Devconnect Convos: Stay Early to Onchain Trends with Flipside's Data Analysis Tech

Transcript

welcome back to the rollup High Roller Community as always nothing on the rollup is to be ConEd as Financial advice we are just an Educators in the defi space toss us a quick like give us a share a repost send it to a friend who may want to watch this and finally give us a sub or a follow it means the world to help support credibly neutral educational content in def5 we're here to educate Empower and enrich you and let's jump in dude welcome to the rollup thank you for having me uh so you were you were telling us a little bit about flip side um the the uh platform has 10,000 users 880,000 users I think you said yeah we've had like 150,000 signups but in practice that's more like 80,000

who actually run queries okay um and essentially uh users run queries um they query blockchain data with SQL yep um and they use SQL to create dashboards um about about blockchain um about all kinds of things right applications chains protocols yes so I mean most popularly you'll see like two different groupings typically it's like blockchain stats themselves how many transactions blocks throughput transaction failure and then on the other side you'll get like protocol specific so for salana data I want to know about Jupiter exchange for example um so protocols versus the chain themselves are the two

generic groupings for dashboards and and you were talking about essentially the business model of flip side which is not necessarily profiting off of data queries or dashboards because data is cheap but instead helping blockchains grow and now starting to focus more on the evm space yeah so our mission is to help blockchains grow the data is a core part of that you know in 2017 we started as a fund it was hard to be a fund in 2017 because all the coins were you didn't know anything different between them so we realized oh we had to actually like have data about the coins and the chains so you can actually differentiate between them we got into the data space and we've been there for you know six years since about to be seven um and yeah the business model for

us has always been that data is cheap and data is really integral to the overall crypto ecosystem to grow so we really don't want to charge for it um so what we do is we have the studio which is a SQL editor and the engine for dashboards that's free the goal is to keep that free forever and then if you want professional Enterprise you know programmatic access to data we have that business offering for Enterprise levels who need all the data replicated professionally blah blah that fancy stuff and the main business line which is helping blockchain grow or we try to be full service so starting at zero we mean like we're there for you in the test Nets we're running nodes once you're at a test net we're talking about running nodes for staking ideally

getting a delegation of stakes so that we can make blocks and then once we're getting revenue from making blocks and being involved in the chain one we're AC crewing tokens in the gas token so we're the incentives are aligned we want growth so where the tokens we get grow and then we have the tokens to do actual programs so whether it be you know top of funnel impression programs like let's make really popular dashboards let's get people really deep in the data doing marketing about data uh talking about what's good and what's bad and why and then also like you know hackathons new project developments on chains um that's the bread and butter for us is because it really aligns incentives so data is like it's important but we're really trying to sell growth makes sense and do you find that

depending on the blockchain they may have different different priorities maybe there's a different target audience and then they oh yeah they tend to approach growth in a different way yeah I mean we we don't publish the four phases that we think about but we generally think about it in four phases um and the reason about is because it differs in every ecosystem but I mean yeah they're all layer ones right whether it's apos sweet salana Avalanche near whatever or maybe they're a really big protocol like Bridges um like XLR something whatever it is they're going to have different kpis key performance indicators they're going to have different goals different reasonable targets you know oh we want a million users every day hey good luck there's three chains that have that

number let's get a reasonable number let's also talk about the quality differences so a lot of it for us is like not just helping them set up kpis that are useful and relevant for them but also like what's unique to them what's their Edge and how do they maximize that edge so a lot of my work on data science is research optimization what works what doesn't work um and really treating these things as like unique places to do things it's not it's not all def5 all nfts like these things have different use cases for different groups uh we want to like maximize those differences and you said there's four uh ways or approaches or kpis oh like phases so like you know phase zero is like okay you actually have to be up and running test net you need the test net to work

if you're proof a stake you need the stake to work if you're a delegate to proof a stake you need the delegation to work all that core foundational stuff uh you know we do that we do test Nets and things like that we we want block games to launch and we want it to launch well we want to be one of the early people making blocks and being good at it because we want those ense of the line um after that you know it's getting developers getting growth measuring growth a lot of people love GitHub commit measurements it's not my favorite metric I look at contracts deployed or programs deployed um because those cost money and when people are willing to pay money to deploy a contract you know they're serious um my background is economic so I care a lot about that economic value of the settlement things

like that um once you have contracts and protocols you know you not just want to have them grow and have user growth you also need to govern them um you know we have an ecosystem team that's deep in governance across all the ecosystems you've probably seen proposals that we've made um some pass some don't you we're deep in like everywhere that we can be on like expertise that we have on tokenomics designs um not really airdrop designs but we have helped air drops like when they've broken and we've come and fixed them uh then phase four is you know just like monitoring and continual growth and measurement and keeping things on track um similar to the data science pyramid which we're not going to talk about right now but so sorry data science pyramid is just the foundation

is like what happened like how many then you go up a level and it's like okay well how often or like is it predictable give me like the average oh you go level up okay now forecast it give me prediction and top of pyramid where you get into like the artificial intelligence machine learning that the idea there is not that you just have ai the idea is you have an expectation that's driven by data and deviations from that expectation are signal and that's the top of the pyramid when deviations from expectations are telling you something whether it be like oh crap like the last 180 days me's daily active user count has gone up like 15x in the last 180 days they're just they're just hitting a product Market fit in an

emerging market with an um so when you have a deviation from expectation it's telling you something you're not going to be able to get there if you're counting the number of transactions per day like those are two different levels of data and those are two different levels of kpis two different levels of maturity um so that's why we try to think of B in phases as we help blockchains get to that phase eventually and ultimately like th those four phases are each means to an end the end being product product Market fit yeah I mean I mean so chains have their product Market fit and then protocols on those chains have their own product Market fit and sometimes they conflict a lot of my researchers on Unis swap for example you do the math right Unis swap token holders haven't gotten a

lot of Revenue I don't think they've gotten much of any they'll let LPS get revenue from fees but really the big winner is eth the layer one which gets the transaction fees so you know when people make jokes about like oh what eventually they'll be their own chain they'll be an app chain we're going to have an app chain future they're hitting at this reality that like the goals of the chain might be different than the goals of the protocol and they can conflict and that causes Downstream questions that you have to answer so we're we're not trying to pick winners we're trying to figure out how everyone can just get to that future State on a timeline that makes sense for them yeah and also also helping projects construct and think think through the future State

because it's going to be different depending on each project you kind of help like present the questions and the data to help them measure like what their ideal state is in terms of like math and numbers and then also uh the framework to help them get there yeah and I mean a lot of the time and I don't say there in a negative way but like when you have a new chain or something and it's a small group and they get funded right like it's hard to build that team and let alone building that team that knows how to do product and data and engineering and devel let alone having that skill set and also the crypto skill set so for us because we're so into the data we try really hard to grow that analyst community that I

mentioned into being so crypto Savvy that they can be planted at these foundations and be useful a lot of the time again nothing negative I don't see it that often but you hear about projects or chains and they have a lot of like oh xbcg X McKenzie X consulting which is very powerful but they're powerful in mature firms and we're in like a startup world where like you might have the lead of data somewhere who isn't super deep into crypto and that's okay we need to be that crypto expertise for them and in the best case scenarios we do get to plant people from our community into being the leads of data at these foundations um that's the full cycle if we can get there that's like the dream

scenario yeah have you been able to hit that full cycle with any of the projects that you've worked with yeah I mean we have ambassadors that's like a separate program where like people from our community who just specialize in a pafic chain they literally just get paid to study that chain and their their entire goal is to do topa funnel so make dashboards that are about current events or make dashboards that are canonical like this is what you this this is the dashboard for that protocol this dashboard is linked on their website it's what they consider the source of Truth um and in the best case scenarios yeah we have had people with Mark Community to rise the ranks and they end up in the foundations in very lead data rules where they almost become our Customer because they went through us and we just have that depth of

relationship um so yeah that's a dream scenario we've had it a few times not as much as we like obviously want it everywhere yeah and and then what what um what what is the thing if if you can point to like one or maybe a couple of things that you've seen that really move the needle for these chains obviously a lot of them are different there's nuances in their goals and in their kpis and how they achieve those is there a common pattern that you've seen that if a chain will do this collection of things they tend to achieve their goals yeah so at a high level um there's kind of two distinct path there's one there's the data side of curation every chain makes data that's not enough um and for two reasons and I'm going to use a theor of the example when a contract emits an

event it's telling you whatever it wants to tell you it doesn't have to tell you the truth and we do see it we do see contracts that because of errors or because of just weird ways of implementing a certain function they deviate from a standard you can check the state difference of the trace and you find oh like they say this happened but when you actually check the blockchain that didn't happen um and with that when I say that what I'm saying is like you have to curate the truth not just curating like the results or statements or log out the blockchain so for us if you go to our studio for example you'll see you know something like 20 different chains every single one you'll see almost the exact same schema core this is like the actual event the facts the transaction hash the

boring nitty-gritty ingredient okay now defi we've curated the swaps for you um nfts we've curated the mints the sales the transfers uh governance okay we've curated the delegations the staking uh and price data okay we've integrated like synthesized prices from a variety of price feeds so you can have token prices those only one of those things that I said is like from the chain directly and and it's the bottom layer of the def of the data science pyramid yeah the facts like the core like contract said this and that's all you get um and some of the time it's like you can't read it it's like 0x00 said 0399 two contract z x like like a human can't read that so the work that we do

to curate I mean labeling we have a billion labels or something like that across all the chains like so you know hey this contract is a pool it's a Unis swap pool it's a this Unis swap pool you has this fee like all those numbers have to be like brought in so that you have a nice clean row so that's half of the work is just that half the work is can you make the data about your chain and protocol usable or regular people the second half is like okay well now your data is clean you need to have people paying attention like what do people care about they care about what's popular what's growing what's fast so you know I'm going to shout out near for a second n's worked really hard on something called the BOS the blockchain operating system and it's their

mechanism to say Hey you come to our site and you want to know what's happening do built on like flips side data and I think other data as well like this is what's happening these are the contracts that are most popular this is what people are doing this is how many they're doing this is like links to the top projects um and having these stats available where the end user is not in a dashboard they you have to go five click from Twitter to our dashboard site then you click the link from our every click is 50% of users lck so putting that data and that knowledge where users are is the second part of the puzzle um and yeah and the best case scenarios having like really clean curated data has meant a world of difference for people who are like a

small decks who want to convince professional market makers to LP there hey like you you know if you're a market maker with like 100 million in assets under management and only 20 million is allowed to be in crypto you're only going to pick so many dexes and you have to make sure any tokens you threade you're hedging on coinbase or bance or whatever so like you're you're very limited on where you're going to play in defi if you're playing in defi at all so for the protocols trying to capture like the attention of these whales you think the whale is going to go click a random dashboard on flip side no way like maybe they'll go to like a website but you got to you have to have it ready you have to have it ready in your pitches and your email and your marketing uh and it's

critical for them it's critical for them to have it curated clean easy refreshing automatically so they can use it um it we've seen some success and your mental model for this data science structure and the mental model for flipsides B business structure is very similar so I find like particularly cool because it's a common mental model that you apply to both schema on the the technical and then also the business is there a general mental model that you have that kind of underpins both of those uh I think that at a high level it's just that everything's a spectrum like I really don't really believe in like binary stuff like some people say oh this is popular it's not popular oh that token's going to go up it's not going to

go up and like I find that if you're more tolerant of like the spectrum of potential results you can have a more fine-tuned idea of like okay like this is not great but it's on track or this is bad and it's horribly off track and like having like going from binary to like even just four potential States just makes like it makes planning a lot easier it brings you closer to reality of like what might happen um and it just makes easier to game plan like for us like you know if a chain goes down or if like one of our infrastructure things go down like we are sourcing from so many places that we've already had those

backup plans so we're not going to be in a scenario where like oh yeah that's just broken it's like okay let's this piece of that is broken the price feeds are broken but the token transfers are not broken so like you know we don't want we don't want to be one or zero and that's the main model is like TR to out four plus scenarios for anything very cool it uh it involves uh redundancy liveness and it applies a lot of these things that we hear in these panels and these talks that apply to blockchain networks to a business and a business Network which is a network of people and there's still something at stake whether it's time or it's money that's flowing back and forth yeah I mean in terms of business liveness like I think that is a very transferable concept um you know my

background before crypto was monitoring like Ebola in like East Africa for um you know Public Health clients uh and also like childhood blood Le poisoning and in that world data is very very small so like we had very difficult situation with liveness because for us it was like you don't find out about a situation until after a tapen months after it's tapen um and that affects the entire business because like well how do you stop an outbreak if it's you know three months later and he just now found out I don't want to bring up covid but like that's a as a more common example people have but I saw it from lead and Ebola and raidon and other problems so I mean yeah liveness censorship resistance comes up I mean who's allowed to submit data to

governments about heal um why only them what are the minimum requirements to even submit that data is it better to have partial than none um sometimes the answer is no sometimes partials like actually messes up your conception um so yeah I mean a lot of these Concepts fit in a lot of different places um that's my Consulting side coming out of trying to reuse a word in a different context back spot yeah that's cool no I mean this uh this is interesting particularly to me because I was speaking to this guy from Cy's Ai and they do on Jing monitoring basically they're able to or they claim that they're able to uh detect malicious events uh way before a protocol came

like at the at like the state level and then also uh there's other companies that are out there which um actually uh use m in flash loans to shr run malicious hackers hack the protocol themselves as a white hat and then offer the funds back for a small commission and explain the I that I've SE that happened actually I've live uh it wasn't I mean I've seen a lot of pet Shield alerts but there that's not what they do I forgot who did it but yeah um the way you do that right in the data pipeline is everything I said about data it also exists in the m poool so before the transactions confirmed when it's just

ready to be ready to be planted you can monitor them mle because nodes are talking to each other when they say hey I have a transaction here it is note that this transaction is available if you're one of the nodes or you're listening to that gossip you can see these features when I say feature I mean in the data sense you got imagine a column in a table is it a flash loan yes is the address new yes is this address under by tornado in the last 48 hours yes okay well you're starting to have a lot of yeses in sketchy columns now now this gives you an opportunity to be worried and then what you do next is okay what transaction did they submit who did they submit it to okay they submitted it to flash Bots we're you know they're friendly they're not going

to like they're not going to necessarily sendor transaction they're not going to hide a transaction unless he do their private relay so that's when you get an opportunity to say okay they submitted what is clearly a hack at this gas I'm gonna 10x that gas and send it to flash Bots first and Flash Bots will put mine first so I get to front run the hack um and some of that's not off tank like it's not like they like have this agreement to do it it's because like the mechanisms are available to do that um and the reason I bring that up is like it's possible it's possible to do all of this very Advanced machine learning and forecasting in meem pools um and even if you do before then which is where I think the government's trying to go with like pushing wallets and RPC nodes rbcn

points or whatever to like know their customer quote unquote is that if IP address from North Korea he wouldn't have this if an IP address from North Korea sends a sketchy transaction you know it should be blocked not because of a white hat it should just not make it to the nodes at all or the node shouldn't propagate it that's what the government wants to see if they knew the technology well enough that's the kind of stuff they'd want to see it's not real yeah and I think on the first point also the incentive mechanism for business model which just kind of shocking to me was that they would threat run these hacks which is OB obviously saving the protocol a ton of money and then they'd basically hold the funds Ransom and negotiate some sort of commission with with the not like I mean

perhaps it's predetermined right but there's very little like upfront payment it's like yo if something happens we're going to front run it and then we'll get XY and Z out but anyways kind of moving towards uh flip side with with this and kind of summoning this all up are you guys using AI or kind of like machine learning within your data structures and if so oh so I mean right now it's just pipelining and testing so like you don't really need it for something like is the data missing right like if you don't have data you don't have it you don't like in other business context you do something called imputation we like oh like hey I don't have the number of sales for this product that day but I need a number of some kind for my master to work get the average of the last

seven or do some fancy forecasting to info that number that's like the classic use one of the classic cases of AI and machine learning is like fixing data errors with like imputation some data scientists hate that but whatever sometimes it's actually really good and useful in our case like you won't see us do that because we're Le looking for prices and stuff from sources um where we see AI as useful um in our business specifically you know you use AI all the time and you don't even notice it that's what we're trying to fit in so like if you're writing a squel career and it breaks um Dune has this shout out to them Dune wand and their Dune AI you know if your query breaks have a button that says fix query that uses AI to read your code read the

data tables you're trying to parse and make an educated guess on why your code didn't break and does something new that kind of text based natural language fixing with AI very plausible future for a lot of ter protocols um not just flip side not just do uh other AI use cases like we said monitoring that's not super in our wheelhouse the things that I see as useful for AI are really the educational side of like great you have data uh do you know how to use it because what's going to happen and you already see this people say like oh look we have ai come talk to us use natural language what's the number of users on this chain on this day you don't really need AI for that you could just like type you almost

write that SQL directly yourself but there is a world where people want AI to translate their ideas into like computer code that is an extremely large gap that it or anything that's not a toy example is extremely difficult so what we're trying to do is in AI for things like search um you have an idea you want to you want have an idea for a dashboard or you have a question we don't want to just make up code that might not fit your use case we want to link you to analysts who've done that research or have done similar research or'll link you to dashboards that answer your question partially so that we're not using AI to make you do more work duplicative work it's to help you discover kind of like Google right you're just discovering things that are

relevant to you you so that's what we think of AI mostly is just like how do we get people maximizing the resources that already exist as opposed to using AI to duplicate work yeah sweet um gonna kind of wrap up here but what's the most dgen use case for uh me as a retail power user of flip slide how can I maximize my DJing onchain with so I have a Flipside Pro account which I use which lets me bring like a bunch of data out of the system and build my own API and R to do uh forecasting for UNIS swap so pretty much every day every 24 hours I get like recent volume wied average pricing across a bunch of Unis swap pools I do

forecasts on them I test those forecasts and the output is a recommended high price low price and allocation between like eth and Bitcoin so I just like funnel in the data extract it and on a schedule it tells me hey we think the not it's not really a forecast of like the price it's a forecast of the highest profit range possible because in unisoft V3 the tighter the range the more fees you get so that's where like some not AI but that's where some of like the really fancy Maps comes in up using our data integrating into another system and getting like a really easy number so for me I actually used it for Harry Potter Obama token sorry for sure think it's go to a dollar uh do do you ever use gelato

to automate smart contracts because if you have that you could like somehow automate the smart contracts based on that that data set every 24 hours on L2 you just have like a a bot working for you basically I've been thinking a lot about L2 Arbitrage but not Arbitrage because it's not it doesn't have a typically Arbitrage is a zero execution risk it's Atomic like Buy Here Sell here or don't do anything I've been thinking more about like okay if these prices are deviating on different l2s because they have different liquidity and I know that they mean revert in this way I should just just be buying on op selling on ARB and being okay with like the risk of holding ethon accident like there's there are people doing it but because people are thinking in a traditional

Finance mindset of like don't have exposure risks to the underlying asset if you're not scared of that there's like a lot you can do to your point um definitely not scared of that yeah I'm not scared of holding did I answer your question I forgot your question I use zelato yeah I have it's all good I launched a project on eth M net that uses gelato to automate the uh curve rewards for example it's a pleasure to be in your company Mr B big brain I appreciate your time and uh shout out to Dev connect shout out to oap and Uma for gast's governance shout out to flip side am I'll flip this side now and a second shout out to Uma for across also because

across is like this shouting out a data thing I found out recently there's no bridge that has as much volume per tvl as a cross and it's like not close like go check out that stat absolutely based if you want to find more stats go to flip side and get your flip side on thanks all bye bye