Hashing It Out
Hashing It Out

Episode 61 · 1 year ago

Hashing It Out #61 – Solana – Anatoly Yakovenko

ABOUT THIS EPISODE

In today’s episode, Corey and Collin talk with Anatoly Yakovenko, the Co-Founder and CEO of Solana. Solana is a blockchain platform attempting to scale transactions without the need of sharding and we want to know how they plan to do it. Join us as we dive into their strategy, their current progress, and where they plan to go. We ask the right questions to get you to the right answers on how this technology works, so come listen and enjoy!

Links:
solana website
Anatoly’s Twitter

Donate to Hashing It Out!

Wow entering a CNCAST at work, welcome to hashing it out APOTCASS forrethought to the ATTECH intevators behind blockd in intrastructure anddesentralized networks. We dive into the wee foget at whyand how peoplebuild this technology the problems they face along the way I'm listening andlearn from the best in the business. You can joind Orreck, O gus mote back nation out as alwayshung your hose tarted cory, petty with me, my Tressico Hose Call Cochet, soit'sot everybody calling what's up everybody calling wo gats ame energy.Today Yeah Damrata got some spunk dod, I drink my coffee, it's not Tan seveno'clock in the morning, Ci you're not tey're, not drin dude. I am dreined.Actually I've been working really hard. Like I got four hours sleep the nightbefore and last night I got maybe five or six so yeah. It's a crunch time justpopped up on watch of CAFFIE IN ECENT. I'm just excited O at hi cast man I'Iho in Tilli has to say that speaking O, which we had today. We have anatolecofounder CEO Selana Um. What do we do? The normal kick off and that youintroduce yourself how'd, you get started into the block chain base andthen wove from there we'll move into kind of what Salona is and whyit needs.EIST ES also to be here so yeah. This is Annatoli, go CONDERSOSALOTA! and U how I got started in this mess. I Guess Oun twenty seventeen. I had too muchpoppy in herself tefore in the morning and H. I had this like fever. Dream ofthe encoding passage of time is data socreating a the datastructure, the represents time passing skind of aweird thing, if you think about it, m like in Metaphysical Ramification Indi, but what Chool Otis it's no time is afoundational component to distribed assistance and at the time the coin was like at seventy dollarsper transaction or something like that and everybody was talking about scaling.BOCHIN UM AT II had spend most of my career working, a Qualka which is a winrless semiconductor firms, so well aware of how wireless networksYoure sell, Yor thatwork scale to you, know millions o participants and oncean a source of time I was really Gona. I knew that I had something that couldactually sell all these sdaling problems and that's really what kicksstarted the project so with like what was your background in like thelock chain space pride to that, or did you just like, have a knowledge aboutit and have this sudden insight and decided to t jump into Um doingspecifically blockchainwork? So I mean it was a engineer like most of my lifebasically but spent most ofmy cuall amy like worked in operating systems. Someyou know wirdless Gals, a bunch of Riggo things and I bickle an can. Icame out. I was like well aware of it and theere.You Know Trat to Ge Miting r kind of thinking what if we wrote a keep Eyuteexpesion of this, we can get all the Hash power, but I wasn't really likeserious about it. It was Kindo like this is a neat thing and it's neat fromthe sense that, like we have this permissionless open way to synchronize information, but Ididn't really think about the ramifications of what it was. You KNOandgineers often missed a social kind of phenomenons twey. Think about theteck from the sex perspective Um, so theals. Remember that Heriamaso andkind of thinking out this is like sadly, jealouscripts for this really coolthing. They could have used the Mi Tet, avirtual machine, a better language, Aga totally missed the whole. You knowSocia Liv back to O and a like moltenna revolutionaryaccomplishment of as to get that working so yeah, just just when I know a lot of peopleor KINDOF like Misso. Sorry about that. Ah It's hard to Sigrei talking whenyou're yea can't see each other but um you. I know I told of people t actuallycame in for the social reasons that a ton of people who actually came in justfor the purely technical reasons and m. If you're like looking for the perfectscalable, you know fully decentralized system like dicquining it therymeaningso yeah. I can understand why that'd be kind of a turnoff if you're highlytechnical and just interested in how transactional throughput you know. Ifyou can meet the demands of of the users of the system. You know I kind othink Yeahye yeah. I think this whole...

...idea, Thati Cubeas, store value, justwasn't even o my rator yet either and that's something that is almost obviousand retrospent to he ha um and that's like an interesting thing that I missedcompletely so around twenty seventeen. I was in San Francisco an this bigACIOBOB happened and H, like literally I Hav too much coffee on to him. I wasup before and the skin of this revelation homan. I can add a source oftime in the snotering that a trustless source of time right to withoutcheating, it's a source of time that doesn't rely on an external clock. It'spurely the scrypographic process that gederates an data and they scoured theInternet for anyone else. Working on this and the couldn't really findanyone an ou now to commit convincing myfamily that that wasn't crazy. I clik my job an like kind of startedthis project, so you actually got past the point of convincing your family.You are crazy: that' at's, a Gen that alllos itis I's a feet of a Coptre itit took about a month, my wife's, an engineers. So it took me like Okayyou,got to listen to EA SS, like I think this'll work, an O. I o Man Thata,let's get down, do that t seems to be a t, different trader in what you'retrying to build versus what a lot of other people are doing. You said thatdatastructure O that I guess encodes the passage of time. Can you elaborate on that a little bitmore yeah, so m? It's a lot of fogs have actually started working and havebeen working in this. Even at the time when I figured tha, you know heap thisrevelation and it's called a verifible delayfunction before that there wasn't a lot ofpapers published on it, so hiy couldn't really find anything, but at twentyeighteen, a lot more papers, stare getting published and what we'reactually doing is Wao't call like a foorman verifiable delay function. So the technical term is Yoave a some some mathematical process thattakes certamont some puzzle. Kenlike a pefor puzzle. TAT takes cerain amountof time to solve, but the proof is much much faster and we have the same property but um.The proof takes the same computational amount of power, so if that make senseof what what we're doing is um we're using this, you know shouto fifty sixfunction, the same function that Vickine uses they were running in in aloop. So output is the next input and you run this thing in a single core asfast as you can, and because it's this recursive, I you knowlup, you can'tpartaalize that process right. So if I told you here's a bunch of tamples ofthis thing running for an hour, you know that I spent somewhere someamount of time. Doing t is- and maybe, if I had a slightly faster single cordmay take like in ow forty five minutes or SLEP, it slow run an hour and a half,but you know that actually spend real time to do it, no matter how much moneyI have, because I can't you know, go by a million Mares and making a milliontimes faster. Cou. Only really, you know super cool, Ell Lik get a reallyreally fast day, sick from Hehas, some s and super coolet. The mittle stillgave me about. Maybe two three x peet up so because it's found by the physicallimits and electrons passing through this slic single tredit, you knowsingle circuit, but what we're doing to verify I os hetake all the samples and an a modern TPO card and idy has been like supersuccessful, a the Scali Mas things you have about four thousand course, O yousample this four thousand times a second. You can verify a second and aquarter mill, a second so for for all practical purposes. We have a kind of ayou know, practical Vedia, which is very secure because tes space on chater,fifty six and very cheap to bear it by. If you have a modern M processoravailable. What do you mean by sample on the hoc? I don't think letme make a key different shade tohere in terms of sure, like when people think a proof of work in the time youassociated with blocktimes. That is a statistical average of what the amountof computation it should take to find something within a random sampling.Rightand s like non average, a bunch across a bunch of different, triesyou'll you'll converge to a given amount of time. So, like N, whathappens with prove of work is that you optimize the difficulty to go to acertain time of it. So every seveny blocks that coin will rea readjust thethe difficulty to just for ten minutes of computation time, based on how longit took to self blocks. But in reality you could have you have kind of theirAsia Ness, so you could end up solving a block way way away less than tenminutes if you're lucky and that's...

...normal, that's just a part of thedistribution. What you're talking about here is not that it's not a statisticalum thing. It is literally it takes this many times to run shaw given tos muchCPU power, and since we have a pretty good idea on what single Corpeprocessive power could be youhave. A pretty good idea on how muchcomputitional work and time is going Otack to do it. That's right! Ye Ut! I still don't understand when Y. U,when you brought up the part about Um, you know the single like linear progression ofthe shatwofifty six. Over and again, we have an idea of how long that'll take Iget. What I didn't understand is apparel. You know like we have GPUcores and there's like four thousand cors OA, whatever they are called andwe can run like, I guess, running, that same function. INPARALLEL. Is thatcorrect? So what is the like? Why? What does the word sample mean there like?What is the Advanagh? I knowing it that way? So it's it's because shouts offifty six is Prema resistent r. You Ave no way to to predict the output. That'swhy we can do the Sigal thre process and guarantee the Tima spent, but thatalso means that there's no way for us to verify it any faster than by justrunning it again. Um to imagine you have a single core right single circuitrunning as fast as I can, and you simply record the number of times thathas elapsed like it's run, the circuit a million times to million times andthe current state. So these samples, you record, is data.So now the STATA structure right would say: exampled, you know a thousandtimes a second. You can take the start, an end of every sample and run it inparallel on a different core, of course, finish ight and there's no errors. Youhave guaranteed that that takes your check, pointing every as you runthrough this this circuit, and then I at is checkpoint. You have a start, Gint finish and you can enparalyze that process as you go through so so folks,a vediat research that wor lk dampany and a bunch oter researchers in thosare building much more sophisticated approaches that use you know liketwenty forty eight bit square modula Um to to to co, to build a mathematicalfunction that takes real type to chenerate, but faster than Polylogutime to verify so the speed ondoesn't. The verification doesn't require thisGPU neralization, but those all those constructs right now have some very funky cryptographic. Trade oftsthat are hard to deal with. Also the hardwors, an readyyet and Um from our perspective were not religiousabout our approach. You know as soon as that, harboris ready, iwill switch it but shut O. Fifty six andkind of me isan engineer. That's not a you know, VEDIAF researcher. I can I understandthis an it's very secure and it's very easy to work with, because until tha,the indeed both ship shouted Te, six specific instructions, havtoboth can doa single round of on one point: Seven five cycles, probably the most optimatesmonction in in the world right now things to bi. We have. We have a prettygood idea how fast Thas in can go and therefore, for somebody to try to attack thenetwork would require much greater. You know investment and resources andsomething new well to understand what those attacks would look like I'd liketo understand a little more about how it forms consensus using this clock. Soyou have this. What isessentially and I don't even wan T it's kind of weirdlike I assume every noses running this on their GPU, so they're all runningthes same thing, but they're, not necessarily at the same time, in otherwords, you're, not running the exact same thing. At these same time,soskindof like th, Eis, clock property. Is that kind of a weird thing to say,or does that make sense and n? So how do you use that to form consensus? So will we do with this thing? Is a ISO that every every noliday in e hnetwork runs a single CPU cord doing this process Um and this Shouto ift six process? What'sinteresting about it, as if y take one of these samples and a write, a message:Riht pick a message: You know I use that sample in the message, ats, just adata and assigned that guarantee that that message was created after that Umsample was generated because shouted fivty six. I can't guess those numbersahead of time, riht its premage resistent as a script, strong,tripographic property that the numbers are unpredictable. So if, if I reference it, it's AIPITICThe neuope times picture myself with it, and now everybody knows that I wasalive after that New York Times of Polish, I think, have to keepline a long recordof these AE. You only have to keep recond o the Checkinglin right. I wanna.I want Ta want to try and rephrase what...

...you just said: The way that I'mpicturing in my head, um, it's everyone who's participating inthe network who wants to like liaspalidators, we'll receive a messageand they're running they're, basically, all running the same veedia they're allrunning a single CPU cycle of Shazi, forty six, that's they're getting they're getting hashesout overtime and when they received something theed like to vbalidate it,then, in reference, the current Hash, which will always be a part of whatever gets submitted tolike whatever cononical ized blockchain people t er referencing later on downthe line SOS the part that iffrent reference one ofthese hashes that guarantees. If the messages created after that makes usright, H, OASO, imagine I take a bunch of these messages and Hashd them andthen a pund, the hash into this process and every cord that account ten million.I had all these messages and I inserted them n my single thread single CPUthing just by pending it to the curent state and theyr recorded that thatthing Thano, Thi Shouto. Fifty six is now modified in this unpredictable wayand that guarantees thand all those messages were createde before thatmodification occurred. So yes, so th, so actually in the end of the day, whoactually gets to submit transactions to the Bloche is, is cause right now witha Proov of work, everyone's fighting to solve the giantswho do who puzzle andthen they get to submit things and then, whatever they end up submitting, is EPbecoming all the transactions that they validated. Is that the similarsituation, where everyone's vine to win a game and to winter that game thatgives you gives the rest of the community? What messages get includedin the next block so N, like so back in like the earlytwentieth century, people figured Oun, radiuns and they observed is that Hou and I are inthe same frequency we transmit. At the same time we get noise collision right.So what they did is they gave everybody a clock and they said you transmit Imen mint one enconmet anever like even minute and you've transmitted. I everyodd minute, and then we don't now politions right and nir participantscan basically grow to Habal. Our qloks can be synchronized m. So that's verythat's. Basically the foundation of Scaling Wireless Networksthats. Thefirst thing we do is we divide everything by time and now they do muchfancier things wer. They divite things FY fequency and its frequency hoppingthing, but basically, like you, have a common channel and you split it by Touc in your common channel Oinade, that's aproblem because you can only have, but so many likechunks of time available to so many people trying to broadcast messages. SoI'm trying to summit a transaction. How would I do that in your network? If theblocks of time are already kind of allocated? Do I have collisions? Isthat a problem amost understanding something so um? Basically Blockproducers are round robin and who gets to be theproducer as fast as we can and right now, that's four hundred Milla seconds.So ever every for, in a mill of seconds we have a new block so wheoure as as aclient of a network, pieceina transaction, it will get encoded in soblock. You know fairly quickly, basically not wik to live o okay, sothere's a list of of so basically the leader is given it's like N, afrequency, Spectiam kind of way of thinking about it. Like I do like this,this particular you know, Modulo of time is is, is is given tothis particular validator right I Tispendino A, and sothe people who are participating in this network m know who that is either in advance or by polling. Thenetwork CRRACKYEAHYE INA, no sae faces them. God tothis there a sacrifice for liveness,because that, like, if that note, goes down like what happens to the networkMSO, here's. The interesting thing is ah, because we have this media. What thenext note does is simply submit a Proov that they waited long enough to get totheir block right to their slop to their Hash Com, and even if the pregous note ice down,there isn't really this classical Timeo, where everybody waits for messages andomtimes amount, simply the whoever gess to submit their block first with theappropriate proof that they reach their e required height. That's the one thatgets process and because everybody staggered intolage right. Ifeverybody's up and alive, then everybody goes according to schedule,but as soon as somebody's down that next Valinatit, just you know, skipsskips up right right. So what a SOI...

...withholds their proof. I guess itwouldn't matter, because everybody in the network went have already advancedfar enough. But what? If that DU network delays or anything, does thatmatter at all like this person did not get some like all the messages thatwork, so it seems like you're, sacrificing safety, then instead, soI'm going to submit my message to to you know my peer and my Pere sees itand he's the current elected leader. Whatever he's going to be the onesmitting message to Decko out to all the peers or like how does that work DEy, bon top of this is like you know, single Shard en so no Shartingis arepaying even im a potgast cilled. No shorty is a really fast epeted statemichion. So every porn anil of seconds wo produce a new block. If that leader,Isn n there, the next one can simply providea proof that they waited to ge othe ment block and whatever Datad it arrives. First Um Valadid is vote on the NETMORKcontinuous movieg. So we don't sacrifice safety in the in like the long term M,but we have is asecret of safety, so as blocks are produced because they'repretty so quickly, individual uinners really no therustnotreceive them, so they vote, and then they observed the rest of the networkvoting and they continuously increase the commitment to safety and that growsexponentially, but at the Teno at the tip of the chain you have this Um,maybe lowest commitments to safety that quickly becomes exponentially a grostyFall II. I think I think the obvious question from thereis, who is the Valevator set and had? IsOne enter and leave it? So we are basically a proop of stakenetwork, so the valinators Ar gover has enough steak gets to be, you knowscheduled and a stake waited round robin. She simply writes sometransaction. That say me: I wouldn't be avalidator. Here's some stake, that'sSalegat ID to me and the next step up. He get skinchilled in and right now, they're. Actually inthis H, buth the network Tas, basically to go to Maret Launchbaye, so we wereworking with a bunch of folks from the Cosmos Communitie Ane folks, from likee es block producers. Um Cosmos feel like created this group of folks thatare more like the professional valodatorsa and it's been really amazing. Workingwith them and kindon boarding, all of them and were doing dry runseffectively like the do dry runs until we can't crush in that work and thatits may not right and yeah were wr we're following theirmodel Ho oamergion a gave Masake it's calle through Tho soul, know a bunch ofdefects or cycling. NERDS, an part of it is stress, testing the Nutmer I canChon Demonstrate thisis. This is the man that same Hardwai, people o Ramaa.You know totally permissionless tesentialized, but can do this. YouKnow Forhnda Millon second blocks, fifty thousand TPS s their target willsee if we hit it and that should be a fast lotched thatwill never have seventy dollars. FORTRAN's action, FIE bt. What can youdo on it? You have this. You have this link list of of as of right now ambiguous transactions.What what can you do with a transaction? Can you make spark M Raimar contract?Has the state grow, Um forrespect of that type of stuff yeh, so we're Um somy background is: like operatingsystem actually worked on the soascaled brew, which ran like every flitpone upthere hehad a CD may like Moterla Rasir at was a corconal engineer. The sameand a a bunch of the team is actually from that project and we're likeoperating system virtual machiners, so our smart contracts, language orbike od, is the pike of in the t part of the linix kernel. I think e R,almost a decagnoicalbkmpacket filter, its design for pipe performance packetfiltering, but not has not been used for ind, O more general purpose. Securelike a U so this thing in a single like single machine, can process sixtymillion package per second and a forty gic of it oor connection it. It's designed for really reallyreally fast, like processing rigt N, and this is running that there's even like implementationsof this running a Hartwar, so anything that you can confilethrough allbeon. We can execute and weare using rust and see as asprogramming languages like native ones,...

...but we just ported them move beum andin our vision, for this is that we can enable a bunch of irtui machines. Youknow like we're looking at adding a th with spoodnick BM, which is a reallynice clean, rust implementation of BM, but also stuff like sapling, I think itwould be really cool to to run ASE cash ucial machine alongside as well so fromour perspective, Ere cin o have this. You know really really fast facelayer,that does you know consensus and low atency blocks, and you can program itin this. Um might go that decide to be fast and what you ron on that bike. Oteis up to you and if you want to use HAGHR level language like Etab, whichwill you know, take more instruction sexecute than c Ar us, he can do so heaa the same. Look at you website.That says like basically like the way you do. This is latelyor. One is thiskind of stream of blocks don on the way O, which you just said with t averifiable delay function and we just build layers on top of that Sr.Decoupling, the transaction or th other virtual machines from the consensuslayers, a right Y and there's there's like I mean, there's, definitely tradeoff there right. So I think the folks that are working on making massively sharted systems areSallin a really difficult computer science problem and their goal is tohave you know the smallest amounto computer power to add resources intothis in a Messof NOWR. You know slic messure computers for us, we'redependent on these more professional validators to runvalidators that are Um, like you know, have more hartwor. So almost like youknow the way I think of it is like we have a you know, we're boot strapping,the next Internet with a bunch of homegrown nice speeds, some people thatkow like attack a little bit but n, or not about it and can Penago with thisthing at bee: local colication, spase and whok. U A bunch of Peepse forsignature, Var vagation, a bunch S S for Storang and make the thing reallythoughtful, but it's still totally open an permissionlist Tho people can'tenter you know at any time. So this is something that that kind ois a problem that I know of a Vidi Action. I really want to hear youcomment on it and I think that UST sounds like you're network. Just fromwhat you're telling me it sounds like it might still actually suffer fromthis, and it's Kindof like the keeping up with the Jones is problem AK. Thestate inviltrator proby, we have you see, mentioned early you'd- have todump in a ton of computing power to like outpace the network right, buteventually, like your validators need to kind of like upgrade their stuff,and it seems to me, like you, really do,require a heavy amount of like metal just to make sure that the the systemis kind of secure like you're not going to be running this on a Brasgray Hie,it sounds like so so. Single core speeds have beenflat lined for a long time in terms of the VDF itself. It's it's going to be pretty Toro. Toget two three t peatup like hlike, a large investment. Then you need likeliquid, Naturen, cooling and stuff like that. Beyond that, you have to spend abig pile of money to tape out y whole new chip, ut Thsm C, the latestapication process, and even than you might only get agn another two threeucks and two three XS not enough to to make an difference. N now ha athephysical limits of increasing like hatting these. These threcal stets of architectural changes within courselike this. It's not going t like we're, not we're not going to go any furtherafter that, so, but but what's interesting is thatthe number of actual you know what I remember these things being callthis non uniform memory or Numa course his growing, because he can just slapmore cords, an Loks faker that have no dated dependencies, and this is whatNvida does but th you know. cudacors are basically you know, they just treadway more stuff in a single weigt for every year, and that doubles everyevery two years in fact Um. So when we started, we needed four to Eigtytis toprocess a million signatures per second right now, one twan eighty ti can do awhole million. So, for my perspective, our like, what's actually happening iswe have you know the amount of parilel computational power is going to keepdoublic, because you know just the silicon wakers are going to get bigger,so they can just put more stuff on it and the process is still shrinking, sothings will get slightly faster and slightly you know more dense, but justthe amount of Siligin Mican ship is going to continue growing sobut ty.

I see what you're saying I'm sorry, Ididn't mean it. I thought you were don o apologize Um. The thing that kind ofalso Kindo is this: Assuming a computational like paradim that I don'tthink it's necessary. For instance like there are, there is research into likeships that are light based. You know what I mean like thy Thyr they're usingF photons is their method of of transferring data, and I'm not like I'm not saying thatthat those are anywhere near where we are but Ho H. don't you think, bydepending your architecture on h, our inability to innovate on oneparticular computing paradime or one medium of doing? Compute is kind of Um,not long term safe. I mean that's hard to predict right ifpin o computing does something andexpected in in twenty years, but Ithink you know for the next twenty years. I imagine the amount of computean each malidator is going to double right, so the cost of doing you knowfifty tps, fifty thousand TP s is going to sthrink by Haf Right, so ten, ourcapacity should double right. So the goal for this is like Prost to build anetwork that, if you have this normar spike in usage, that it can just handleit yeah I mean also like you got ta putthis intoperspective of the community you're serving L e. These networks arenetworks right there they, if, if, if what you're doing serves the use cases that are built ontop of it sufficiently, then good, okay, theoleoly problem with Um potential new paradimes in compute,which I don't think we really have to worry about, hitting the limits likecatching up to the limits we currently I have a silicon silar wafers Um iswhat they can say about consensus, whether or not they can take over likethe Suto, the security assumptions of the baselayer and doesn't seem to be the case like imagine. If we have like machinofbase communication and everybody in the world could be superconnected toeveryone else, best consens assalt, we don't neednwe, don't need anything elseR, but that's you know, probably a thousand years away, O Warwworr Y, like wwhat's, going onnow whet. Do you expect to go over next over the next year, sole Lo Werd, in a launch and that'spretty scary? That means that CNGAL will be will be running and don't be valued tothe underlying resource of GOVERNC, like the simil resistance right of theIO computer and if wee grew up an ovalue might be destroyed right thatthats scary, just as an engineer, you're, always thinking about whereWorkan things failed and U building a a network from scratch over the last year. It means you knowyou're going as fast as you can um R, I'm always like concerned aboutsecurity. A like do. We forget something Towe, not think of somethingyou know like: U, B thence that to me is like the biggest wory that where doyou go like? How do you? What do you do to to allow you to sleep at thaht? Whatsteps do you take to try and give you stronger confidence that what you'vedone is at least sufficient for the for the community you're serving, and yourhave things or processes put in place so that Um? If something does exist, it shouldeventually show up by by the happy path and not the bad one so mean like we're.Like Yo know, good engineers, we do like separation of function and likeany kind of cryptographic, operations use their own separat keys and which todo like Ita the limit, the privilege of everything youknow ' tot as much as we can they're also doing it face roll out. Soinitially we're probably going Alan with just staking and consensus working.We can observe that Kno it' Sov in the wild and it can't do much, but it's atleast cinsistent and and nothing is falling over uexpectedly and ten teringup transfers and tentern on the contract sention. So Tha. Those arelike the responsible engineering things to do, but you know: There's alwayssomething bagging in the back of the Lin ere, always obviously like they're,also diing an onit with a reputable security for but, like you know, like people havebeen working on firefox and chrome for decades and there's odays own. Thereeveryyou know remomor two, so you brought up well now you broughtup two things 'cause. I want to hear about your upgrade path and like Owll,that affect things if you need neneed one, but before we get to that, I'mactually kind of curious about your staking Um.

What is the model you have forvalidators in H, network to Stak? Is there slashing conditions? How do theyjoin? What is their lockup period? What is the? What is the staking incentive Um? So we have cooled down in WARMUPERIOD FOR STAKES A it's about two two weeks and the incentives are global inflation.So as avalitator Yo joined greate, you know a contract that participates inConsumsus is what thepans of valinator and then US ourself stake can delegatethe stake to you and if your votes ar violating a doubleboat or produce two blocks non eurs, but for the same slot you can getslashed and initially a slashing is pretty light. WAYIT's, like twenty fivepercent, there's no likeness, slauhing oranyfeellike that, but eventually the goal is to have a hundred percentsslatshing. Think a slashing is really what defines of security they network, but to get there. You know we need tohave a kind of a soak ton for for the software and for the operations forvalitators initialy Yo Tosee Mbut Sago. So why is slashing so sential to thesecurity of the network? I mean: Do you actually need so slash? Or can you justsay we just for that for that PTALA period you're just not going to getwhat you were going to get like. You know you have to be honest. If you'regoing to get your ouro stuff an we find that you're, not then you're kicked outof the validator pool. So honestly, like one of the worstpunishments he can do to balodatorses force their clients to redelegate right,because that means I have to go reach out to everybody again and deal withhim. Tell ep: Go ru this command or UN. You know, run the softhwar and Restat. That's pretty tough right! Thatrequires human, human time braves and communication, but slashing is ultimately like onceupstate the capital at risk that defines the security, the network, andif you don't have anything at Resk, then the security is very weak andthat's true about everything, Tet quin included. You know if the difficultyis low, then,would you trust a billion dollar like I don't know if yous noticed there is aresenb set of transaction to put a billion dollars into one account. IPEFON L T's. That's definitely like that's the security of Bitcoin and profOC work. Is the total capital put up front to try in Mine Thaks, which then gets you know,embedded into the recration of whose blocks of someone wanting to reproducethem? But I iin thinking Oi got I respondthat one I I' Sorry Cori. I think that's actually has more to do with thenumber of validators in the network and the fact that you have basically theleader election system. So you know, if you have this sort of like it's, thatleader election technically in a classical sense, but it's like you'veallocated this particular FREQUENTC. Let's ban, I don't know what t exactlycall it to specific this clock ranged as call you know: M Timeto, validator m.You know they can cause havoc in that partaler period if they're busiting Um,but if you have because he you know with different mechanismsfor for for consensus, you don't necessarily need to have that sort oflike leader necessarily to do all the you know, transactional orderingessentially so, and the commitment say that'sbecause of the way that you did the or it's inherentod the protocol that youneed slashing, because you have one person making a decision at any giventime. I think so. There's a couple portocolesthat are not prepeor Te, dovet Lashi like Algaron and I think Ablach. If I'mnmistaken and honestly think that it's a mistake because imagine like she gots firedyeah, I proof of war yeah. I think I got to exactly where I wanted to gowith this on full of sclosure. I work in O elapse. You know, I think, that's it's a interesting choice to make andit's a controversial Weatan, that's awesome, but people are exploring about just from the way. I think of it isthis one billion dollar transaction tha, one billion dollar account in pickitright now. Somebody published a privaty for those tickets basically stalled bitpoint for I don't know a hundred days 'cause, that's the amount of theinflation like from block rewards that you would need to add up to a billiondollars, but miners re are not really taking the heaviest for the picking theport that gives them the most money, they're intrinsically creedig or self,serving so if Youal potical without slashing imagine a twmillion dollars, avolum Flo thot and that's n less than...

...one percent of the gross payment volumein the world. So that's more tha, that's one of like the key differencesbetween proof of work and s sake as if you have to. If you have to stake a resource that is external tothe system, then you can never do slashing and that's how prefe work.Systemcs work as you're up you're up ther, putting up front an externalresource at the chance of of of being a block Rorducer, which thenallows you to give a real world value upon t the assets that are careated.Based on that Stiak with fstake. You have Tdebili a slash because you'reputting up in an internal resource, namely the token on on the thing, butthen, where do you get the? Where do you get the kind of maybe idea of what the value is Riafirst, sothe prof of word? I think that volume capacity of the network, I think, isobvious at least to be at least the inflation reward right. If, if YEU CONgenerate twelve million dollars a day, O pe should reward, then tole milliondollars a volume just the stink could probably handle theitte more right, butpind intutively t that makes sense. But if you have a trillion dollars a volumea year going through a network, how do you secure it? Like? You know, anAlgerond right, like I bribe a bunch of you know: BGP arounderengineers, so iy get a bunch of notes out there to create a partition and Icreate two blocks and then I also short Algo. At the same time and ye av been,I in the companies that ad depended on the trilly dolla worth of payments voll.I just want to point out there that I mean I have to defend the honor of. I don't think I ni think I think th.The real cause of that issue has nothing to do with t e, like the factthat you need that you need slashing conditions is more inherent tosituaations, where you, you, basically have deciders a group of you know,either a group of them or a or a single decider, that's kind of either elected,or they mine it or something like that. These kind of situations, even withalidators Li k, they're just validating like. If, if it feels to me like whenpeople are able to make decisions on their own, then you need to be able toslash them, but if they don't have to make any decisions, really they justhave to say this is what I know and the world kind of comes to consensus aroundthat. Then you don't need to to to to penalize them for saying somethingwrong. You just need them to get in line with the final state, and so that in that case, you just owna purely incentive facemile. Instead of like a penalty model- and I think thatthat I think I think keeping your open mind-I just want Ta- I have to be careful 'cause. I don't want to be like a shellfor like the company, I work forbut. I think I something to do that. You knowwhat I mean yeah. No, I agree. I think that so for us, it doesn't even make senseto slash you in the in the very like N, making mistakes atthe tip of the of t e spear riter h at the Front of the ledger, because there,like the network, is not settled yet and it doesn't really matter if youmissvot or produce a block. So you can actually like it's just way more workto construct the steat machine to a for not slashing there, but slashing lateron, if you're trying to be malicious and tryin to trick in exchange rigt andaccept a bunch of fake money thou through a partition, because obviouslythe time frames are much longer and then you're creating a much longer forit, but so that those kind of separating those and in terms of likewhat are the actual attack, lector a you- know, who's executing them. Ithink it's like security analysis that needs to be made. I Aa Asamn M. I don't know I'm Aeris. What you hadyou said like this sake is your security and distribution of that of a coin hasa lot to do with who has like who who's put up? They amount associated GIPI forget allthat Wyo have one security model associated with Um. How much steak the elevators have I'vego curious about like the real world, Um physicalled needs associated it beingabalidator. It seems to be relatively expensive. If I would like to beavalied, I no longer have O noanly have to put up a sustantial amount of moneywhich I'm willing to lose. If I, if I misbehave, but also, I need asubstantial amount of physical hardware in order to keep up with the networkTCETERA, so the hardor is actually fairly cheapbecause of Anvidea a single twenty Eightti, seven Hundr yN. twelve hundred bucks can process a million to fifty five ineteenthingisust per second, your modern d p. s for, like you know, can do like. Ithink a taribit have been with dat o...

...this quite a Bein of hardor and an likea machine. That's five thousand dollars an that's GOINA recommend it set up, so it is more of an investment than but people run Demos and they runsomething in the raspberry fire lopd up. But I think the reality is that we kindof have this professional set of valitators that are coming up: theCosmos and eow the block producers, fer mother, fror mother networks, but alsolike the people that are hobvious minors. Theytypically run. You know Iraq, AF, Peope, use right. T an they have a fastconnection because they want their blocks to be inotopicated at a decentrate. So there is actually quite a bit offolks that are in the space that are already way a way more than qualifiedtoround, a valitator N. on top of that, like ssay, you have the valiators tatset up and they indee the creation of the Blacche is done in a a in a safeway. What about people who just want to use the network and run nodes thatgather information? What type of harbor they need is such as the simple likeRaspberry Pie, mobile phone? Did they need a PC ine to rely on someone elseto give them data and hope that they do a good job with it. MSO like most Um, like kind of a commonapproach to the stuff, is using like clotes how you generaltle a prof of anevent that occurred on change and that's really like the foundation tolike I blache and communication and th. That's basically what we're building aswell, so between networks themselves or between you know, people that are justlike wallets that are using a wallet they don' need to Runa a foll validated.They can simply get proved that the network is doing the right thing orsomebody should be flashed so um. What about upgrades to the network h? So is there any advantage to like anewgpu coming out, for instance, so tthere's a hardwor up grade side ofthings like do? They have to keep up with the Jones as if they're, aprofessional validator on this network, an video comes out with an architecturethat still has paralyzation but they're able to do a tremendous amount more. Itsounds like it just benefits the network because you coan run morechacks Aracks. He is the interesting thing like people can deploy an thecheapest hard ro that can run the network so there's, but we can pricetransactions based on the most optimal capacity of the potocal. To imagine youknow we did a really good job. O Softer Engineers Hell be, can actually handle.Forty Gigabit persective networks take a big pile of work, tright not going tohappen this year, amaybe my years right. If we just had had the time to do it. That meansthat we can price the transactions as if we're handling twenty Ni ancientatens per second, because as soon as there's demand for that it'sridiculously easy to scale up your Hartbor, you just you know, take binnerpunch of commands and all of a sudden you have a data center and whatever youknow, club provider you want, and if that capacity sustains, then yougo investin rocks right on collication space or whatever so, but that's likein Thi, like extreme euscape, that all of a sudden weard the financial fabric.If the world that were Halk, you know twenty millions, an sections per secondto be insane right. So the network anly needs to keep upwith demand, but we can set the price based on thecapacity of what can be achieved great available, harbor B. You want to dothat, though 'cause. That seems as though like I'd. Imagine. The incentivefor validation is going to be based on transaction fees. If on said, it'sinflation schedule Right H, I I like I, I'm not sure transaction feesare long term sustainable model foor any network. So it's only a SP, a formof span protection versus Um, an actual incentivization scheme. It is, it is definitely anincentivization scheme. If we are at capacity because then let's say theyget, you know o fifty two hundred percent margin, which is totally fin atlike those twenty million numbers Yer. The network is making billions a yearright, it's justified because of those numbers. It's doing something crazy,itsishandling a lot of financial transactions. I don't know if trunsaction fees in the road can do it,because if you have a bunch of low volume networks that are handling youknow, billion dollars, transaction on biccoid howhow are the validators tat.We hearding, you know sixty cents or whatever, for ninebillion, for securing a billion dollars...

...worth of ashets. So, to me like there'sthis misalignement of incentives that well need to be corrected. D Am Not ahundred percent sure how that will? U Actually Resolve itself, but that'sthat's not a problem. That's unique to us. Even I think that's Upherihe ointas well O that Acte, I e, I think a ext WAA like to to move into somewhat of adifferent subject. That's tagentially related m. none of this matters. If noone uses Salana right, we have a lot of I've Gan, so many different projectstrying different schemes to like find the thing. That's GOINGTA scale to whatneeds to be scaled to allow for people to use it basically like as a universalchain across across these things. What's your goal, an Um like the community would like to useSalana and then how do you get people to come and join you build on top of itto create the value o trying to get so, I'm, like you, know, insanelybullish on the space like. I think we have this like huge opportunity to justget rid of ads from like everything that we do. If wecould just do that, TAT's plenty, an its Goin to say in everyone likes like m. If you look at like howmuch revenue finencials plike payments processors make it's two tilliondollars a year in revenue like the entire Gogle facebook Adzicov system,makes like one fourth of that seen fivehundred million. So there's a lot more money in the Camids Processing Si andif we can start building silly tings, like you know, red it orCarmas te token, and you can lik each other for mes right like or searchengine where Lix, instead of ads you'R micropay for the Secoure and it becomesfast and flawless and Um like Kend of transparenty it. I think we can startbuilding products that are these. You know, have mastive numbers of usersthat are self sustaining without stealing your data. So to me thatthat's like a enormous opportunity for this lnd, he next pace of the Internet and again like I, was like a teenagerin the nineties, and I saw the Kindo the hockey stick of the Internet. Wehad think forty million users on the Internet globally in ninety six andthat's about no many wallets star right now and I think only a hundred thousandwallets have more than point one bedcoint, an O. It's pretty smallmarket right now, but if it doubles right, if it actually doubles everyyear when we hit those like two threehundred million number of self custody wallets out there yo'll start seeinglike homegrown crypto oly, like phenomenonshappening like the frensters of Cripto. Like I don't know, if you remember thefrents Tar Right Lik, I had unye six. My six degrees of friends like it waslike the coolest thing and I had they had a scaling problem for a computerscience. Roem Ey had to recompete the scrap and retise Somban joyit. I did not know that about Tham, that'sinteresting, and it's funny. You brought up the wallet to to theInternet count owner from n from nineties. I mean I remember those daysand 'cause. I also was a teenager in those days and I would say that thebiggest leap forward for the Internet was you know. While we did need improvetechnology, better prowsers etceterectr and it all kind of started cometogether, the the Bagt Leep was the infrastructure broadband. Once broadband startedgetting adopted and th and people were using their cable wires, the littlecopper wires through the house. They were already us an watch. You knowtheir favorite T v shows to also browse the Internet at a wopping two megavitsper seper. Second Um, you know fr up from the fifty six k they were doingbefore. Suddenly things like you, two became possible, you know and that'swhen things started getting interesting M in my mind, and that didn't happenuntil early two thousands and then I thinkyou Tuve appeared about two thous of five Um oright anyway. Point is I s that, like it W S, it's aninfrastructure play right now, um, that's my feeling towards the world andyou are in the right direction. So congratulations ar building stuff and M.Thank you very much for coming on and Talkin with us about it. Yeah. For sure this was super phone.How do people learn more? How thehow d they gettingcontacted and for and soforth, Um so Godo Salana atcom, there's a discord group wor. All the engineersing o. We have a telegram Chanel, you know potcast, no shouting pot cast so achatovos yea osactwe. Actually wehad like the year protical Uy on realso.

So sharting is a lot of work but needto deal e new for Kal eiter for other things, but it's not a rear. The CAL rigt'sgon.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (108)