just I can start listing all the names and I'm sure that you're familiar with but um I think like there's a lot

of like really kind of crazy stuff that you don't realize without getting into it and I think from your side you can talk about the technicals but from my side I think it's just like the realization that for all the data that's created in the world it is like completely inaccessible to the decision makers that like could actually do something with it and I I would love to just hear some of your Reflections and your thoughts on like as like the internet and technology and computers are so new um relatively speaking like how how have you been watching businesses transform as we've entered like the information age and like really do you let me just be like really Frank with you my job was to get

as much data into dashboards as possible so that like Business Leaders could make decisions and the reality is is no matter what we put in front of them they were making decisions based on like gut instinct or like the the conversations that they knew they had to have with customers or like just or like a great data engineering business analyst team that built great reports but the executives don't even trust the reports they're like you might have hard on this but I'm not sure I even trust your data exactly so very open on the question but like how do you think about the transformation of like corporate business during this information age and do you think that like um there's a lot more like opportunity to unlock decision- making uh using you know big data or do you think that like the

actual just getting things up and running we've already like kind of achieved that goal we I think we've overachieved and now we're actually trying to scale back uh maybe mildly controversial take or hot take if you will but uh listeners we'll get into ZK in a second I'm gonna ramble for a minute on datawarehousing and business analytics we'll we'll get into zero knowledge and why that's relevant in a moment uh I think we've over indexed like I think we spent the last decade just adding tool after tool after tool you you experienced this when you probably PC five different bi tools including click viw and sap Hana and name your other and I'm sure micro strategy at the time I'm tongue and cheek saying that but um what we did is we we started to build all these data

silos of like pile on database tool after database tool for storage and and like processing the data like actually tools that actually stored data and execute query and then we piled on BI tools analytic tools I click view on top of that and then we piled on python data science devs on top of that just building who knows what and then we piled on Jupiter notebooks and it go you know and then and then and then mlops and it when you added this fifth layer called mlops it just got out of control right I mean the corporate spend the it budgets to do all these Green Field projects where you're just stacking on tool after tool after tool and then by the time you built the fifth layer the first layer is already Sunset and

deprecated by the time you get to mlops the original data warehouse you built isn't adequate you have to Sunset it and build a new one and it got out of control and my career has been selling and building building and selling those products and enabling that like um proliferation of too many tools for too many jobs rather than right tool for right job and I'm hopeful that AI actually solves this we've been over the last few years scaling back like there's this philosophical movement or this like push to a small degree with tools like mother duck and other database solutions that are actually purposely built to be Micro Data warehouses that are like hey remove all this blow you can run this thing on

your laptop like you don't have big data you don't have 40 terabytes of data you could just run our database on your laptop with your half terabyte and call it a day I don't think that I still think you need a data warehouse I still think big distributed systems are you needed because the data volumes on chain are are large like we're talking terabytes per chain and the data volumes at the Enterprise grade and at and in in video games and in social media apps are all large so I still think you're going to scale past the laptop right the the little Micro Data Warehouse solution is that's the antithesis to this 10 years of stacking on data warehousing tools is not the right solution but that's the trend we're kind of trending back towards removing these

tools now I I'll end here I apologize for the ramble I I believe AI actually helps I actually think there's a potential for us to use AI to solve this problem because when you have ai it can replace you know 10 of the 12 Tools in your stack you need somewhere to store structured data you need somewhere to store un unstructured data and you need somewhere to Vector search you need somewhere to retrieve via SQL and retrieve via Vector search and that's about it as long as you have both of those you could just put an L llm on top and the llm is your bi tool right it is your mlops tool it is your python data engineer it'll do all it'll be all of your business analysts and your data Engineers yeah and I think when we start

talking about what space and time is and like the UI that like I can already see today like that will become much more apparent but uh just to wrap up this part of the conversation like I think a huge part of this is that you're exactly right like we over tored for just like putting basically more impressive looking graphs in front of Executives and and then telling them like you have like business tools now and like I I totally think everything you said is right like the the next phase is going to be about refining that and making it smaller but the other opportunity that is so clear to me so quick antidote um after I left uh Technology Innovation which is not a real job that's um I I worked in cash flow uh forecasting which

is you know like payments have to go out payments come in and then every single day you just need to make sure that there's enough money that um you know that that business can continue flowing and for anheiser bush in the the I was in charge of Canada and us that's 20 billion dollars a year of cash flow and like every single day we had to forecast and then around the end of the year Things become like particularly important because not only you worried about like the actual dollar balances you're worried about like hitting your yearlong numbers right so one year we were just like off every single day and finally like the CFO is like go figure this out like why aren't any of our numbers coming in and so I flew to our call center in St Louis where like all

the guys who actually did the work were you know doing the inputs and we finally figured out that the way the second largest cpg company in the world was like creating their like intricate daily cash forecast was like some guy who'd been there for 20 years was guessing every day and he's good at it he's very very good at it because he''s been doing it so long but like it was literally just a the Excel spreadsheet from last year copy pasted over and then him like altering the individual numbers based on like Vibes you know think after a decade of us trying to build tool after tool after tool and automate ml process after analytic process we still do business on Excel that's how most executives are still like just

guessing no I mean I can go all day on Excel but I I guess to wrap this up like the the Insight that I see here is that not only is there a huge opportunity for AI to help with the like like finding insights there's a huge opportunity for AI to help us like actually take all of these like data inputs that are just like too dense and like not structured and not like consumable by humans and just like get them into the databases in the first place and so like that's like just a huge opportunity that um I I feel like it's the flip side of the analytics three things that AI will solve all related to this mountain of AI of of data tools that we lost ourselves

climbing three things that AI will solve is data ingestion like you said getting wrangling structured and structure data into a format that can be stored and retrieved uh because that's a miserable painful and a boring process today like noers don't want to do that yeah I can do uh it'll it'll also solve the the analy problem right you instead of instead of a mountain of bi tools you just need a vector search database and a SQL database and you just give the llm access to it you just ask the LM questions great or have it generate Python scripts for you and then finally it'll solve the um just the the complex question problem which goes beyond SQL or python or MLA it's it's a question of

like hey a business user or a business leader has a question that's a complex question how do we solve it where it's not deterministic it's not as simple as just writing a SQL query gives you the answer to that question it's a question that requires a deep understanding of the business looking at structured and unstructured information and making like subjective and objective decisioning well where AI can replace an entire analytics team classically it was like okay you have analysts that are more business focused that are looking at data in charts and trying to map data to business decisions you have you have Engineers that are building all the charts and building the environment then you have business users that are like setting the requirements in well AI can be that whole team like my requirement

is answer this very complex question about my business that requires SQL Vector searching and machine learning processes all together to answer llm can answer that that that's that's good inference for the llm to do for sure man super exciting so let's uh let's move on and try to get us to space in time so you're this is in 2020 2021 you're messing around with D5 this is also I entered during that cycle as well so probably at the same time that you're getting smart I was like buying polka dot and cardono and stuff I've never touched r yeah all both those both of those ghost chains yeah so can you like just give us like the brief story of how you went from like messing around maybe with some smart contracts maybe with

your money to realizing like there's a huge opportunity to build like core infrastructure to make you know this really crappy blockchain computer into world the shocking realization that smart contracts don't have access to data um we were my co-founder Nate and I were fortunate enough to be uh introduced to web3 data by chain link uh my co-founder Nate was an adviser and working with the executive team at chain link and chain link became our first partner and kind of our first supporter and we kind of born space and time within the chain link ecosystem and we're close very close with those folks over there and go to smartcon every year when we looked at chain link we saw okay

here's an opportunity where you have a very strong business very good Tech very strong team and really all they're doing is just finding a reasonably secure way to get some data to a Smart contract because smart contracts have no access to data and we're not talking about data availability don't get confused data availability is a whole another term that got sort of terrible name terrible name I don't know if it's terrible it's just this it's just it's just too um it's a name that will confuse people because they'll think about data availability in within their own context um anyways getting access to onchain data for a smart contract now defi summer was um facilitated by really by oracles

bringing price data to Smart contracts like not a lot of discussion about this but I mean if you really think about it what what really facilitated defi for the first time was smart contracts being able to know the price of a token right like once you have liquidity pools that can quote prices then you need smart contracts to those prices hence chain H this idea of like yeah sure a smart contract in the evm may not be able to ask questions about activity on the evm like a smart contract can't say hey show me how many transactions Scott has made and the total volume and the total lifetime of Scott's wallet that's not a question that a smart contract can ask on chain but maybe you know they could ask what's the price of avalanche

aggregated across all Unis swap pools that quote Avalanche and so that Genesis of tii with like providing data to Smart contracts to make them smarter allows you to step back and say okay what's chain Link's real business the real business actually just making smart contracts smarter and building a whole Suite of tools to facilitate that the first tool that took off was price feeds so then thought all right well let's take that a step further like what smart contracts really need is a database where they can store and retrieve data they can offload computation and storage from this from the chain to a database but it's cryptographically secure it can't break the trust model of web 3 it can't break the security of the blockchain it's got to feel like an L2 you know it's got to

it's got to have that L2 security connection and then finally it's not going to consensus probably isn't the right approach and uh it needs to be pre-loaded with index data from that chain like if you're gonna allow smart contracts on polygon to ask questions about activity on polygon whether it's price whether it's defi questions or just like General Activity questions like show me all wallets that have at least a million dollars of madic and have made 10 transactions on chain if you want to enable a smart contract with accurate cryptographically guaranteed tamper proof data that answers that question uh you're gonna need to index the entire polygon chain into a database that can run that query and answer that question and so we thought okay

well zero knowledge this was the early days of plony and Halo and people starting to think okay maybe ZK can present alternative to consensus the chain link approach which I love is let's set up a network of offchain nodes that are not really a blockchain but a decentralized network of nodes that come to consensus and all validate data and gossip with each other that they've all validated the same Data before relaying it our contract that asked for the data the ZJ approach would be this index all the data from that chain into a database in a ZK compatible format and then allow a smart contract to ask ask any question in SQL against that data build a zero knowledge proof

once on one node in the network but verify it in a contract on the L1 that requested it verification can be redundant via consensus on the L1 that requests but proving should be done once via ZK why because oftentimes we're proving against large data sets I mean just the ethereum logs table we have in space and time is nearing a terabyte we're we're talking about reasonably complex queries against re reasonably complex blockchain data sets and um you you kind of need like a whole GPU you need you kind of need server to handle that and you're not going to handle it via consensus you can't have

20 big servers all redundantly running the same massive query against a terabyte of data each and then come to consensus on that output a going to work the reason it works for for chain link with consensus is because they're coming to consensus on small data sets it's nodes each grabbing the current price of avalanche that's doable and they they rocket that the problem is if you want ask a complex question you can't answer that complex question 20 times redundantly come to consensus and then relay apologize for the beled answer but just want to make that clear and when we realized that that was the realization of okay let's go build a database that's ZK powered and let's go solve this problem and build like the web 3 native

database where a smart contract can talk directly to a to a database and offload all of its Computing storage but also retrieve blockchain data that's space and time man that it's super exciting and I think like the the if I'm like framing space and time kind of in like The Narrative of where we are in the development of crypto I think what's really exciting about the Insight that you guys have and and let alone the fact that you actually built it is that we've all been talking about the modular blockchain thesis since like literally since I've entered crypto so in 2021 like it was already there I remember there was a bankless episode about that I was super excited like it's been there this point I'm like too afraid to ask is modular literally just these did these

dudes just remove smart contracts and then just call it call it modular I'm I'm saying that kind of tongue and cheek but I'm als serious yeah no I mean I I do think that um I don't definitely don't want to throw shade at bank list but like I there and and and just like that whole millu of conversation but I do think that a lot of people that use that terminology have no idea what they're talking about and therefore we all have no idea what we're talking about but I me I don't I think of modular as like a blockchain minus smart contract execution yeah well I think I think you know yeah I think that's like a good way to look at it especially like in in regards to the L2 frame right we have like we take um the execution out

of out of the um main net and then like because there's a whole execution Engine outside that kind of plugs back into mania net I think that's supposed to be like the modularity and now we're entering this like data availability Paradigm so like you don't necessarily have to use the straight up ethereum data availability so that's like another module plugin and so like whether or not you really buy it saying there's good good point data availability is a good point because what you're really saying is there's another layer of there's another module it modularity isn't as simple as just removing smart contract execution it could actually be more keeping smart exec smart contract execution but removing chain State blobs exactly and I think the whole like concept behind the modular framework is

that each individual application can pick and choose the pieces that like work for them so let's say that you want to be an L2 that um like really really is like ethereum secure and so you're not going to choose to use like a Celestia or igen da because of like your purposes like modularity gives you the f F ability to make that choice while this other like defi generate chain is like we just want everything to be as cheap as possible so we're going to like kind of handwave at decentralization by like throwing data out wherever it needs to go and then move on right and I think that literally space and time is the first um company or or group or whatever whatever we are in this industry that um

looks at a new part of modularity and that's like actual databases and I think it's like before space and time we just we had databases I guess they were equivalent to like our TI 81 calculators just within the evm like framework but like what you guys are doing is allowing these this modular um system and like plugin to say like okay if for my D app I need like really true deep databases that um can support like a game with 100 million people or can support like we can get into some of your examples later but support tra ing like space debris or or any of these things that are like wildly complex every bullet fired in the game

that's going to leave onchain reward yeah and like what you guys are doing is saying like Hey we're creating the interface and the first example of like a database module um and if if ethereum is to become the world computer like we're going to need databases yeah and we we actually just published a white paper today our first like big announ like hey here's the space and time white favor and that's pretty much the whole thesis of the paper that's pretty much the you just you just summarize probably the first seven pages yeah well just full disclosure I did not read the paper but I just think that what you're building is like so kind of clear and once you like really have clicked into the fact

that ethereum is not really just like a casino or this like weird internet thing that is about like Defi and tokens and locking it both of those things it is only both of those things yeah no for sure but I think that like once you really buy into the concept of ethereum is about decentralized compute uh like the the necessity of like well how do you build like really deep modern complex databases into decentralized compute like space and time like the question is answered and what you've built sure and and like there was a a moment in 2020 2021 2022 arguably even into 202 three where I truly believed in decentralization and in ethereum as that

like it future world's computer that will decentralize the world's compute now I'm just like bull runs back baby full DJ know pump my bags I'm sure we'll go back to building next Bear yeah no yeah it's I hear you I hear you and that's sort of the thesis of our paper is that like hey we believe uh especially now in the generative AI world we're walking into like verifiable and Community operated compute is Paramount because it's the antithesis to a generated world to a a a very non-deterministic like everything's fake because nothing's because nothing's generated or everything's generated therefore nothing's real in that kind of

world you need a cryptographically verifiable system to move money to perform to move data to do anything where data is valuable if you have valuable data where there's a financial incentive to manipulate it tamper it change it tack it breach it in that case you need to go way way above and beyond to build cryptographic new modern web 3 facilitated cryptographic uh um capabilities that counter this generative AI World we're walking into and so like you brought up a video game hey hey Scott like you can't just store terabyte to video game data on the world computer that is ethereum so space and time becomes this conduit where you can store your terabytes of every bullet

fired in your first- person shooter and connect that to a Smart contract just in time when needed where the smart contract can say hey how many how many hits what's what's Rex's kill death ratio and should I give him a new reward for having a positive kill death ratio the same uh take that a step further and say all right where are all the use cases where we feel like data might be manipulated or where the data is going to secure Financial value or where Financial transactions are going to recur as as a result of this data being changed manipulated tampered anything right and you you enter this AI world where all of data is going to be

synthetic it's all going to be created generated and so the only thing that matter will be the data that's not the data that's not synthetic that's actually facilitating real transactions and real value and that data has to be cryptographically secure so if you just take a postgress database or a mySQL database or snowflake or my almaat teradata and you run your terabytes of video game data there and then you try to ask questions about that data from your smart contract on ethereum you've broken the cryptographic security of web 3 you've broken the trust model or trustless model of of the blockchain and

your data will inevitably be man ated by your own employees or third parties reachers or even partners that you give access to your database and forget to revoke their access and then six months later you realize they've changed the price of t Tesla stock on chain and minted themselves like a million dollars of rewards I'm trying to what I'm trying to hint at is like we need a database in ethereum but that database has to keep the trustless model of ethereum yeah I mean I'm so glad you said that because I've said this on this podcast a many times but like I just I I don't know where the intersection of AI and cryptography truly is like go to Twitter you'll see like a thousand tokens will

tell you exactly where it is but I think like what's so clear to me is these are two sides of the same coin two pieces of the same technology because as exactly what you just said like AI is about generation and abundance and like just like pumping out as much stuff in the internet as possible and crypto is about scarcity and blockchain is about slowing things down and like just to it it's to me there's no coincidence that these two technologies are really like blossoming like literally within the same months and years of each other um I just uh I think there's like a lot of like questions to be answered two sides of the same coin is a tough one for me because

you're crypto is the coin at the is the is the the cash that you hand to the cashier the blockchain is your cash register and crypto is or sorry AI is the cashier in my my very humble opinion and if you if you and make it an if the if it's analogist to that the whole two sides of the same coin thing may not be a the best explanation I'm only I'm only laughing because I saw a tweet about that the other day about someone's whole argument of like is AI and crypto two sides of the same coin yeah I mean look I don't I don't know if we need to spend that much time on this but like I will push back and be like no man like it they definitely are and I think that within

the context of like the world we're in like you're totally right like crypto and blockchain and AI like have like specific roles to play in finance and all that stuff but like if we're just talking about the greater pieces of the technology like the the best example is like generative images like how are we ever going to be able to tell what actually happened in the real world once like chat gbt 10 and mid Journey 6 and all these stuff is out there and it's like that to me like as we create these generative Technologies we also need to create scarcity Technologies and like yeah I think there's going to be some like as a eth maxi and a bag holder like I think there's going to be some

connection back to like ethereum blockchain like for example like we could create a startup that just goes and finds all the content on the internet as it exists today create a commitment and then get that commitment on the blockchain so that in the future like you can have a browser add-on that says like hey here was every single time this image was seen and therefore you know that it wasn't created you know this historical image was actually historical but like blockchain and and eth and like all of these tokens that we're creating like that's almost Superfluous to the technology conversation and I just like really believe that um like all the things that like the worst things that could happen with AI like

are totally possible but the reason I'm not worried about it is because like in conjunction the smartest people like yourselves or like uh Jason Morton from Ezekiel or like any of the the people we've talked to on this show are all like realiz are all growing these Technologies like kind of in Tandem and so um I guess like love to hear your response but like that's why I really do believe like two sides of the same coin in a in a in a generative world world where most data is synthetic the only thing that matters is CK yeah yeah yeah wow very very well put yeah and uh you know crypto's the cash blockchain's the cash register daps are the point of sale system AI agents are

the cashier what does that make ZK ZK is the the iron the armored truck out front the the armored truck out front the grocery store or I don't know I'm I'm I'm really reaching for an analogy here but all right well let let's move on from the tortured analogies and like let's get actually into space and time so I um was watching your product demo um at consensus I believe and I you you put it very beautifully it you said that it was five things in one so I wrote it down I'll just list it for you so you don't have to come up with the top of your head but you said it's a beautiful a sorry a beautiful UI a very powerful database a validator consensus layer and a a

novel ZK proof and a blockchain indexer so with that kind of as the framework can you uh just like help us understand what are you guys building yeah yeah yeah this is our biggest challenge honestly space and times biggest issue as a business is not a technical issue it's not a it's it's a messaging issue we know exactly what the market needs and we're building that but it's a very it's a platform it's not a specific little product right and that platform has five or six sub products in it or you know layers to it um and so we've really struggled right because because because people go on our website and they're like seems like you guys are like kind of like an AI company doing like AI

SQL or you're like an indexer you're kind of like indexing blockchain data you're like the graph others are saying are you kind of just building like a database you more like Dune doing Analytics what what what is space and time and the reality is most of our R&D is just zero knowledge right we're mostly a cryptography company uh building ZK for SQL what just having a cryptographic circuit to prove SQL does not provide value to to developers and certainly does not provide value to business users so had to build like a whole platform around this cryptographic the crypto we could

have just said all right space and time your whole business all of your engineers every dollar you spend is just working on proofs but to make it valuable you need index blockchain data right you need you need a two-way street between your prover and these different chains so have to index data from all these different chains in a ZK compatible way so we have to build and invent a ZK compatible indexing Solution that's cross chain uh and then might as well build an awesome user experience if you can nobody wants to write squel so if AI can write the squel for you in a in a in a beautiful UI that would help we're not trying to compete with Dune necessarily I I actually think

pretty highly of Dune but hey you've seen our user interface you see you know it's possible with good dashboards and with Aid driven analytics my I did spend the last decade building analytic tools it is my passion uh long answer to Simply say space and time is a ZK proof our core engineering our smartest folks are just cryptographers working on a novel ZK proof to prove queries but to make that useful to web 3 and connect it to every major chain and make that a you know a cemented institution of a primitive in the ethereum ecosystem

we're going to need a platform around that proof so we have our own native database we built for the kind of workloads we expect dap developers to put on our work our database you know backend database for daps we pre-load it with index blockchain data we give you a nice AI powered user experience on top but the real core Tech is ZK yeah that makes a lot of sense and I the Insight that I came to really last year hopefully we're entering a new paradigm now but like at least at the end of 2022 and 2023 like ZK is so new and moving so fast that like it it it didn't really make sense to really be building ZK products yet like all of the companies that were getting funded and we having a

lot of momentum were just like trying to turn Advanced like crazy number Theory into like just packages and tools that were familiar to app developers right like apis or sdks or that kind of stuff co-processors that kind of stuff and I think like what you're saying is that you guys had that same Insight that like let's create this ZK that does something functional let's wrap it in a in a package that is actually usable but then you actually took the next step which is let's build yeah let's build actual useful applications that leverage this technology um and like and yeah I mean I think you hit that like eight different

times on the stack whether you're talking about the you know the everything from just the actual movement of data in and out and how that happens to um you know like complex queries and that kind of stuff but like really you are let me say leading the charge in what I think is like the next phase of ZK development which is like how do we take these like incredibly impressive tools and Primitives and how do we make actual applications that people need I I I kind of like worry that some of the ZK like Ezekiel you mentioned them really cool really exciting project never met the team would love to awesome project

very exciting I I do Wonder though if a lot of these very very exciting technology projects with very technical leaders are maybe going to struggle a little bit to go to market because they haven't moved up the application stack and they don't have killer apps building on them yet now I could be totally wrong maybe they've solved the killer app problem if so hell yeah no and and this the point of this conversation is not to call any like succinct Labs is doing the same thing a is doing the same thing but let's talk about succinct for a second because they are moving up the stack that is an example of of another ZK project that's an actual platform like Space in Time succinct went straight to platform building rather than straight

to uh R&D around like new circuits we did both R&D a new circuit patented it invented it and are bringing it Target and we have a team on the circuit but we also have the platform so sync is a platform play that I think is really going to be valuable if I was a VC I'd put some I'd be calling up C synct right now um and and and by the way me personally I will be calling up C sync because I want to partner with them and like make space and time a circuit available on their platform this is this is like a this is a great example of moving up the application stack where they're they're bringing experience to the to the developer on chain and they're creating a nice interface and a nice user experience and they're

abstracting away a lot of the complexity that's the key here when you move up the stack it's about abstracting away complexity axiom's try to do that via JavaScript hell yeah the sync is doing that via this nice platform for like like ZK Ops if you will space and time's doing that via SQL we're all database folks and we figured hey sql's easy it runs the world it's a great interface for accessing data let's make our interface be SQL no but and again love all these projects every project that we've talked about I've had on this podcast and So like um I I but I do want to distinguish with you guys because like yes axum is synced and even aiel they're all like basically trying

to make complex things less complex and allow developers to actually focus on developing applications instead of number Theory but like you guys are are taking that further up the stack and like that is what I think is like so awesome so can you talk a little bit about like the UI that I've seen and for those that haven't seen it just start from Basics like when like what what's going on in that like you can ask like a plain English question of like hey help me understand this specific thing about the blockchain how does that get translated into like something machine usable how does that hit your database and like talk a little bit about like the true like human level um applications that you guys are already

able to build it from that yeah yeah absolutely we we uh we're fortunate enough to be backed by Microsoft they LED our last round and they're a great partner with us and they've been looking out for us so they gave us early access to gp4 earlier in the year we did a collaborative project with the AI Lab at Azure our our Engineers work with the Azure AI lab and we built a pretty sophisticated I mean arguably the most sophisticated prompt to SQL system that exists today now that's a y combinator application right like that's that's a whole startup in itself if we if we had the energy we could spend off a whole another startup just doing AI powered bi

AI VI right um I just don't care crypto is so fun the crypto where where we need to be so my focus on AI is really only to enable better access to blockchain data all of my energy that's going towards prompt to SQL which I'll describe in a second is really just to facilitate a better develop experience finding your blockchain data and cering it and understanding what's happening on chain but if we had the energy we could take that Global to just be like a better bi tool for anybody if we like it so here's how it works if you're familiar with database tools they're awful and they're boring

and you have tables and you got to go explore them you got to find you a new employee at Walmart on the business analyst team has to First go log into six different databases and find the table they need and understand where the data is then they have to write these awful queries that start to join data sets together and then they can start visualizing things and putting it in charts and sorry just as like a as one of those analysts like half the time too and this is why it ends up in Excel is like you have your like social media like modern tool and then you have your like terrible tool from your like 70s Mainframe and then you have like the new like terod dat cluster that we just set up and so like none of these things are

in the same formats or even talk to each other or whatever so you just dump it all in Excel and then like hope that the 21-year-old like brand new analyst is able to craft it together in a way where you can kind of create a story that the CEO will kind of buy yeah exactly yeah that's well said we saw an obvious opportunity to fix that with AI and if we like I said if we we had more energy and more engineers and weren't as into crypto maybe we'd go build a business around this because it's gonna be the not is it it is the future is here and it looks like this you just log in this bi tool that is called space and time and you just

start asking questions you know if it's we'll use Walmart as a dumb example like if if it's Walmart data in space and time a business analyst at Walmart logs into space and time they have a text field and they just say hey what's going on with West Coast sales this month why why are things down so bad in the west coast that's a complicated question that's probably going to require a lot of query you know six different queries to three different databases underneath once those results come back as you know then probably some data engineer needs to run some python pandas process to like and it just becomes this like complex question where whereas the llm can just figure out what are those four four SQL questions we need to ask and just kick them off now we're not quite

there yet we will be soon today what we've deployed is just like one query at a time ask a question at a time you got a question ask it uh SQL gets generated for you and that's accurate space and time SQL and then you can choose to route that accurate space and time SQL request to a ZK prover or just to a regular database engine that's cheaper you know we have a normal you know a centralized database as well as our ZK proven which more expensive but cryptographically guaranteed so you as the or you just choose where your query goes AI writes your query for you you just choose I want do I want to ZK prove this query or is it more just to power a dashboard I don't need ZK Bruce and you chose your costs associated with that that's what exists today and we've done

a we've done a great job it's in the market got a ton of daps using it already about 8,000 active users actively running queries and writing prompts and it's exciting but this is the tip of the iceberg right the next step is just prompt to dashboard not prompt to query prompt to dashboard today we're kind of in the middle we're kind of prompt to query to chart we gener we get we get you all the way from a simple a natural language question all the way to a chart showing the blockchain data visualizing it but then the final step we're going to make in 2024 is get you all the way to a dune style dashboard from a single question the question would be hey

what's going on with chain link ccip uh yield like what's what what's give me some insights into chain link ccip gas usage and then an entire dashboard gets generated for you about chain link ccip Dash gas usage and it's roughly what a business analyst would have asked um lots of things to talk about here but just uh on the distinction between like uh running your query in a ZK prover or not like I think what you're saying is you're essentially checking you're quering against the same data you're using like all the same technology to um like get the answer but in the ZK prover situation you're also generating a cryptog sorry a cryptographic proof that

says like if this verifies then you know that that data came from this database is that correct yeah do do you want to talk for a second about that cryptographic proof yeah so I yeah I guess like the the the couple of things I wanted to ask you about this first are like um I would love to hear a little bit about like what is the ZK proof that you guys created and talk through the process of like what all the work that you had to do and I'm very curious about that but the other thing I'm curious about is like data is only you know the the famous saying right garbage in garbage out and so like there's one thing one side about proving your queries were executed against the database and then there is another side about proving that the database like keeps its Fidelity over time and so I

like I'm curious about your thoughts and how space and time is like the the proving the query very clear to me but making sure that like yes the query was run but was the data that it was run against like still maintaining the Integrity that we expected to like how does space and time solve for that so two questions very open-ended I'll let you kind of go from there sure I'll I'll answer the latter first um what we're what we're proving is that the underlying indexed blockchain data or offchain did that you loaded from your video game it doesn't matter the that once data entered space and time it hased that's the first we're proving that the underlying tables have not been

manipulated also included in that proof as part of what we proving mathematically is that the actual computation of this of this SQL query execution was done accurately so that we're we're proving that the node storing data didn't manipulate it and the nodes Computing the proofs did so correct like actually ran query you wanted them to run and didn't leave anything out because a simple attack Vector would be don't tamper with the data but just tamper with the query result against it right like add your W sneak your wallet address into the queery result when it's when a smart contract is saying give me a list of people to reward just sneak your address in there so you get a free reward we're proving those two things

we're proving that the data hasn't been tampered at the actual answer against that data is correct um and then built into that proof of course is proving that our indexed data is accurately indexing to the chain right we double check that that before ingesting it into space and time when we when we grab index data from all the major chains we're we're double checking on injest that it was correct we're proving that during our CK proof so it's really three things proving blockchain data validity proving the underlying tables haven't been tampered and then proving that the actual query execution was done accurately and all of that can be validated on chain um and then I forget the first question yeah just about building like

talk to us about like so you built a proof from scratch right or maybe not from scratch but like you've built something novel here so can you talk a little bit about like what's particularly scary about ZK when you're at the Primitive level is like the math is so complex and so exact that like any tiny tiny error could end up being catastrophic down the line so can you just talk a little bit about like you've now explained a little bit about the purpose of the proof and all the different pieces that are being drawn into the commitment so that it can be later verified but like what's it like to develop ZK from scratch and how do you ensure that the the properties that we

expect from ZK are actually there wow that's a that's a that's a very good question a very tough question answer because now you're really asking hey how did you design a novel zero knowledge proof I mean you always start with prior art Right Step One is look at prior art uh we started with some academic research out of the University of Michigan um around virtualized squl had nothing to do with ZK actually our our initial honestly our initial probably year a whole year worth of R&D we didn't even touch anything ZK related nothing to do with ZK for probably a whole year of R&D before we even opened up plony or Halo

to to even take a look we we we didn't use those Frameworks but we learned a lot from them of course yeah but yeah the first year was just like looking at academic research on this topic and figuring out what kind of commitment schemes make the most sense we didn't know what we were doing nobody does I mean no one no one has any idea this is all Green Field there's no best practices right this is truly novel technology no built approve nobody's built ZK to to process terabytes the entire industry is pouring now billions into ZK mainly for transaction processing right mainly for roll mainly for proving simple computations or sorry proving reasonably complicated computations on reasonably small data

sets a small set of transactions like the most recent batch since the most recent rollup most DK research is is simply how do we prove we did we did this set of 5,000 transactions correctly not what we're doing what we're doing is more the Axiom style which is a new generation of ZK that has come up over the last three years pioneered mainly by space and time but or maybe not Frontier by us and now we've got some awesome players in the space um building around us and it's exciting uh what what this generation of ZK is doing is not

about comp relatively complex computations on a small amount of data relatively simple mathematical computations on massive amounts of data like running a query on the entire ethereum history and saying show me all ethereum wallets that meet these five criteria that's a massive Full Table scan against two terabyte we have to prove in zero knowledge nobody's done that before so the first thing we tried was just like different commitment schemes to see what would work I have to admit even to this day we're still trying out different commitment schemes to because there's still there's still plenty of R&D to do this can be endless there's a whole I see a decade of optimizations that we can make to make

this faster more practical well integrated into different database services and blockchain Services I mean this is a a proof you can embed in Snowflake or ter or big query and we're already starting but there's going to be a whole year a Whole Decade worth of endless R&D and optimizations and trying new commitment schemes and as New cryptographic Primitives can invented and published we can attempt to maybe use those to make things run faster it's very complicated the last thing I'll say is this you might have the perfect architecture for proving SQL queries against a large volume of data but that Arch that perfect architecture for making proving

efficient might make verifying extremely inefficient expensive you need your prover to be fast as hell and cheap as hell and scale to terabytes but you need verification to be gas cheap affordable like you know low low gas gas optimized was the word I was looking for so we made a mistake early I'll admit like we we actually didn't take the right approach from the jump we uh we our initial approach was trying to balance the two balance how much work the approver does versus how much work the verifier does because we weren't actually confident we were going to even

do onchain verification we thought maybe it'd be more like verify on chain link which we're also doing it's another option and and as we learn more we realize nah the future is on chain we should we should verifying so then we did a lot of work to switch cryptographic curves and re you know out a few pieces add a few pieces back in so that have put more work on the approver and verification be cheap yeah I mean I think like that is the answer because like we have the Moors law in the real world right and you can always like kind of count on the fact that um proof verification like will get better but kind of the point of ethereum is that that within the evm it's not going to

get better right like maybe marginally maybe by like a order of 10 but like what we kind of know what we get and it makes so much sense to optimize on for verification that and that's also the elephant in the room right the elephant in the room that nobody wants to discuss is verification costs on ethereum just so you know little Alpha here 2024 is g to be a very eye openening year in terms of onchain costs for ZK I mean if you think inscriptions are bad wait till you see ZK proofs being verified on chain I mean the ethereum gas CS are going to 10x well again I think uh I mean as a bag holder that doesn't really do that much on

chain like I'm happy about that but um I do think like yeah we are entering like an entirely new paradigm and it's because like everything before was just about like oh like it's happening like what little like chachki things can we put on chain and everything that's coming is like how can we deploy like the full force of like 60 years of modern Computing Research into these like tiny proofs and um I was having a conversation with seram from igen layer like about a year and a half ago before um became the most valued project in all of own the entire narrative for 2024 yeah yeah I uh you know I have thoughts

on the like the The Narrative but um the order on the on the narrative around order three derivatives yeah and and it's just like I I think Igan layer like is or was interesting because it was about like how do we like continue down the trustless Computing like and honestly like I thought the coolest thing about igen lay was like oh you can do proof of stake you just don't have to issue your own shitcoin like you can just use eth and I was like oh that's great insight and somehow it's turned into literally just put your eth into a a smart contract to farm a potential shitcoin it's like I don't really know how we got there but I I the comment that he made to me was that if you

actually look at the numbers like we can only ver like from gas limit standpoint we can only verify like 15ish like uh snarks per block and like what happens when we build all of these like ZK oracles and ZK rollups and ZK data availability layers that like yes there's much more computation offchain but like still need to run a snark every single block like just things are I I know what's gonna happen I I'll tell you if you have time I'll give you that answer I mean I assume it's be like a fraud proof Paradigm but yeah I would love to hear an answer oh no no no don't say that word fraud fraud proofs that's a bad word well sorry okay what I mean was that like basically

we'll all just no no what just to expand what I meant like people will just send their commitment on chain and will accept it as true and then if someone sees that it's fake no okay so my point is not the bad word that that is you know uh that to me is problematic because it defeats the entire purpose of ZK what you're doing is they introducing optimistic you're you're introducing optimism to something that was purposely designed to fix the need for optimism optimism is not secure it's optimistic it's it's l optimistic security by default and and when you start introduce fraud proofs you're introducing optimism

to a technology that was that introduces the hope of not needing optimism yeah all right man what's the answer I'm optimistic that we won't need optimism anymore not not optimism the EOS says I love optimism that's how op I mean I'm optimistic that someday we won't need op we won't need fraud proofs for ZK that's all I'm saying there's reason it's it's completely antithetical the solution is something like this you know maybe I shouldn't maybe this is Alpha that I shouldn't be giving away this early because it's something space and time's going to be delivering to the market but essentially the equivalent of like an oracle Network for verifying ZK low cost low cost commodity

Hardware that's extremely decentralized where you as a devel uh you as a user of the network you you you pay a certain amount of gas for verification depending on how many nodes verify and come to consensus and then relay you're already verified data payload to the chain that needs it like a Celestia for ZK prove it for ZK verify prove in your own network prove wherever you want hell prove on your laptop but you won't be able to afford to verify on ethereum so you'll verify on essentially a side chain for verification that's purpose designed essentially an oracle Network that's purposely designed for

lowcost verification that's driven by consensus yeah I guess so I I think like that is like structurally sound I think the the realization that I've come to is that like I don't think that like decentralization is like that viable outside of ethereum main net I just like as someone who is a node operator like oh my God is it a pain in the ass like I was on vacation for three weeks like five days ago my node went down and then I just like had to sit there and panic for five days right and I think I think we can hope and expect that like we have one network that's like 10 or 100 thousand people but like I do get

like a little bit concerned when like we start to build like more and more systems that are surrounding ethereum I I I agree I know I know what you mean it gives me a little heartburn yeah and so I guess like you know maybe we that that just is how it works or maybe like we figure out how to really use ZK to make it all like economic or or something or or or maybe it's just like I I don't know maybe things are changing so fast that there's going to be a new solution but I guess with our final few minutes here I would love to talk to a little bit about like how do you think about um the the decentralized network that you want to build with space and time and like specifically I kind of just hinted at it maybe breaking a

little Alpha early on this podcast but that's kind of that's that's kind of it like decentralized network of space and time is this set of nodes that can run a sing it's just a a a rust binary that can be run on a single server or kubernetes cluster on what you have is available as a node operator and this server is uh a network of database servers that also come to consensus and generate ZK proofs it's servers that do proving and you know generating queries and proofs then talking to each other over consensus the the talking with each other over consensus

facilitates using those same servers as a network essentially an oracle for verification as well so I know this is kind of wild but like one single rust binary is a SQL database of ZK prover a consensus Network like has consensus and um like is essentially a a verifier Oracle right it's it's a network of servers that can verify proofs and come to consensus that they each redundantly verified the pro so you could have one node proving and the other 50 or 100 or 500 verifying or SCS you could have one Noe you have maybe there's 500

validators in the space and time Network and one of those nodes proves you know per query not one one total for the whole network for each query because it gets routed to one node to prove and then A Shard of 30 of them verify and come census and then relay the verification to ethereum now that's not as sexy as verifying on ethereum but this is a response to you saying hey Shram mentioned that when we have when we we run out of space What do we do well you can offload a space in time space and time is all about offloading baby offload your storage offload your compute offload your data access and now finally offload the verification of your

proofs to a lowcost service that's cryptographically secure I'm really glad you continue talking about that because that like specific Insight that the space enti or the the proposed Space and Time network um is not really about like just creating another blockchain Style Network where every node is just replicating the same computation and then coming to consensus but it's really leveraging the ZK succinct prover uh verification Paradigm like makes so much sense and I guess um I that I don't really think that's Alpha I think I learned like that you guys are doing a a decentralized network uh just reading through your docs but like I would love to just fish for a little

Alpha and like really help understand like how are you going to get like a bunch of like people like me or or node operators to um get involved in so I guess I'll just ask you like the direct question like do you anticipate some sort of like token based system or or how are you thinking about creating a decentralized network I'll ask you the direct question wi token yeah yeah okay that's the direct question yeah hoping hoping to bring it into 2024 I'll like we we got more work to do we're not quite ready we have arguably nine months of engineering before we're ready to drop a token and I'll admit like in nine months it's gonna be if we do drop a token in

nine months if that's going to be a crazy very aggressive nine months so my my guess it'll take much longer but uh all I can say is this aspirationally I'm hoping to bring a space and time token into 2024 or else I'm probably going to miss this bull run yeah no yeah fair enough uh yeah at the end of the day like a lot of work to do we got a we have a message to tell right we have a story to tell we have a main net to launch and that's a lot of work to do in in less than a year yeah well I guess as a closing thought I mean I I hear you um and I think we opened this conversation by saying that like the the biggest problem that space and time has isn't really even an engineering or a company

or a building problem it's a communication problem and I think that you have any recommendations to solve it after hearing me I think like look I think like I I can give you recommendations and like you're smarter than me and and you've got a whole team behind you like I'm sure you've thought of it but I I just a comment of Hope is that like two years ago literally the only way anyone could conceptualize ZK was like oh we used to have plasma and that wasn't really working and so like ZK is going to be the answer to make this like other esoteric thing work like that's all ZK was and like so far like we we've come so far from that and we're

already talking about how ZK is like a paradig Paradigm changing technology that is about offloading like creating modularity within blockchains and so I I'm just so confident in all these conversations I'm having on this podcast that like the story is being written right in front of us and like yes we all have work to do um like we're all contributing to it in our own way but I just like am pretty confident that over the next year the the problem of like it's very hard to communicate what we're doing is going to be start to be solved by just like the community actually building the things that we've been uh just like theorizing in our head and and

whether it's space and time or succinct or modulus or like all of these companies that we keep talking about like like the the ecosystem's coming together and the narrative is coming together and like none of us are having to work as hard as we did even a year ago to explain like what is even the purpose of this maybe the maybe the community maybe the ecosystem will sort of Define half our messaging for us well I think I think that is like the law of crypto and like let me put it this way like back when like compound first you know compound a like how much work was needed to explain like decentralized finance and like these Primitives around like um like the the one that's coming to my head is VE tokens like how many

hundreds of thousands of threads were explaining like what like locking and four-year and emissions and all this voting stuff was and now like it it is part of the expected like if you don't know what that is that is like a signal that like you're not really in crypto yet and so like to think that that's not going to happen and like that that you know SRAM and Igan lay was like our Unis swap moment in ZK is like I think just like a misunderstanding of how this industry works well set well hey it's been a blast yeah I know this has been awesome and before I let you go can you just give the audience uh like just let them know how to find you how to find

space and time and like if they want to like learn more about what you're doing and like start to build applications that like plug into this like what should they do yeah okay um find us on find Space and Time on Twitter at space and time DB we have a pretty popping Discord as well if you want to get more involved there and like I have no doubt that a point system is on its way uh TVD um I'm at Chief build not not build build I don't even I don't even know how to pronounce my own Twitter little uh and uh come follow us we're always talking Ai and crypto always talking ZK and then

um pretty active on on Twitter and Discord would love to grow the community and love to to bring you in and if you guys want to learn more about uh what we're building tons of documentation tons of content on our website as well so we'll see you there awesome man well Scott I I uh I was telling you before like I'm so excited about what you're building and I'm so excited by like as we've been saying all projects that are taking this like literal magic of hiding numbers behind curves and then using that to create business applications and turning that into like real things that are changing the world I'm just so proud of you guys and impressed by you guys and can't wait to watch like especially the next year so um if you want to use the app it's app. spaceand time. aai if

you want to see this next gen user experience we've been talking about for the last hour and hour or so so go to appspace and time. aai connect your wallet write your first AI query awesome well perfect all the links will be in the show notes and um again just Scott thank you so much and uh Happy New yeary New Year talk soon my friend