Discover more from WE'VE MOVED TO thedefiant.io
"Blockchains are an Extinction-Level Event; They're Like an Asteroid in the Sky:" Emin Gun Sirer
Sirer is betting his blockchain AVA has the best chance yet at becoming the world's new financial layer.
Hello Defiers and happy Friday! This week’s exclusive interview is with Emin Gun Sirer, Cornell University computer science professor who has been deeply involved in the Bitcoin and Ethereum communities from the very early days. We talk about how he got started in the field, building a cryptocurrency long before Bitcoin. Gun, as friends call him, then turned his attention to Ethereum, notably catching the bug in The DAO, but failing to alert the community about it.
After years scrutinizing existing blockchains, he’s back at making one himself with AVA, whose testnet launched two weeks ago. He explains the inner workings of this network, the first built over the Avalanche protocol, and how it can process several thousands of transactions per second at latencies of under one second.
He also talks about his plans to release a subnetwork called Athereum shortly after the AVA mainnet launches in July. Athereum will be almost identical to Ethereum; it will replicate its smart contracts and assets, and ETH holders will hold the equivalent amount of ATH. Sirer says the intention is for Athereum to serve as a safety net in case something goes wrong with ETH 2.0. If this sounds like Ethereum’s old friend wants to bring on some serious competition to the second-biggest chain, that’s because he is. Still, he’s quick to highlight he wants Ethereum to succeed and that he’s not after Ethereum dollars, but rather after money flows that are outside of Ethereum. This should make a great read for the weekend :)
You’re signed up to the free version of The Defiant, so you’ll get only part of the interview transcript. Subscribe now for complete access. Click here to pay with DAI at 70 Dai per annual subscription ($100/yr normal price).
🙌 Together with Eidoo, a cryptocurrency-powered debit card and platform for easy access to decentralized finance.
CR: You've been in the space for a while. You were deeply into Ethereum, you know the community very well, you were one of the few to alert about The Dao hack, but you were involved in cryptography and cryptocurrencies even before that. Tell us a bit of your background and how you got involved in this space and, and then we'll get into your latest news.
ES: When I became a young assistant professor one of the most exciting areas was peer-to-peer systems. And in every peer-to-peer system, there's this big tension between people who want to just get things out of the system and then people who who are more altruistic, more nice so to speak who contribute resources through the system. It's most stark in file sharing systems where people want to download a file. You have this great asymmetry where everybody wants something, nobody wants to put anything into the collective good. People were facing this problem for the first time in around 2001, 2002, and thinking about innovative solutions.
Bram Cohen at the time was coming up with with BitTorrent, which is a barter-based mechanism for solving this. If you want a block, you have to put up a block. And I looked at this and I thought, well, that's cool but barter is not enough. It doesn't reward you for being a nice guy when you don't need things. Most of the time I'm really just genuinely nice. I'm happy to provide resources to everybody else, even though I want nothing. But I want the system to remember that so that when I do want something, I get preferential treatment. I thought the easiest way of doing this is to come up with some kind of a technique for keeping tabs on how nice people have been. I thought we would do this with the help of a token that is issued by nobody. That was created out of the thin air, but not in infinite supply. I came up with a system called Karma and it was meant to facilitate peer-to-peer file exchange on the internet. 2002 was the year when we did the work and 2003 is when it was published.
Image source: Twitter
Proving Satoshi Wrong
CR: Wow. So even before Bitcoin.
ES: Way before Bitcoin. I was doing this before Satoshi was a thing. So I've been around. It was very well-cited.
CR: Were you part of the cypherpunk mailing list?
ES: I was, I didn't pay much attention to it. I just thought it was kind of boring. A lot of the people were circulating around the same ideas. I just pushed it out as a solution to peer-to-peer systems, magic internet money that's created out of nowhere and is used to keep tabs on people's behavior online, whether they've been good to the system or not. And there was no controller, there was no central bank. And so that was an okay idea. It's very well cited by academics, but it was missing two key pieces. Satoshi came six or seven years after me and he came up with a consensus protocol based on proof of work.
And the other thing he did of course was he came right after the 2008 crisis at a time when people were really anxious about the value of their money, of their fiat currency. He came in with a very strong narrative saying we must replace fiat currencies, down with the state, down with the money printers, etcetera. That narrative found traction within the libertarian community. And the rest is history. His timing was impeccable. His his narrative was a great fit for the zeitgeists then.
And after Bitcoin took off we ended up looking carefully at Bitcoin to see if Satoshi really got the protocol right. And I, with my colleague Ittay Eyal, who is now a professor at Teknion, we found that Satoshi was wrong. Bitcoin is not perfect. In fact, you're better off not following Satoshi's prescribed protocol, but following a slight variant which gives you more coins. This was a mining strategy called selfish mining and then when we published this people went crazy. They attacked me, they attacked Ittay, I got all sorts of threats, all sorts of attacks. They crowdfunded a simulator to show us that we were wrong. And then the guy who wrote the simulator said well, they're right. Everything they said is correct. And the graph he drew is exactly the graph we drew. So it's really cute. This is why I love being a scientist.
CR: Wait, so, so did somebody ever use this selfish mining system to try to attack Bitcoin? I mean, because if it's a better system, then maybe you can use it to profit, right?
ES: There are a bunch of risks associated with engaging in any such thing. And so I'm not aware of anybody that has used it against Bitcoin. I do know that it has been used against other altcoins but not against Bitcoin because to do it right, you need a substantial number of mining rigs. So it's not trivial and it also requires know how, and until recently miners did not have the know how to do this. They were barely capable of running the code as given to them by the developers. They weren't able to develop much code themselves until quite recently. So I think we're going to see more attacks.
The DAO Hack
CR: At this time, were you already at Cornell?
ES: Yeah. All of this is at Cornell. After that we started looking into various different aspects of improving cryptocurrencies. I worked on improving the security of coins at rest. People get their coins stolen all the time. And so we came up with this notion of vaults where if your money in your vault is stolen, you can get it back from the hacker. Then worked on Bitcoin Next Generation, Bitcoin NG, which is a new protocol that's much faster than Bitcoin, but preserves mining.
And and then I turned my attention to Ethereum and smart contracts. So I was involved in The DAO. Me and Vlad Zamfir and Mark Dino, we found nine different flaws in The DAO and we cautioned everybody, we had a paper called A Call for a Moratorium on The DAO. I also found the bug in The DAO, the reentrancy bug, but I had a summer cold that day and I was feeling miserable. So I fired it off to my grad student, Phil Daian, and said, hey Phil, take a look, see if there's a reentrancy bug here on the line 666. And he looked, and then he said, no, I think there's something there, but it cannot be exploited. And I was like, okay, I've got bigger worries, you know, with my brain coming out of my nose, I felt terrible. I've worked on layer two. I built something called t-chain. It's still the fastest Layer 2 out there. It doesn't suffer from the problems that Lightning suffers.
Bitcoin’s Stifling Narrative
CR: Is that a Layer 2 for Bitcoin?
ES: Yeah, it's a Layer 2 for Bitcoin. We were doing 125,000 transactions per second on Bitcoin. And so it's kind of cool. But I veered away from the Bitcoin community. I think that that system is in the hands of developers who are a little inwards oriented and they aren't changing the system much. Bitcoin's charter is to compete with the dollar. And that's a fine charter. It's great and somebody should do that. And good luck to them. I think that's a tall order. And I think there are many other much more exciting, much more immediate uses for blockchains ahead of us.
CR: Do you think that they'll be able to compete with the dollar in the longterm?
ES: I wish Bitcoin the best, but you know, what are the chances of replacing dollar? I don't know. Bitcoin had its moment in the sun and maybe six weeks ago we had a big, big, huge crash and it moved in conjunction with every other asset. So I think it's not living up to what it's built to be. I think it's going to be very difficult. It's going to face regulatory challenges and to really see motion, to really go somewhere, you've got to have adoption. If you are simply a buy-and-hold kind of a thing, well then, how far can you take that thing? You can maybe sell it to old people, people who really love gold and then you can say, well, this is kinda like gold. I think it's kind of a boring, stifling narrative. But as I said, if it works for them, that's wonderful and I hope Bitcoin goes to the moon because we will all travel together.
Blockchains as Extinction-Level Event
CR: You mentioned there's more exciting applications for blockchain technology than potentially replacing the dollar or becoming a safe haven like gold. So for you, what are those bigger goals?
ES: I believe that blockchains are here to transform the world. I believe that they are an extinction level event. They're like an asteroid in the sky and there are a bunch of these dinosaurs on the ground and these dinosaurs go by names like Charles Schwab and they go by names like Swiss Re. And what's the common feature among these dinosaurs? They're incumbents. They are sitting on their laurels, they are not innovating at all and they don't offer any value proposition to anybody new. Blockchains will disrupt them and they will provide a huge opportunity to savvy startups that know how to position themselves.
And it's not going to be a transformation overnight. It's not going to be like one day we're going to wake up and all the stocks are now on the blockchain or something. It's going to be gradual. It's going to be that every new company will have fundraised on a blockchain, so that by the time that we're older, suddenly everybody's stocks or valuable assets are already in digital form on a blockchain. So it's going to be a slow transformation, but inescapable and completely surrounding us. So I'm really hopeful about the area. And where will the disruption take place the most? It will start to happen in finance. Everybody is focused on the money use case and that's the hardest use case. You can see how far far Bitcoin has gone in 10 years. I cannot buy anything with Bitcoin where I live. Absolutely nothing.
I believe that there is enormous amounts of assets that can be digitized out there. Trillions and trillions of dollars worth of it. They're sitting on people's balance sheets, they're hard to trade. Their reach is limited. Reach is a technical term in finance, it just means the number of people that you could actually sell an asset to if you tried. You will find that it's actually very difficult.
It’s Impossible to Buy Putin Bonds
I have a funny story here. I tried to buy bonds issued by the Russian government. It was impossible. I mean, they were paying 14%. My bank was paying like 3%. I was like, I don't want your 3% measly little interest rate. I want 14% and they said, well you know, sir, if you want to buy that kind of thing, you're going to have to talk to our investment advisor. Okay, sure. Let me do it. And they were like, well, he comes only on Tuesdays and Thursdays. I was like, alright, I'll come back.
So I came back on a Thursday, physically go into an office, this was in the early 2000. It was a portly gentleman, with the coolest mustache, a suit —he looks like an old school-banker and he gave me one of these handshakes and everybody who's been to Charles Schwab knows this handshake well. They give you a banker's handshake and the reason is you have to trust them and they have to engender trust. The entirety of that operation is based on trust. And so they give you this firm handshake and like every programmer, I've got a whole bunch of problems from typing too much. So I put out my hand, I'm like, I know what's going to happen and this guy tries to like crush my RSI-ridden hands. And I'm like, okay dude, you hurt me, fine.
I want to buy some Putin bonds. And he goes, sounds to me like you want to buy some junk bonds? I'm like, no, Russian sovereign bonds. I just want to buy that. Just find me the vehicle. The coupon value was a million bucks. I didn't have a million bucks then. I said I wanted to buy a fraction of it. How hard could this be? And he's like, well if what you want is exposure to a junk bond, I suggest you buy some GM stock. And I was like, dude, GM makes coffins on wheels. The other guy, he's got radioactive stuff, he's got nukes, he puts people in coffins, you know, I want to bet on that guy. Anyhow, so this is the state of the world even now.
Ethereum Falls Short
CR: It's crazy that like the financial assets are still constrained by geographical borders in a world where everything else is digital, information is digital, but finance isn't. It makes no sense.
ES: It's astounding. It's absolutely astounding. I double dare anybody listening to this to go out and try to buy some sovereign bonds issued by another country. It's impossible. And I don't understand why. There are a lot of assets like this out there and and we need to get them onto blockchains, we need to extend their reach. It's good for everybody. It provides funding, it provides liquidity, it brings down the risk premia. It's just great for everyone. And and that's where the action is going to be. But to go after that, you gotta, you have to have a lot of innovation under your belt. You cannot repurpose Bitcoin for this and you cannot repurpose Ethereum for this. It's just not the right model for these kinds of assets.
CR: Okay, that's interesting because I think many people in my readership will disagree with you that Ethereum is not meant for this, because this is what people are trying to do with DeFi. So I'm really interested in your take on why isn't Ethereum capable of being the financial layer of the new economy and how are you proposing to make this a layer?
ES: So I wouldn't say it as strongly. Ethereum is trying to do a whole bunch of things. Just so everybody is clear, I love Ethereum. I spent a lot of time working on it. All the Ethereum people are very close friends of mine. I am very much a part of the Ethereum community and I've always considered myself as that and I'm not going away from the community for any reason whatsoever. So let's establish that. And I happen to love decentralized finance as well. But the Ethereum model is not suitable for a lot of the assets out there because Ethereum does not give you control over the network, that's the short of it.
Control Through Life-Cycle
CR: What do you mean by that?
ES: Ethereum gives you a lot of control over the behavior of your assets. So suppose you want to issue Camila Coin. Suppose you want to fractionalize New York City real estate. Real estate obeys certain rules. You could then code those in a smart contract. But what you cannot do that's very, very difficult in DeFi and just about impossible in Ethereum is provide that for the entirety of the life cycle of the asset.
But let me tell you what is actually outside the model that my system AVA is looking to fix: You can define all of these rules. You can come up with a whole set of business logic that works just for Camila Coin as of today, but tomorrow when there is an unforeseen event, when somebody comes to you and says, this person holding this thing must necessarily be cleaved off your system. We have determined that they are, I'm just making stuff up, that they are terrorists, they're on the OFAC list. And so now you have to blacklist somebody. And if you did not think about this particular scenario in your code, then there is no accommodating this.
Ethereum in terms of architecture, in terms of operation, pretty much copied the entirety of its system from Satoshi. They changed only the virtual machine, made it Turing complete, which is a huge step forward. Ethereum is the same way. There's only one EVM and there's only one smart contract you deploy on it and you will find it very difficult to control what happens inside the network. This was exactly the case with The DAO. Something unforeseeable happened with The Dao and then people found themselves having to do a global revert to make one small change to one smart contract.
So this cannot do if you are a large company in charge of issuing assets on a blockchain, you cannot take the risk that the assets are outside of your control. That there are unforeseeable events that will cause you to have to ask for global help just because there was a, there was a mistake or something in the smart contract. And back to my legal point, what is the legal set of laws? There’s no legal set. Bitcoin is extra legal, it's its own system. And we try to sort of make it work at the margins, at the interface between Bitcoin and the legacy system, at that interface lie the exchanges and they come under regulatory scrutiny. The system itself does whatever it does.
If you look at what people want to do is they want to issue assets, but they want to control what happens to the asset after issuance within lim its. There's no blockchain out there that I know of that gives control to the participants over the participants of the chain.
Issuer Designates Participants
What AVA does, is it allows the asset issuer to designate a certain class of participants so that you can say, this is Camila Coin and it represents fractionalized real estate in New York City. And the nodes that validate Camila transactions are say, purple nodes. Now you get to designate who becomes purple. You can say things like, well, to become a Camila validator, you have to be in the US and therefore under OFAC regulation. Or you can say you can be anywhere you like, but you have to sign an agreement that you will enforce OFAC and anything else you might want.
You can express things with AVA about the network that you could not express in EVM byte code. Legal agreements top the list of things that people want to express but cannot in code and there are other things as well. There are resource requirements. Look at the Bitcoin craziness. Look at the way Bitcoin splits up. Look at the whole the blockchain debate. The blockchain debate is all about people saying, Hey, all these nodes have enough resources for a bigger block, and there's people saying no they don't, and there's an unsurpassable fight that caused the fork.
The greater AVA network has a very low common denominator. But the Camila Coin can say, look, to participate in my network, you have to have a 64 gigabytes of Ram. And suddenly you can say, and smart contracts on Camila Coin that relate to Camila's fractionalized real estate, they can be much bigger than normal.
CR: So the issuer is able to decide what validators or nodes can process its token and also decide on the resources required to transact with them.
ES: Yes, exactly. So this opens up a brand new world of possibilities. So for this real estate example, you could say things like all the purple nodes, they have to sign an agreement and put up a bond that says that they will keep archival records for the next 20 years. It's real estate. You gotta do that.
Mini Blockchain Nations
CR: Are these nodes known entities? Because if they're taking this kind of liability and if they have to check whether the token holders are meeting these requirements, it is actual work. Is it something that can just be automatically handed over to thousands of nodes in a network or are they, are they supposed to be known entities doing this work?
ES: That's a great question. I don't think they should be known entities. AVA the network is very similar to Bitcoin, Ethereum, and so forth. We're talking about assets built on top of AVA issued by third parties and those nodes are determined by the third party issuer of the asset. You're issuing Camila Coin, you're going to determine who gets to be purple. It's entirely up to you what you demand of them. What I think we will see is that there will be some common large sub networks. So there might be an American sub network, there might be a GDPR sub network for the European union and there might be other smaller ones.
And of course once you have the capability then we're going to see people get really creative with this because there's going to be a market for paying people for archival, for privacy, for security, for extra resources and so on. And then people will come up with cool ways of discovering these nodes without necessarily having to know who's behind them by name.
CR: It's a really interesting concept and I think it'll be really useful for issuers to be able to determine different rules and regulations for their own tokens and especially to change them as things change along the way. But one question there is, doesn't this have the risk of having the same problem you had with the Russian bond? Doesn't this create these separate jurisdictions that limit the flow of value and assets across the network?
EG: Great question. No, I think that this allows a whole lot of things to be digitized into blockchain form that weren't, or could not be with the current systems. And then there are many other things we are doing so that assets on this network are discoverable. So reach and discovery go hand in hand. We're working on a decentralized exchange to facilitate the transfer of these coins so that somebody in Singapore can use digital gold, actual backed gold, to buy your fractionalized real estate in New York city or vice versa. You know, somebody wants to buy into whatever else, somebody wants to sell Russian bonds in exchange for Tether or whatnot, that should all be possible on a system like this.
And absolutely, I think two things are critical. One of them is being able to discover the assets on different sub networks. And the other one is being able to trade them. We see the job of AVA as some kind of a narrow waist. So just like the internet, the IP protocol on the internet, the IP portion of the TCP/IP protocol is the common protocol that allows many different networks to come together. AVA the global network allows people to bridge the Camila network with the Kevin network, with the Collin network, with the Michael network and so forth to allow all of these things to transact with each other in a global cohesive whole.
CR: Got it. So maybe according to each token's specific rules, not everyone will be able to transact with these coins, but the goal is that most of the network will.
Avalanche Consensus Protocol
We've talked about this quality of increasing the reach of assets and being able to modify the requirements and rules that an issuer has for their own assets and tokens. But I know there's other features that differentiate AVA from Ethereum, Bitcoin and other blockchains. Can you walk me through these pillars?
EG: Of course. So there are three pillars for AVA. One of them is a much, much faster consensus protocol. And a lot of people talk about this and a lot of people are frankly just bullshitting everybody. I'll just be very frank about this. They make a lot of noise. I've even heard Justin Sun talk about all kinds of things that I'm going to talk about and I know the technology that these people have. It's nothing. It's technology from 1980s or 1990s recycled with a little bit of a blockchain lipstick on it, but it's the same old pig.
The are of distributed systems has been around for 45 years. In those 45 years, the first 30 to 35 years, were spent on protocols that we call classical protocols. These are protocols where everybody knows everybody else and everybody asks everybody else for agreement, for consensus. And we knew about these protocols. Satoshi knew about these protocols and he thought, eh, these protocols are boring. They're super fragile, they're not a good fit for the kind of open systems that we want to build at internet scale. And he was right. So there's classical, that's one big idea and then came Satoshi's consensus protocol, which is a second big idea in this whole space.
The third big idea is the Avalanche consensus protocol. It works differently from everybody else's. It works much faster. It achieves latencies on the order of one second or so. And it achieves throughput on the order of tens of thousands of transactions per second.
Faster than Visa
CR: And you've already proven this to be right? It's working, it's actually doing this level of throughput?
ES: Yeah, we just opened up our Testnet last Friday, so I'm talking to you on a Tuesday. So the test net has been live for four or five days right now [two weeks at the time of publishing]. And so in other settings, prior to testnet, we have achieved as high as 19,000 transactions per second. The current testnet with all the instrumentation and lack of optimization is achieving about 6,500 transactions per second. That's a hell of a lot of transactions. I remember talking to the Bitcoin developers about the blockchain debate and they were telling me, look, we're currently doing about 3 to 7 TPS and Visa is doing 2,000 to 5,000. We will never be able to compete with Visa. And therefore it doesn't make sense to go from 3, to 6, to 9. That's about the kind of thing we can do by expanding the block size. We cannot compete with the Visas of the world. And in fact, no consensus protocol can.
Well, it turns out they were wrong. In May, 2018, the Avalanche consensus protocol was released by an anonymous party that calls itself team Rocket. And those of you who know Pokemon will know that team Rocket was Satoshi's biggest enemy. So team Rocket, essentially one upped Satoshi. They came up with a protocol that's much faster, much higher throughput and maybe the clincher, this is probably the most important part, is it's much more inclusive. You cannot meaningfully participate in Bitcoin anymore. You can buy the token and that's about it. To participate, to become a miner that has any pull, you have to put in tens of millions of dollars into mining equipment and Avalanche is not like that. There is no mining. It's super fast as I said. And anybody can participate. The protocol can accommodate tens of thousands to millions of participants.
CR: Does it require staking ?
ES: It does require staking for AVA. That's correct. But the key part is, there are other proof-of-stake protocols that use classical consensus at the root because everybody has to talk to everybody else. Those protocols can't accommodate more than about a hundred people, because a hundred people trying to talk to a hundred other people, that's 10,000 messages in flight, a hundred squares, there's, there's 10,000. It's a lot of messages. That's why those protocols are slow and limited in scale.
My student is the lead author of the protocol that is used in Libra, Facebook's Libra. And he improved the heck out of classical protocols and and managed to get it to scale to a few hundred and Libra adopted it because it scales. And then Ted who did it over the summer came back and said, look, these classical protocols, they aren't going anywhere. This is not where the action is. I want to work on Avalanche. And he's been working on the Avalanche protocol as a protocol architect.
CR: This team Rocket that published avalanche, are you in any way related?
ES: They talked to us, but I do not know the exact composition of team Rocket.
CR: So you're building on top of this Avalanche consensus protocol, but you're not the author of this protocol.
EG: I am not. You're going to have a bazillion questions about the identity of team rocket. And I have received a lot of inquiries from other academics by the way, because they're like, whoa, there's a lot of math going on here. What's happening? Who are these people?
CR: So using this Avalanche protocol is how you were able to build this new blockchain called AVA. And what this allows is a much higher throughput than Bitcoin, which is at 3 to 7, and Ethereum which is at about 15, so this is orders of magnitude higher. What about compared with Eth 2.0?
EG: Potentially. I have a lot of thoughts on ETH 2.0. I happen to love Ethereum and I have a bunch of misgivings about the entire idea of sharding. ETH 2.0 is based on the sharding idea. So when you have a problem and you are new to distributed systems and your problem involves scaling, right? You want to scale the immediate idea that often comes to people is let's divide our problem into K little pieces and and so we can be K times faster this way. So ETH 2.O is trying to shard the blockchain into 256 pieces and it's hoping that that way they get 256 times 15 or whatever it is, kind of an improvement, maybe more, in terms of throughput.
It's a fine idea. But it's the most naive approach to trying to get scale. If you had a problem and you thought sharding it K ways would would help you achieve K-fold improvement, what you really ended up doing is you had a problem and now you have K problems in terms of K different shards plus one additional problem of inter-shard coordination. So every single problem you had on one chain still exist on every one of your chains on every one of your shards.
CR: If you have that many shards, then you're still increasing the number of transactions that you have on the blockchain overall.
ES: Right. Perhaps that only happens if you have a uniform distribution of the workload and that we know that's not going to be the case. We know that the workload is exponentially distributed. We know that some shard is going to have Dai on it and the shard with the Dai contract on it is going to see the most use. So everybody will want to be on that shard because it's going to be quicker to communicate if you're on the same shard as Dai. And so you might very well have 255 unused shards and one Dai shard that everybody wants to be in on.
I see a lot of the risks with ETH 2.0. I wish Vitalik and other friends great luck. I have tried to help them as best as I could and it's very, very tough. And if they succeed, let's assume they succeed. If they succeed, then Ethereum is going to have latencies on the order of five seconds and throughput will go up depending on the workload, I'm not sure how much it will go up by. So that's what we're looking at. I wish them luck, but that's not the approach we're taking.
CR: So how is Avalanche different? How can it reach that much like higher throughput?
ES: Oh, it's a completely different consensus protocol. It works based on a different core idea. ETH 2.0 within a shard, everybody uses a classical protocol where every node still talks to every other node. It's a simple signature accrual protocol. So within a shard, it's very similar to EOS. It's like they took EOS and divided it 256 ways and now they're trying to make the resulting mess work. Avalanche is different from this. Avalanche's pattern of communication is much more efficient. Instead of me as a node having to get approval and okay, from every other node individually, what we do in Avalanche is we do a probabilistic cover of the entire space.
What does that mean? I poll. I pick a few people and I say, "Hey, do we agree that Alice paid Bob?" I asked this to three people, maybe five people, but not more than 10 people, a small number of people. I might have tens of millions of people out there, but I'm only checking with three to five or to 10. But no more than that. The magic is, and this is the magic in that initial paper from team Rocket, the magic is I don't need to do this all that many times for me to have assurance that covers the entirety of the set. So I ask a small number of people and I get okay from them. You ask a small number of people and you get an okay from them. And together we cover the entirety of the space. And of course there's much more to it than a lot of subtle math.
The reason for why this is correct is really the most difficult part of the equation here, but it's all laid out in the initial paper and it yields a system where, you know, after a modest number of rounds, you get you get consensus. Now, modest number of rounds translates to about a second or so in terms of latency. Everybody talks about TPS, transactions per second. That's a sucker's game. All of my graduate students in their first two years, they're all focused on transactions per second. There are a billion games one can play to lie about transactions per second to fake boost those numbers. We're seeing this play out with new coins coming out this year. There are lots of people claiming millions of transactions per second. And then when you look under the covers, they're barely doing a few dozen. So TPS is not where the action is. The action is with latency. Ask anybody who has tried to build something real on a blockchain. In Web3, they will tell you that they want latency.
CR: And latency is how long it takes to confirm a transaction?
ES: Right. How long does it take after you submit a transaction to finalization. How long is that period.
CR: And why do you think that's the metric to look at?
ES: Blockchain systems when they are trying to be used the typical bottleneck is the throughput bottleneck, but the thing that really gets in the way, in a way that isn't so obvious unless you know to look for it is that they take forever to confirm. So Ethereum, for example, takes 500 seconds if you want to really be sure that your transaction won't be reverted. Nobody has 500 seconds. You can't just wait one confirmation on Ethereum, it can get reverted. It's cheap to revert that. If you want to have a an interactive application one where people click on things on the internet and it responds, well, you need to get into sub-second latencies.
And I keep advertising about a second. The actual number we're seeing on the network right now is 90 milliseconds. So it's less than a 10th of a second at the moment. Now, of course, that's not going to stay there. You know, as the network grows the latencies are going to increase a little bit. You should see how fast this thing is. And once you see AVA in action, there is no going back. Everything else looks archaic in comparison. I try to use MetaMask these days and it's just those latencies of having to wait for the transaction to confirm and so on are insanely, insanely long. And a single confirmation of course is insecure. We haven't seen anybody get hit by that yet, but we will. But even assuming one conf is good, the time it takes on MetaMask with Ethereum today is 15 seconds.
CR: And the way to achieve this is by having this consensus protocol where it's not necessary that all of the nodes or validators talk to each other, and you do this with this mechanism where you have two groups of people asking nodes around them, what the right answer is until that grows exponentially and the whole group of nodes answers the question.
ES: That's exactly right. So the consensus protocol deep down is one of repeatedly polling the audience and then everybody in the audience trying to align themselves in a big agreeing group and it turns out that this process is super fast and that gets us very low latencies and very high throughputs. And as I said it scales to a large numbers of participants.
[ … ]
Paid subscribers have access to the full interview, including sections on:
AVA’s on-chain governance
The planned “friendly fork” called Athereum
AVA not going for Ethereum dollars
The need for healthy competition
Subscribe now so you don’t miss any of The Defiant content. Subscribers reading this post: Head to posts marked with the little lock to see the full content.
The Defiant is a daily newsletter focusing on decentralized finance, a new financial system that’s being built on top of open blockchains. The space is evolving at breakneck speed and revolutionizing tech and money. Sign up to learn more and keep up on the latest, most interesting developments. Subscribers get full access at $10/month or $100/year, while free signups get only part of the content.
Click here to pay with DAI.There’s a limited amount of OG Memberships at 70 Dai per annual subscription ($100/yr normal price).
About the author: I’m Camila Russo, a financial journalist writing a book on Ethereum with Harper Collins. (Pre-order The Infinite Machine here). I was previously at Bloomberg News in New York, Madrid and Buenos Aires covering markets. I’ve extensively covered crypto and finance, and now I’m diving into DeFi, the intersection of the two.