How Blockchain Tech Advances Benefit Other Industries

Transcript

[Music] welcome to web 3 with A6 and Z I'm sonel and we're excited to share with you our 50th episode of this podcast for this Milestone we decided to share not only a conversation all about blockchains crypto and web 3 but also about how the field is driving advances in several important technologies that can be used in and applied across other Industries well beyond crypto our special guests are Dan Benet a6c crypto senior research adviser and Prof professor at Stanford University and Tim Ruff Garden A6 andz crypto head of research and professor at Columbia University the two discussed topics such as automated Market making credible auctions collusion and other mechanism design not possible before as well as zero knowledge trusted execution

environments and much much more we also discussed the recurring theme of how web 3 provides a laboratory not only for experiments and governance but for macroeconomics and much more throughout Tim and Dan offer many useful explanations for anyone to these Technologies are seeking to understand why they matter in the big picture as a reminder none of the following is investment business legal or tax advice please see A6 and.com disclosures for more important information including a link to a list of our investments the first voice you'll hear is Dan's followed by mine and then Tim with the two voices alternating with the first half and the second half of the episode so I guess what we wanted to talk about is the fact that blockchains

overall have caused a lot of technology new technology to be to be developed it's kind of remarkable how many areas of Technology have been pushed forwards by the blockchain community just as an analogy I kind of like to draw this analogy to the to the space program where you know NASA was going to the moon but along the way they invented all these technologies that were beneficial to the rest of humanity you know things like the microwave oven things like velcro of course you can't forget Tang all these things were developed for the purpose of going to the moon but they were beneficial to humanity overall and in some sense similar things are happening in the blockchain space where all these Technologies are being developed for and in service of you know the goals of the blockchain but in fact they are helpful for many aspects of our

society that have nothing to do with with blockchains they have applications way way outside that are quite interesting and relevant to the rest of uh Society so before we go through the examples could you guys also just help me think through a little bit more broadly just even more context setting on why that's the case yes there's always this kind of exhaust that comes out of existing innovations that can lead to other Innovations but why do you think that blockchains in particular at this moment in time are leading to these developments in other fields is there something specific that's happening right now or why now to build on what Dan said with this space program analogy I mean I think it's almost like the definition of a Grand Challenge I mean obviously you're going for a goal that

you really want in its own right but that almost like what defines a Grand Challenge is the requirement to invent more broadly useful Tech and ideas along the way and so you know when I think about you know the vision of of blockchains and web 3 is you know sort of a a robust powerful computer in the sky that sort of Open Access anyone can use that can do sort of you know computation that you can count on I mean that's the kind of Grand Challenge that basically requires many different disciplines within computer science to come up with new ideas or take ideas that existed in principle and really make them really make them practical so I think it's because that vision is so ambitious there's no way it can happen without fundamental breakthroughs all

across computer science and maybe to to build on that I I would say that it's not surprising that web 2 came before web 3 in that centralized systems are much easier to build than decentralized systems you know decentralized systems are just complicated they're hard hard to build hard to get right what this community is doing is building things that have never been built before as a result it raises challenges that actually haven't you know we haven't faced before and so to you know to solve those challenges we have to develop new technology and that's exactly what the community is doing developing new technology towards the goal of making web 3 work and scale and be secure and private and so on and that creates all these new technologies that are then useful in lots of other areas so basically what I'm hearing is the what

and then for the why now I mean there's obvious things like just more compute being available more internet connectivity etc etc but is there anything else to add on the why now like why at this moment in time we can maybe view you know 2009 you know the launch of the Bitcoin protocol I mean that was an interesting moment in time where I think the community's thinking in different disciplines had made a lot of progress you know we understood game theory in a computer science context you know better than we used to we understood distributed systems better than we used to you know cryptography is better than it had been so it was a moment in time where there were these things I maybe wouldn't call them mature but these different threads have matured enough right the Nakamoto with a a lot of extra Ingenuity on their part was

able to actually sort of Define this protocol so you know lots of us in the space you know that's kind of the Big Bang moment and then of course once you have the vision you want to really you know you want to realize it as sort of a grand scale with a lot of robustness and a lot of performance so then of course all of those threads which were somewhat mature but not that much we have to push them much further to to get the full vision exciting and by the way just on the grand challenges note Tim I remember you gave a talk at the cryptoeconomics security conference Cesc on this very topic of what are some of the grand challenges but just for the listeners and the show notes I'll include a link to that talk which is on Tim Ruff Garden's YouTube channel as well okay

let's talk specifics now about how blockchains are influencing other Technologies Beyond applications in crypto and web 3 sure yeah I think the big one the big one is and kind of the ones that we start with is really this remarkable truly remarkable progress that happened in the area of zero knowledge proofs right so zero knowledge proofs is something that was first proposed in the 1980s and it's really quite remarkable it's for 40 years almost this was considered a very theoretical area there are hundreds of papers since the 1980s about zero knowledge proofs but they all treat the area as these are things that we can do in principle these are not really things that we can at Le at least a scale we can't quite use in the real world and

then what happened is basically it turned out that that zero knowledge proofs or I should say proof systems in general they turned out to be sort of an enabling block for a lot of blockchain Technology it enabled scalability and bridging it enables privac and compliance so the web Community literally needed zero knowledge proofs or again proof systems in general to scale up yeah we needed to kind of run complicated computations in in a proof system as a result it was a lot of engineering focus and just getting this thing to work and so a lot of innovative ideas in and speeding up the pro time many innovative ideas in how to implement these things efficiently and in fact it's like every every week there are new results being posted on you know ways to speed up these proof systems

ways to implement them more efficiently and so on and at this point we have 10 different libraries programming languages and entire software Stacks that are devoted to making proof systems easy to use again this was done for the purpose of the web 3 Community because it's needed for scalability and privacy in the web three space but now that that whole software stack is available anyone can use it I know it's obvious to us but just sort of give me a quick definition of what are zero knowledge proofs and more broadly zero knowledge proof systems yeah sure so I guess just as we start let's just quickly say what proof systems and zero knowledge proofs are uh so a proof system basically lets someone prove that a certain fact is true and then the proof is very short and it's

very fast to verify so even though the statement that's being proved is very complicated it could have taken you know potentially many minutes or many uh many hours even to generate the proof itself is very short and very fast to verify so the verifier can just look at the proof and immediately be convinced that the statement is true we can add the additional zero knowledge property and that just means that the proof itself reveals nothing about why the statement is true so I can prove to you that a certain fact is true and you will learn nothing about that fact and the web 3 Community needs those tools we need proof systems for scalability because if we want to verify that 100 transactions are valid rather than having the block producers or you know the nodes on the blockchain themselves verify those 100

transactions one node can just produce a proof that the 100 transactions are valid and then everyone can just verify the proof rather than verifying the transactions so that helps with blockchain scalability and the zero knowledge aspect helps with privacy in the sense that uh we can have transactions that don't reveal what they actually do but the zero knowledge proof will prove that the transaction is valid so money is not being created money is not being lost and so on right just as a quick example on buying a house right now I have to share all this private financial information with my real estate agent and the bank and a bunch of other people and I'm kind of annoyed that I have to give all these strangers access to information that as long as they just know I'm above a certain threshold or that I have a certain score

or that I have a certain as long as this is true that should be good enough they don't actually need to see the specifics of all that info yeah absolutely absolutely in fact there's a there's a system called ZK docs yes of course Sam Ragsdale yes exactly Sam built this beautiful system called ZK docs where you can actually write a document that has a lot of private information to you in the document but you can read out all the private information and then just prove to whoever needs to verify the document that the information is valid so they know that you know you're eligible for a mortgage but they wouldn't learn anything about your financial records okay great the one thing I will just say for those who want to learn more about zero knowledge proof systems we've actually compiled a beautiful zero knowledge cannon on our

website that our research partner Justin Taylor who's helping make a lot of theory into practice has been pioneering and he also wrote up like a beautiful curated Cannon of the of work that's contributed to all of this and evolved so I'll share that link in the show notes great yeah so actually one more thing I'll mention one interesting fact that the again the web 3 Community wanted to make zero knowledge proofs as efficient as possible so they introduced this prize it's called the Z prize which was a competition to actually have people come in and improve implementations of certain aspects of these proof systems the Z prize actually had Fairly significant Awards associated with it the awards were paid off off by the web 3 community and it's really

interesting a lot of folks came in and started building these uh improved implementations in particularly particularly how to get things to run quickly on a GPU yeah programming in GPU is not that simple and so folks that were like former Nvidia employees jumped in on the Z prize and it's not too surprising that kind of former Nvidia employees actually won the won the prize and as a result now we have like super fast implementations of cryptography of these proof systems that can run on gpus be really really efficient and perform much better than we could have done without the zprice so this again it's an example of how the web 3 Community is contributing to developing software that can then be useful in other areas it's also a fantastic example by the way of

what you started off with about sort of the exhaust that comes out of existing work and then you can use it in other ways because obviously you think about gpus and the evolution of gpus in particular you know they were initially like really commercialized for gaming and then it kind of came back to more other you know parallel Computing projects I mean obviously at Stanford there was like folding at home and other uses for applying it to protein folding then obviously applied to Ai and artificial intelligence and now we're talking about them being hacked for use in ZK to make them scalable so anyway it's another example of that too oh for sure for sure yeah let's talk now then about how zero knowledge is playing on in other fields Dan okay so maybe I'll just give two examples uh there are many

many than one can give so one interesting example is goes in the area of what's called Blind signatures and Anonymous credentials so here there's a company called Spruce ID that's taking California driver's licenses and making them available basically as a digital object now why would you want to do this or what could you do with such a thing well if you remember your California driver's license has your birth date on it but now suppose you want to enter I don't know a bar or whatever that require that has an age requirement so you want to prove that you're over 18 or over 21 or whatever it is you want to prove today the way that's done is you have to show your driver's license to whoever you know is enforcing these rules and they all they need to know is just you're over 21 but actually be by

showing them your driver's license you're really showing them your birth date your your your home address and so on and so forth and so what DK proofs lets you do is take the credential that was issued to you by the state that says you know this is a valid driver's license this is a valid birth date home address and so on and all you can do is you can produce a zero knowledge proof that says I am over 21 yeah and the reason you can prove you can the reason you can trust that statement is because it came from a valid you know credential that would is issued to me by by the state or by some government so the zero knowledge proof says I have a government credential that says that my date of birth is such and such but the statement is I am over 21 yeah and so the only

thing the verifier gets to see is I am over 21 and they see nothing else about my data right and by the way this is not just Theory the California you you know Department of Motor Vehicles the DMV is actually actively looking at using this yeah there in fact there are a couple of states that are that are looking at digitizing their credentials and one of the benefits of doing that is now we can have these Anonymous credentials that let you prove exactly what you need to prove without revealing the entire credential to whoever it is that needs to verify them great the only reason this is possible is because ZK proof systems have gotten so good that now they can be used to prove complicated statements like you know I have a credential that says that I am over 20 one so that's one example the second

example I'll give actually has to do with disinformation it's one of my favorite examples which is basically using ZK proofs to fight disinformation on the web the problem that we see out there is that there are newspaper articles that contain pictures in them and the problem is that people write a newspaper article about one thing let's say the war in Ukraine and then they use pictures from a completely different event yeah maybe some other war from 5 years earlier and so you know how do we make sure that the pictures that you're looking at in a newspaper article actually are from the time and place that the article is claiming that they're from and so the journalist Community has been worried about this quite a bit and so they developed a standard it's called the ctpa standard

contents Providence and authenticity which basically makes it so that when a camera takes a picture it adds a signature to the picture and the signature is a signature over the location and time where the picture actually took place yeah and so so every every jpeg will now have its metadata yeah the location and time and and so on along with a digital sign signature by the camera to say that it was really taken at this at that location yeah so that's the ctpa standard and in principle the newspaper could just embed the signed image in the article and now ver browsers can just verify that the signature is valid and they would know that the image really does belong to that particular newspaper article the problem with this is that that's not

what newspapers do yeah we edit and yeah yeah exactly so newspapers will go and crop the image they will go and downsize it you know images that come out of cameras actually are quite large in fact uh the camera as to do this Sony has a camera that embeds these c2p signatures and images it produces a 30 megapixel picture which is too big for newspapers to use so what the newspapers will do is they'll down sample the image and send you a much a smaller version of the image the problem of course is now that you the user have only received a down sampled version of the image you can no longer verify the signature right to verify a signature you need the original the original data but you don't have it anymore and so the solution of course is a zero knowledge proof so what the

newspaper will do is it will downsample the image and then attach a succinct zero knowledge proof to say that the modified image actually came from a properly signed image that was taken by one of these ctpa cameras yeah and now the the browser on the user side will just verify that the proof is correct rather than that the signature is correct and again the proof is a synct so it adds very little data to the image and it's it's very fast to verify it's just a few milliseconds to verify and so it takes no time for the browser to actually do this and again I want to stress that the only reason this is possible is because the web 3 Community put so much effort into developing practical ZK systems yeah 10 years ago this would have been science fiction

there was no way we could have produced a zero knowledge proof over a 30 megapixel image it's just not would have not been possible whereas to today it's actually very practical and in fact newspapers can use this uh whenever they want I'm really glad you brought up that this would have been science fiction 10 years ago because for those who are not in this field if you go back to our Canon you'll see like exactly all the advances and all the many many many steps that I've gone into that which is kind of nothing short of astonishing okay so then one more quick note on this before we move on from ZK what was really great about the example you just shared on misinformation Dan is that something that honestly not just journalists worry about but politicians worry about it policy makers worry about it any public figure actors actresses

people who are just in public teachers radio personalities it can be anybody who cares about this because you want to know that you know images of you and and narratives are not fake so it's really important on that level I want to also add just one quick note on ZK applied to elicit Finance you actually helped co-author a paper with Joey bures on our team and Michelle corver on privacy protecting regulatory Solutions using zero knowledge proofs and at a high level what's super fascinating about about that is that traditionally a lot of illicit Finance world people think of privacy as a really quote bad thing because you can hide Secrets but there's actually good privacy that you want to preserve and what's really great about ZK is that it can allow safeguards for privacy while still allowing the

requisite access to information yeah the way I like to say it is that it seems like if we have private transactions then it's very difficult to enforce compliance so if you don't know about ZK it looks like privacy and compliance are at odds with one another yeah but in fact ZK allows us to square that Circle and with ZK we can have privacy and prove that we are compliant with local laws actually quite efficiently and quite easily yeah privacy preserving compliance that's exactly what ZK is privacy preserving compliance we can have both without any any problems great I'll link to that paper in the show notes as well cool so let me um add another example in general I'm going to be adding examples more on the side of economics and and incentives Less on the

side of cryptography and I wanted to talk about mechanism design which includes auction design as a special case so what is mechanism design it's often described as inverse Game Theory so in game theory in the spirit of Bon noyman and John Nash a game just sort of falls from the sky and you study it right so you say oh do players have dominant strategies oh what are the Nash equilibria those are the kind of questions you ask in Game Theory and mechanism design even though it comes from economics one really has one's engineers's hat on when when you think that way you really say look you know there's this outcome I actually really want to happen like maybe it's like a scarce resource and you want it to be allocated efficiently maybe you're developing a protocol and you want everybody running the protocol to run it

in the way that you're hoping they're going to run it so you start from an outcome you want to see happen and then the question is how do you back out how do you reverse engineer a game such that that desired outcome is in equilibrium so that's kind of mechanism design broadly like a special case would just be like an option that you see on eBay or a sort of SES like that one item for sale that's a scarce resource only one person can have it you would like it to go to the person who values it the most and so if you don't know who that is in advance you run an auction to sort of figure out who it is that vales it the most who'd be willing to pay the most to acquire that item so you know mechanism design been around forever you know if

you've heard of a second price or vicory auction you know from 1961 so we're talking over 60 years very rich history very rich literature in economics right why would you ever care about that you know If you're sort of Designing blockchain protocols and applications on top of them well actually there's there's a lot of reasons they come up all the time so they come up at the application layer like if you're doing an nft mint right maybe it's a limited collection and that's really kind of tailor made for running an auction but it's not just at the application layer and this is one of the things that's so fun about doing mechanism design in web 3 is there's also really interesting mechanism design questions all the way at the lowest layers all the way at the consensus layer so for a blockchain like let's take ethereum as an example the

the scarce resource there is just ethereum block space so just computation in the ethereum virtual machine it's a scarce resource lots of transactions would like to get executed and there is a room for all of them so again you need some kind of mechanism to decide which ones get in which ones are left out and for the ones that get in what is it that they're going to pay for that privilege so that's mechanism design broadly that's why you care you have to care about mechanism design you know if you're sort of if you're trying to design sort of good robust blockchain protocols and applications on top of them so next question is okay cool we need mechanism design I just said there was like a 60-year history and like lots of amazing textbooks on mechanism design so you know I guess it's a solve problem

I guess we'll just read the textbook and we'll just sort of copy paste the standard Solutions and we'll just sort of deploy that on the blockchain and and then we'll be we'll be done right yeah right and that's a fine first reaction like you definely want to try that first right because sometimes it works and you don't have to work any harder than that totally fails trying to Port just traditional mechanism design over over to the blockchain setting really it's initiating a kind of I think quiet revolution of sorts around the evolution of mechanism design I want to talk about some aspects of that so why is it hard what are the obstacles that come up so let me just highlight two of them one concerning trust assumptions around the person running the mechanism like an auctioner and secondly I want to talk a little bit about about collusion so

pretty much almost the entire history of mechanism design the Assumption has always been that you know the mechanism that you're participating in is known and cannot be changed after the fact so if you're participating in a second price auction the auctioneer has somehow credibly committed to the fact that they will indeed award the item to the highest bidder that they will indeed only charge the second highest price to the winning bidder because that's how we've always thought about second price auctions so so literally to the point that like there's a great and I'm happy to say very influential already paper that was published in econometrica just in 2020 not motivated by blockchain so I actually think blockchains are a better

application than anything they actually mentioned in that paper this is a paper by Muhammad Akbar and shangu Lee of Stanford and Harvard respectively so they introduced the idea of a credible auction which is auctions where the auctioneer actually is motivated to carry out the mechanism as as stated that was three years ago that started to become a thing over in the economics literature but in the blockchain context you can't not solve this problem you can't not worry about this issue right so let me maybe talk a little bit more about why that is because if you're just also pattern matching it may even sound kind of weird actually what I just said that somehow like the trust assumptions are like not okay for the blockchain setting because yeah because you might also say wait a minute but like I thought the whole point of the

blockchain protocol I thought the point of this computer in the sky is you didn't really need to trust anybody thought that was the point right right and that's sort of true right so you know when you have decentralization you have permissionless blockchains you're not trusting any one person any one entity you're sort of trusting a population on average right you're trusting that a suitably like large constant fraction of the hash rate or the stake or whatever is running the protocol as intended you don't you don't know who are the honest ones who are the dishonest ones you're just trusting that the population as a whole is more honest than not so that sounds like it's just make mechanism design easier the trust issues easier but the big irony at least for you know the biggest blockchain protocols like Bitcoin ethereum the big irony is that at the moment at which a

new block of transactions is chosen for processing at that moment in time there is actually a single individual with unilateral control over that decision right it would be a minor in Bitcoin it would be a validator or proposer in ethereum and so on there's going to be a time you know where where they went a proof of work lottery in Bitcoin where they're just sort of you know chosen as the current slot leader in ethereum where they get to pick exactly which which transactions enjoy the scarce resource of that blockchain's block space so that is why if you're running an auction at consensus layer to auction off block space that is why you cannot trust the auctioneer it is fundamentally an untrusted person that that's carrying out the mechanism interesting I never

heard it that way that's great so that's so that's what I want to say about so again traditionally mechanisms on you always trust the you know you always trust that the mechanism is being run as intended you cannot generally do that in blockchain applications second thing I want to discuss is collusion and so for starters you can think about collusion among different participants in the same auction and here and this may surprise you now again in the mechanism design literature most I'd say an overwhelming majority of the papers basically ignore collusion assume it away the focus is always on unilateral deviations by single participants right you're worried about a single individual misrepresenting what they want to influence the mechanism outcome like maybe I underbid to try to get a sort of cheaper price for for

winning the auction that's the traditional concern unilateral deviations not coalitions of people acting in concert to manipulate the mechanism now it's not I don't want too sweeping a statement there are certain exclusion has been studied many times in the economics literature but I do want to emphasize it is it is a minority why do I say that may surprise you well you know if you think about sort of traditional applications right sort of you know auctions being sold I don't know for wireless Spectrum licenses or whatever it is like you could definitely imagine collusion happening there as well seems like that would have been a concern in traditional applications also and it is I would say there's two reasons why you know when they actually talk about mechanism design they don't

address it uh reason number one is just that sometimes collusion in the real world is hard to pull off especially if you have things like you know side payments that are contingent on the outcome of the auction it's not trivial to actually carry that out in practice whereas in a blockchain setting if you're working in a smart contract environment this is where the untrusted nature of the blockchain actually makes your life difficult in a different way which is through smart contracts you actually can programmatically arrange for example side side payments contention on an auction outcome you can make it you have a credible commitment to that in in advance of the auction so one reason that's just sort of it becomes more important in the blockchain setting is just it's potentially easier to carry out pollustion than in traditional settings but the other way you know so I've talked to mechanism

designers that have spent a lot of time in the trenches either sort of advising people how it deid in Spectrum auctions or designing those auctions online advertising all that stuff and eventually I just started really pushing them I was like why don't you worry about collusion like why are you not dealing with it and they basic said Tim look okay so first of all if you actually handle collusion inside the mechanism okay if you if you sort of are robust not just to unilateral deviations but deviations by groups it just it ties your hands too tightly like the space of auctions you could think about deploying just becomes really small okay and so you give up tremendous Freedom furthermore in traditional applications you make participants sign a contract promising that they won't collude catch

them colluding you bust them maybe you find them or you hand it over to the authorities or what have you this collusion in the real world go unobserved sometimes of course but there is sort of the rule of law which is obviously not part of the mechanism design that's like a completely separate like legal and enforcement Channel which is sort of taking care of that part of the job and so again moving back to the blockchain setting it would be inaccurate to say there's no rule of law around blockchain protocols we watch people get prosecuted for crimes on the blockchain of all the time right but it does seem safe to say you can't rely on it to the extent that you can in traditional appli so in other words it's not that ction wasn't there in traditional applications but it's just you could kind of handwave it away

traditional applications in a way that you cannot applications in the blockchain so those are two ways why mechanism design is harder fundamentally for blockchain applications than traditional ones and as we've watched mechanism design on blockchains evolve we are seeing new designs that really we haven't seen in traditional applications because they're kind of born in The Crucible these sort of extra constraints right that you kind of have to deal with an untrusted Auctioneer you have to deal with the possibility of collusion so exhibit a for me here is would be EIP 1559 which is the name of the transaction fee mechanism used in ethereum not when it launched when it launched it sort of just copied what Bitcoin does something known as a first

price auction but since August of 2021 I believe it is ethereum has used this really quite sophisticated transaction Fe mechanism known as known as e559 and it is deliberately designed so that you don't have to trust the validator that's carrying it out it is deliberately designed so the collusion does not help you cannot actually improve your lot using collusion and those extra constraints I mean they really made the design interesting in a couple ways so for example famously EIP 1559 Burns some of the transaction fees that it collects the obvious thing to do with transaction fees and what is done in Bitcoin and what was done in ethereum up until 2021 the obvious thing is to give the fees from the users to the person to The

Entity that chose those transactions rather than other ones right you're like why is my transaction executed on the blockchain well it's because this particular minor rev validator put it in their block they're the ones I owe some money to so they should just get they should get what I'm willing to pay that's the obvious way of doing it that doesn't work okay if you if you try to have what's known as a base fee if you try to have a reserve price which is what e559 does you actually cannot pass those base fee revenues onto the producer of The Block exactly ly because collusion can be used to evade um That Base fee and basically render it kind of irrelevant and so it's really those kind of collusion considerations that naturally guided the EIP 1559 design to this very unusual choice of burning some of the

transaction fees which then actually has really interesting macroeconomic implications you know which I'll talk later now to this point you know e 1559 it's used in ethereum it's used in several other blockchains I don't know of a deployment yet outside of the blockchain space but there is no reason why you couldn't use this in certain traditional applications as well it would need to be an application where you have auctions happening repeatedly you know but as our world gets more and more digital that's more and more commonly the case like online advertising auctions certainly you're done repeatedly because e559 needs to sort of adjust its base fee over time based on the results of past auctions um so I think this could be an example where a mechanism which no one would have ever thought of but for the unique constraints of the blockchain

environment produces a novel design that could then feed back into non web3 applications as well yeah and just by the way for the listeners context because you sort of skimmed over this part which is you're the one who authored the economic analysis of EIP 1559 I think the ethereum foundation had asked you to look into that if I'm not mistaken as well as the paper on the transaction fee mechanism design right that that's right so the design was done by ballic butterin sort of evolved over the years sort of 2016 2017 2018 the commity was discussing it in Earnest starting in 2018 there was a lot of you know it was a it was a sort of polarizing proposal it was a pretty major change to ethereum and so my role was to you know provide a very precise description of exactly what

the mechanism is And to clarify exactly what kinds of incen of compatibility properties it has and doesn't have right and I'm told at least anecdotally that that that definitely gave the Ean Community some more comfort with making the change actually that paper really explained what the mechanism is trying to achieve what are the goals that it's trying to achieve why it's designed the way it is so personally that paper kind of illuminated that whole design for me so it was very highly recommended paper to read I I'll link to it in the show notes yeah actually I think the 1559 mechanism has been running now for two years and people are fairly happy with how it's working it's working the way people intended it to work that's my understanding yes and and even and the merge has only made it better the deterministic block times in proof of

work you have stochastic block times but it's DET terministic in the proof of stake version which is actually very helpful for keeping the Bas fee Evolution smoother it's even just working more smoothly post merge so that's fantastic oh that's so cool so one kind of highle take away from what you just shared Tim is that basically blockchains are just a massive constraints and solving for that mass that complex mess of constraints can actually unlock other applications whether not literally on chain like in blockchains it could be online or in other untrusted or distributed setups where you have people that don't know each other that are kind of collaborating or working together trying to do something so I would say that's an example where web 3 is sort of a

playground where people can experiment with these ideas show that they work in a real life settings and now they can be used in in other areas that have nothing to do with with web 3 so why don't we talk about Dows and just sort of the evolution of governance so da basically stands for a decentralized organization or decentralized autonomous organization but really we think of them as decentralized organizations and it really it's basically a way for a bunch of people to come together and decide on the evolution of a certain project maybe decide on on how they want to spend the resources together the goal of D governance is to kind of benefit from the wisdom of the crowds to make uh correct decisions in in whatever areas we need to make uh decisions in and so DS have been around for a while you know

they're working in the web 3 environment to govern protocols and there are many different types of dows but let me start by saying something a little a little crazy oh good I want to hear this crazy is good yeah yeah and then we'll then we'll see how it relates to Dows okay so you know for our listeners who are like in their 20s right now it's it's not out of the question that by the time that your kids join the workforce or maybe your grandkids by the time they join they join the workforce you know anything that a person can do a machine can do better yeah it's not out of the question that in a generation maybe two generations will end up in that kind of environment whether that's going to happen or not knows but let's pretend

for a second that that's that's where things are headed which would say that in a world like that because of these developments we might need new governance mechanisms New Economic mechanisms and so on so how do you run a world where anything you can do a machine can do better and so how do we experiment with new governance mechanisms well that's exactly where Dows come into play kind of like Tim was saying with EAP 15559 the web 3 environment is a playground for playing with this economic mechanism Dow governance is a playground from playing with governance mechanis Ms so if we have a governance mechanism that we want to try out we can just implement it in code build a dow around it and just deploy it and see how well it works for governing a particular protocol or governing a particular application or or

activity or so on so that's kind of the power and the interest of dows so let's talk a little bit about the Journey of dows where they came from and and where they're headed so at the moment Dows are still learning from Real World governance Dows sort of started by doing what we call direct democracy where all the Dow participants would have to vote on every single proposal and even the Greeks already learned that direct democracy doesn't really work people tend to not participate you know I can't devote all my time just to uh voting on decisions and and so people ended up not voting and you see that in the direct democracy Dows participation rates are actually quite low so then Dows started moving towards representative democracy where you delegate to a delegate and then the delegate votes on your behalf

that's exactly what we call a representative democracy and more recently Dows are starting to move towards private elections uh soos is a very famous example of the nouns da that um is developing a private voting mechanism because they realize that public voting is problematic too there are all sorts of game theoretic issues aspects that come up when you do public voting so they wanted to move to private voting it's worthwhile mentioning the Sika project that's another approach to doing private voting and so that was up until now they're kind of learning from you know the 10,000 years of governance Evolution that Humanity has already gone through right we don't need to reinvent the wheel we can just learn from what Humanity already developed and so Dows are kind of kind of quickly coming up to

speed to where we are today what's interesting is if we look forward Dows are starting to develop Now new governance mechanisms that are not available in the real world and so one example is what's called a liquid democracy where literally at any given point you can actually move your your voting delegation to somebody else if your delegate voted the way you you don't like and so unlike in the real world where you have to wait four years until you can vote for somebody else in a liquid democracy you can kind of actively move your delegates around as as you want so there's there are experiments with liquid democracy in the da world and you know we'll see whether that's a good thing or a bad thing you could argue both ways but the interesting thing is Dows give us a playground in which we can experiment

with these kind of governance mechanisms another mechanism that's quite famous is quadratic voting right where you basically the more you want to express support for a particular proposal you know the the more it actually the more costs you in votes that you have to spend so this called quadratic voting where if you want to vote for a proposal key times in some sense you have to spend K squared votes in order to do it so again there are Dows that are experimenting with that with that idea and we're going to see how well that works so the point is that Dows even though it's a web three mechanism you should be thinking of it as a playground for experimenting with new governance proposals and today these Dows are primarily still learning from what governance has learned for the last 10,000 years but as things progress

these dials will hit on better governance mechanisms and maybe the real world can learn from how these Dows actually govern themselves this is actually really compelling and Andy Hall one of our researchers talks a lot about this where you can borrow so much from the traditional history of governance and then see how that's been applied to Dows and then what is also adapted for Dows and then what is then natively created newly uniquely possible made by Dows given the online onchain context so I love how you sort of painted that art kind of going back and forth and now it could go back to the other direction like back to the real quote in real life world what's also really interesting is that we like to think of it more as a laboratory you say playground which I think could kind of sound frivolous which I know that's not what you mean

it's sort of a a really experimental pot is what really what you're saying when you say playground but I really like how the team here thinks about it a bit more as a laboratory for governance because the idea is that you can essentially accelerate all these experiments as you said and and do things that normally would play out over hundreds if not thousands of years and you know every 500 years or so and you can actually play them out in five years or 10 years or one year even so it's kind of great and the other quick note I want to make to Dan on your point just to kind of share how broadly applicable this is is that we're not only talking about like Civic governance when people think of governance they think of government politics but there's also thinking about corporate style governance like how

might corporate governance itself evolve for modern corporations so I think that's another point to bring up it's not just politics and political systems yeah AB absolutely anytime there's a group of people who need to work together there's there's a need for a governance mechanism and yes as we say these Dows are kind of a laboratory for experimenting with new with new governance mechanisms so people who are in political science for them Dows are like a godsend well that's why our guy Andy Hall is like obsessed with it because he's I think a classic polyi professor and danen is teaching a whole course jointly with them too so oh I didn't know that indeed Yeah by now there's enough material on Dows that we actually taught an entire corong course just on on Dows that's super cool great all right so let's segue to our next one

yeah so on the one hand what I want to talk about next is more about mechanism design on the other hand is very much A variation on the on the theme Dan introduced about sort of web three as a as a laboratory and you know he talked about the laboratory for different systems of governments and voting and so on I want to talk a little bit about a laboratory for different macroeconomic policies specifically different monetary policies now why does that even come up why would that naturally happen when you're talking about blockchain protocols and so here let me tie it back to the to the mechanism design story that I started telling earlier another incredibly unique aspect I think you know genuinely utterly new aspect of mechanism design in a blockchain context is the fact that you have a bigger design space in the following sense you

at least if you're working in a blockchain protocol that has a native currency as all the major ones do mechanism designed at the consensus layer has direct access to the monetary Supply how would you use that well think about Bitcoin right since day one right so Bitcoin miners are expected to solve proof of work puzzles in order to publish blocks like why would more than a handful of enthusiasts ever do that well nakamoto's idea was to have incentives right so every time you publish a block and it gets on the longest chain you will get some number of Bitcoins originally 50 now it's down to six and a quarter where do those Bitcoins come from well that is newly printed money okay so really Bitcoin black production is incentivized quite deliberately through inflation through

money Printing and so the black reward you can think of as corresponding to an inflation rate for the Bitcoin uh currency as it happens Nakamoto pre-programs an inflation rate for the rest of time into the protocol you could you could imagine other protocols where it evolves sort of more endogenously and in fact it does these days in ethereum but in any case the microeconomic decision of incentivizing reduction through block rewards corresponds to a specific choice of a macroeconomic policy a monetary policy of choosing an inflation rate for the currency to give another example right we talked about EIP 1559 I talked about how sort of game theoretic considerations forced the design into burning a share of the

transaction fees and by burning I mean literally taking those coins permanently out of circulation and this right is going in the opposite direction of the Bitcoin block rewards this is a form of deflation this is decreasing the supply of money now ethereum also does have inflationary rewards for validators carrying out the protocol correctly so you've got this sort of tug OFW in ethereum ever since EIP 15559 where on the one hand there's inflationary forces that are sort of a you know kind of a carrot for validators to do the right thing the other hand there's this natural deflation coming from just excess demand for for ethereum Block space and it varies I mean so we now have a couple years of data on this there's been periods where overall the currency's been inflation there's been

periods where overall it's been deflationary so that means Bitcoins and eth are now kind of in opposite directions one is a positive inflation rate one is a negative inflation rate and you could ask is one of those better than the other I mean if nothing else we are actually carrying out a natural experiment where we will get sort of real time data about which of these monetary policies what what are the different properties of them and Akin to you know Dan talking about you know on the governance side right yes we've learned over the course of history stuff about governance through trial and error but the time scale at which that that unfolds is very long right you'd look you know years decades centuries you'd look at sort of how different you know civilizations or cities or what have you would sort of rise or fall and how much

could that be attributed to the method of governance same thing here right obviously we have data points on different countries with different macroeconomic policies monetary policies we can sort of observe how they unfolded but this is a completely different level of you know both active experimentation that we're in a position to do right you can literally like build new projects monetary policies to to explore and then also just the time scale at which you can iterate is obviously just you know orders of magnitude faster so so for example one thing I view is basically an open research question now is right so let's suppose ethereum the currency is going to be deflationary for a while let's say for the next decade or something like that is that good or bad and I've asked a lot of people this

question and I get very strong answers that very much much depend on who I'm asking right if I ask somebody who's a very proud eth holder they are extremely in favor of the deflationary forces that e559 gives rise to when I talk to any classically trained macroeconomist yeah they totally freak out they start telling me stories about like stagflation in Japan in the 1990s yeah right now like I said I think it's an open question who's right I don't know I'm not super convinced you know by the traditionally trained macroeconomists making the analogy between you know a cryptocurrency of a protocol that has

all this economic activity on top I don't know that that's so much like the fiat currency of a Sovereign Nation but the traditional story would be you know you don't want deflation because then basically you know the the fiat currency actually becomes like an investment device whereas what you really wanted to do is you wanted to sort of lubricate economic activity on the other hand like in ethereum I'm not sure there's a contradiction between you know eth while at the same time having lots of us denominated economic activity happening at the application layer I'm not sure there's any kind of contradiction there but totally open right so sort of The Gauntlet has sort of been thrown to us as blockchain researchers to explain you know why we should be okay potentially with with deflationary cryptocurrencies in in our major blockchain protocols

that's great by the way have you talked to Alex tabarak about this not this specifically he he's coming to visit us in a couple weeks okay great yeah so I'll bring it up with him then so yeah because you know obviously Tyler Cohen and Alex wrote a textbook book their or they updated their classic textbook on the modern principles of economics with a whole new chapter and sort of focus on examining crypto through that lens of economic reasoning so it'll be very interesting to see what they say as well by the way what what is interesting is that like in the real world it would be very difficult to run an experiment like this whereas here web 3 just provides like a natural environment for this experiment yeah yeah and we're going to learn whe however whichever way it goes we're going to learn something

interesting from it yeah though that's not to say that like these experiments have consequences because there are real people who are you know transacting and interacting in these systems but anyway take it away Tim okay great so let me next go to to yet another example of this um kind of web 3 is a laboratory experiment this time I'm focusing more on the finance side so specifically let let me talk a little bit about exchanges so this is somewhere where you would buy or sell one asset in exchange um for another obviously not a new idea right we're all familiar with the stock market and sort of other types of exchanges of that form sort of in real life but exchanges play a really important role in blockchains and in web 3 as well so specifically you might be familiar with automated market

makers you know like Unis swap being economical example which lets you swap one type of token like maybe eth for another type of token like maybe usdc so I want to talk a little bit about you know where amms come from in particular like how come like when you hear about the stock exchange you never hear about amm how come you only hear about what you might know of as limit order books but why you always hear about amm instead in web 3 so like what's the deal with that so a little history right so so limit order books were sort of the you know standard I'd say you know by the late 20th century of of how to trade fungible you know tokens and assets and so why wasn't that adopted immediately in the blockchain setting why not copy paste the sort of agreed upon solution

from tradition applications over to the blockchain side and the answer is actually quite interesting which is computational scarcity I mentioned briefly this idea of kind of you know blockchain's you know an abst raction for a blockchain is sort of a computer in the sky you know Open Access that anyone can program and run programs on and an issue is that you know this is changing we're making improvements but historically that computer in the sky has been very slow about as slow as computers maybe from the 1950s or something like that you know 10 transactions a second 20 transactions a second something like that so computation is super super scarce now in economics there is usually been very little attention paid to computational resources okay the really all the focus in the economic literature has been on

the economic properties of different designs right so there's some really egregious examples actually mechanism design like the something known as The vicory Clark Groves mechanism or vcg mechanism which is a generalization of the second price or vicory auction which shows that you know basically welfare maximization is a solved problem but then you actually look into the details and they're assuming you can solve empty hard problems and all these other kinds of things and so it's really only once computer scientists got a seat at the table around the turn of this Century that the shift focused more and like well no no we want all the good economic properties but we want it to also terminate like before the end of the universe we want we want sort of computationally efficient and you know economically kind of inent compatible mechanisms so going back to amm this is

a very extreme form of a computational constraint right so not only can you not solve EMP hard problems in the blockchain you also can't run you know quadratic time algorithms in the blockchain it's very really very limited so you need super simple computations at least for now again we're seeing this change as the years go by and order books you know if I you know if I explain them to you they'd sound really simple right you've got these like outstanding buy orders you've got these outstanding sell orders whenever you have a match you just match them everybody goes home Happy what's the big deal sounds like something you could easily cat up and in some sense you could but there is this matching process first of all you need to store all these orders right and storage on blockchains is actually at the moment quite

expensive plus you have to do a non-trivial computation to check whether any sort of new orders match any of the old ones and that was just too much for the early blockchain protocols but but people really wanted to swap you know one type of fungible token for another and so that led them to this alternative design of automated market makers and you know most famously what's known as the XY equals K design adopted by the original version of Unis swap now interestingly automated market makers did exist before the Advent of blockchain Technologies not for trading tokens not for swaps but rather for prediction markets where you have an event you're trying to predict like maybe the outcome that Pres presidential election and you want to elicit people's

beliefs about what they think's going to happen what's the likelihood that the winner will be a Democrat versus the winner will be a republican so prediction markets you know the usual way you set them up is you invent a so-called aerody debru Securities so you'll make up two sort of fictitious um kinds of equities one of which pays off a dollar if the Democrat wins one of which pays off a dollar if the Republican wins you let people freely trade them you look at the current market price and you interpret the market price uh as a probability as a likelihood of who's going to win that election in some ways it's a it's a way that a market can sort of aggregate the beliefs um of a large crowd of people so prediction markets literally are exchange mechanisms in the same sense of

the stock exchange and there are prediction markets that use limit order books okay that is one thing that people do do but one place where limit order books really break down is in what are known as thin markets so markets that do not have a lot of Traders okay for the extreme case you know imagine somebody shows up wanting to you know buy and sell equities in order to express their extremely accurate prediction about what's going to happen but suppose nobody else shows up at the limit order book right then there's nobody to match with right so literally no trades will get executed because you don't have a counterparty and so that's the sense in which like economic activity sort of dies in a limit order book if you don't have a a thick Market if you don't have a lot of people sort of arriving being

willing to trade and in prediction markets you know for some of the things you know some of the prediction markets you did have a relatively small number of Traders so the question then was could you do something different that gave rise to the idea of automated market makers in the sense that we understand it this was Robin Hansen I believe that pioneered this maybe I don't know 20 25 years ago something like that so-called scoring rule based prediction markets and so they were invented this is before this is before blockchains and they had nothing to do with computation you know if you're not on a blockchain limit order books are perfectly easy enough computationally they were invented to solve the liquidity problem the idea being that this automated Market maker will single-handedly serve as the counterparty to anyone who wants to trade with it anyone wants to buy it

it's willing to sell at the current spot price someone wants to sell it's willing to buy at the spot price so it's always ready for business and so that's the big liquidity advantage of an automated Market maker versus a limit order book so that was sort of the story of amm for a long time and you know they were kind of in you know say at the time ethereum came out in 2015 they were kind of collecting dust on the Shelf it was this tool people knew about it thought it was cool not a ton of traction you know wasn't a big part of like scientific discourse in sort of any area at that point and so many other things it was an existing idea that was just resurrected really by the needs of blockchain technology and so in particular again

you know on the blockchain side computation scarce these amms are super simple right you do not need to do some kind of matching thing or keep track of a bunch of outstanding orders you literally just have sort of two piles of coins and someone comes in in a very simple formula algebraically says you know if you deposit 10 here you're going to get back seven of these and that's it that's how the amm works so the the early versions of amm worked and that was the kind of Simplicity people really needed for the most initial and primitive versions of that that computer in the sky now these days we're seeing a lot of innovation so now that amms have become the dominant way of doing exchange of fundable tokens on chain at least we're seeing a lot of innovation right so unisoft themselves have had

multiple versions that they've been iterating competitors are sort of you know trying to iterate as well you know for example people trying to be people are now trying to sort of programmatically sort of adjust trading fees according to sort of the recent volatility of the assets being traded those are some of sort of the nice things that people are looking at now and the way this Innovation is going we will start winding up with I think sort of exchange designs that first of all we never would have come up with but for this path through this computation scarce environment but like with the f559 after a bunch of smart people do a bunch of hard work and come up with something clever at the end of that path we actually have new exchange designs which might well be better suited than

the than the incumbents for some of the traditional applications of exchange you know here I zoomed in on amms but I mean zooming out the broader theme here is just again very much like what Dan was saying about this completely new ability to innovate you know back when he was talking about sort of governance systems of governance you know innovating in traditional Finance is very very difficult the incumbents are extremely powerful whereas again in web 3 like we're iterating you know every couple of months you'll see like a new version of sort of some of the most popular projects so there's just this tremendous freedom for people to explore uh just to give one other example um would be uh batch auctions so for example for for at least a decade there's been kind of a

movement to try to take um something like the New York Stock Exchange and have them settle trades in batches maybe one batch per second as opposed to what they do now which is just sort of a continuous double auction you just always match things immediately as they come in why would you want to just you know have batches of a second well it makes highfrequency trading sort of a lot less profitable right so sort of having a microsecond of of informational Advantage doesn't wind up being that useful if you're almost certainly going to wind up in the same batch with other people that got it you know 10 milliseconds after you did needless to say the New York Saka sh has not adopted batch auctions there's some really nice Theory and kind of lab experiments around it I think it's really sort of great work whereas if you take something

like ethereum literally every 12 seconds a new block is being published it literally is by definition a batch auction every 12 seconds and there are some really nice projects out there in the space that are actually trying to design novel mechanisms um to take advantage of the sort of intrinsic batching property that that ethereum has and again you know I'm hoping this to some extent like normalizes the idea of batch auctions to a broader audience potentially to feed back into to traditional applications that's fantastic I think that the theme that's emerging from the discussion so far is that like web 3 is a really good laboratory for the rest of the world and so I wanted to give one more example where web 3 technology is going to be applicable to the rest of the world and

that's in this environment of what's called a trusted execution environment a t a te so let's first of all Define what is a trusted execution environment ask you what is that right so it's basically an isolated environment that runs on your computer yeah so you can run literally a job inside of a trusted executioner environment inside of a te the data that will be running inside of the te will only be available to a code that's running inside of the te it's not going to be visible to the rest of the computer it's not going to be visible to the operator of the computer it will only live inside of the te so the the vision for this is you know if you run a want to run a task in the cloud in some sense today you're at the mercy of the cloud operator right the administrator

of the cloud can in principle look inside of your the data that you running in the cloud look in inside of the code that you're running in the cloud and learn about your business if you wanted to if you ran your Cloud application inside of a te then basically the administrator is locked out there's nothing that uh anyone in the cloud can do to see your data to see the the the code that you're running in the cloud and so it's a strong privacy mechanism for running jobs in the cloud yeah so tees are an important mechanism and they've been around for a while in fact one of the first tees I guess they go back in history quite a bit one of the kind of the large deployments of T tees was actually started by Intel in what's called uh the sgx so it's about a decade

old they deployed it and were hoping that people would find applications for it unfortunately what happened next is that there were all a whole bunch of weaknesses found in in in the sgx implementation so even though t te are a really good idea there's a there's a bit of a fear out there because of some weaknesses that were done that were discovered on this very first uh popular implementation but you know tees are out there and what's interesting is that now it looks like tees are finding applications in the web 3 environments and so the reason I'm excited about this is that tees are really useful if we could run jobs in the cloud inside of a te we would get much more much better privacy for cloud customers so tees are

really useful for all of humanity but now that the web 3 environment is actually finding applications for Te I think we're going to see what we saw with zero knowledge proofs that all of a sudden a whole bunch of smart Engineers are going to start developing new new designs for tees better designs faster and more secure we're going to see prizes around the designs for tees so there's like in the coming years there's going to be a lot of activity in better te design driven by this uh web 3 interest so why is the web 3 community community interested in tees and the answer actually actually has to do with uh something that stim already touched on which is me yeah maximal extractable value which I think T Tim said it very

nicely it's basically whoever is making the blocks has a lot of power because they decide which transactions go into the block and what order those transactions are in the block yet people are willing to pay to get their transactions early in the block and that translates into this large field called meev in the past has caused a whole bunch of problems on the blockchain so there are a bunch of proposals or solutions for trying to at least if we can't eliminate M at least let let's not make it cause problems for for the rest of the web 3 users and so one proposal is called me boost this is something that is already deployed it's an implementation of what's called proposal Builder separation PBS I imagine many of

your listeners have already heard of those of those Concepts but there's a new version that's coming and that's called me Suave I'll also link by the way in the show notes to barnab monotes post on notes on proposer Builder separation because that has a great explanation so that'll be in the show notes keep going Dan yeah perfect perfect yes so what is the idea behind Suave so if you step back for a minute and you realize in the world of meev there's a couple of actors so one category of actors are called Searchers what they do is they basically look for opportunities where they can issue transactions that will make direct profit direct profit to the Searcher maybe there's some liquidation opportunities the Searcher finds a liquidation opportunity issues a transaction and they get some profit as

a result so Searchers basically have transactions they spent a lot of work they found these transactions and now they want to deploy them on the Chain there are Builders these are the people who actually build blocks out of uh transactions and sometimes call bundles that come from the Searchers so we have Searchers who are making transactions we have Builders who build transactions and then there are the block proposers who actually take blocks built by the Builders and actually submit them on chain yeah so again there are a couple more parties involved but at the high level these are the three that we might want to focus on the Searchers who have transactions the builders who use all sorts of strategies to build blocks that maximize you know certain utility and the block proposers who actually signed those blocks and put them on the Chain so in sve what happens is we're going to

treat these three parties as members of what's called a large confidential comput so the Searchers have transactions that they don't want to tell the world what the transactions are because people could just steal their transactions so they have these transactions the builders have building strategies they don't want to tell the world what the building strategies are because they spent a lot of time optimizing them and they want to keep them to themselves and then of course the block proposers they have to sign the blocks and they don't want to tell the world what the signing key is so these are three parties in this space they all have secret inputs and somehow they need to all work together in order to build a signed block so everybody wants to keep the data private and until the block is available and is signed but

somehow they have to come together in order to construct a properly constructed block that is correctly signed so this is kind of the classic setting of What's called the confidential computes you know the three parties they have secret inputs but somehow they have to come together and compute on their secret inputs but none of them want to reveal their inputs to the other yeah so how do we do this and so there are basically two approaches to confidential Computing One is using what's called multi- party computation and here some of your listeners may have heard of fully homomorphic encryption FHA fhe is in some sense in this environment kind of a particular way to implement multiparty computation so we lump lump one approach as multi-party computation where we can do a large

protocol between the three parties and the Magical block just appears yeah so that's one way to do it the other way to do it actually is using trusted execution environments right so how do you do this using a trusted execution environment basically we will create an isolated environment on a processor that nobody can see inside this environment we'll have the three parties send their data encrypted using encrypted under a key where the decryption key lives inside of the te so only the te can decrypt the data that the three parties sent to it and now the te inside of it it's an isolated environment but inside of it it's able to decrypt the data that came from those three parties now it can

see all the data in the clear it can see the transaction action from The Searchers it can see the building strategies from the builders and it can see the block proposers private key and now now it can run the strategies on the transactions produce a block and then use the private key to sign the resulting block it would then output the block and erase all the data inside of the te so this is a way to do this confidential compute where the three parties can send their encrypted data to the te the te constructs the block uh destroys the data and the block is available to be deployed on the chain and nobody and if you believe that the te is secure nobody can see the data that was actually used to create this block only that's only available inside of the te yeah so this is basically a

proposal for how uh Suave is going to work that there'll be this te environment that they're building that will be used to construct these blocks and uh basically that's how block production will work under swap but what's interesting is this design really shows that the web 3 Community is becoming very interested in the security of te architectures you see if somebody can break open the te here then we have a problem right that all of a sudden transactions researchers will become available to whoever was able to break the te they would be able to steal those transactions and steal steal all the meev The Searchers were hoping for similarly with the with the uh building

strategies and similarly with the signing key that's used to sign the final block so here the te isolation property is really really quite important so what's exciting to me again is if it gets deployed using existing te technology well you know there are people in the world who specialize in Breaking tees so maybe this will get deployed there'll be you know some backlash in that it it'll be shown that the tees maybe are not as as uh strong as we had originally anticipated and then they'll be like like a z prise competition there'll be a te competition for can we build more secure tees that are cheap and fast and so on so I think it's really kind of exciting that the community is going to

build security e and the nice thing about that the beautiful thing about that is once we have security the rest of the world can benefit right you can imagine AWS and Google and Azure all those are deploying these security e so that all of their customers can then run jobs privately in the cloud Dan you already answered this question earlier when you kind of self-answered it about why now why web 3 for the kind of acceleration of te is it only because of me were there other drivers to that because it sounded like you said it was only because of you know the interest and trying to solve this problem with meev coming up that that's why this accelerated or were there other drivers for pushing and advancing te yeah that's a that's a really good question I mean

so the initial driver for it is meev actually dealing with the meev market but it turns out once we have tees that people actually trust there are a ma there are millions of applications it has tremendous benefits maybe kind of to go back to the beginning of this podcast I can tell you that if we have a trusted te zero knowledge proofs in some sense are actually say proof systems become really quite simple because you can just literally have the te check that the statement is correct and then just sign the statement and then the verifier just has to verify the signature and do nothing else yeah and so in some sense a lot if we have tees that where we actually trust the hardware there are a lot of problems that become actually

quite simple so this is as I said this has quite a lot of uh interesting applications so Dan maybe we should talk a little bit about the consensus layer too and the new protocols we've been seeing because you know distribut Computing it's this is another field that's been around you know 40 plus years but somehow again the Advent of blockchains we're seeing big Innovations in that sort of very very classic area yeah so I guess it's it's kind of the bread and butter of blockchains to develop consensus protocols and also I guess since the time of Satoshi right we see new Innovations and consensus in coming out of the blockchain space in particular for in Byzantine fall tolerance right there are a lot of new consensus protocols coming up hot stuff is one example there are a couple there are many others I think the interesting

thing there is that they're being adopted in the web 3 environment but web 2 systems also need consensus protocols right I mean web 2 the world of web 2 doesn't run on a single machine it runs on tens of thousands of machines and they also have a consensus problem so traditionally they use Tim you talk about this quite a bit right I mean traditionally they use paxos like mechanisms right which in particular are not robust to Byzantine faults only to specialized types of faults yeah right yeah exactly and they they kind of only Implement in some paxus is a complicated protocol in in to implement fully so they only Implement a part of paxos and you know the new consensus protocols that are coming out in principle even though they were designed for web 3 they

could also be used by the web 2 companies and you know they would reap the benefits of the you know better performance more you know security against as Tim was saying security against stronger faults and and L less any listeners think that Byzantine Faults Are relevant kind of only in adversarial settings like like the blockchain setting where they are indeed extremely relevant uh the concept of Byzantine fall tolerance predates blockchains by 25 years right and so the reason you might say like why were Distributing Computing researchers in the 1980s worried about Byzantine fault and not adversarial environments well it's because you know Hardware failures you could have a model of right like what happens if you know the power goes out you just stop hearing anything from

that machine and they stop hearing from you a crappy Network okay maybe some packets get dropped some messages get lost but if you have a buggy implementation on a machine what's your model for how it's then going to behave and it's just very difficult to come up with any plausible model that captures the range of behaviors that might emerge from software bugs and so as a consequence the theoreticians that were formalizing these ideas said let's take a consern conservative approach let's say that you know a software bug the machine can behave in an absolutely arbitrary way there it was almost kind of like a you know a modeling convenience whereas in blockchains it's like literally real like the nodes really may do absolutely anything if they want to bring down your protocol

but the formalism is exactly the same which is I think actually a great um Testament to just the power of the right abstraction how it just is sort of reusable across these different domain so you know like Dan was saying you know even though you can get away with a lot just thinking about you know crash faults and related Notions in web to you know why not have this stronger robustness notion if you can get it as well yeah Tim can you give me the highlevel version of what Byzantine fault tolerance is why it matters you explained that it had the long history and whatnot but I think for lot of listeners they might not actually know in detail why that matters so zooming out for a second I mean you know what is the point of the consensus protocol in the first place so you have a

distributed systems you have a bunch of machines maybe located all over the world and you want to keep them in sync right you want them all to sort of believe the same state of the world why is that hard it's hard because one you might have communication delays so you might have issues with the communication Network the other one is that maybe some of the nodes are not running the protocol correctly now if a node is not running the protocol correctly you want to ask like why is why is it not doing that right one simple reason would be just that it crashed and like in the 1970s when people were first starting to think about consensus protocols this really was probably like the biggest thing they worried about like Hardware just crashed back then much more frequently than it does now it still crashes now too so it's still very

relevant so that's a model of what can go wrong if a node is not following the protocol correctly another model would be it has a sporadic internet connection so that its messages sometimes get dropped and then Byzantine fa tolerant means you assume absolutely nothing about what that note is going to do it can deviate from the consensus protocol and absolutely arbitrary ways okay I can't like Ford signatures generally but kind of you know any kind of feasible comput it's going to do and so you like to design a protocol that still works even though you have these nodes that are doing these crazy crazy things right as long as say like two-thirds of the nodes are running it correctly the third that are just acting you know in some just nuts way they still can't get those two-thirds of the nodes out of sick okay

perfect that was fantastic very clear all right some last quick questions kind of quick lightning round Style just to wrap up okay so some recurring themes in this episode are about constraints this idea of a laboratory for Innovation this idea of constraints creating this tremendous Innovation but then also the users and experiments and and the market kind of pulling out the desire to advance things forward one question I had for you guys is if you were to tell people who are not in crypto who are curious about these advancements and they wanted to work on these things like what might you tell them for how to get involved or engaged let's assume it's a college student a person who's a worker in a tech company like a researcher in a

tech company what might you tell those profiles for of people like why they should kind of come work on some of this stuff or be interested I think one reason for the why is just you know you only get so many opportunities kind of in your life in your career to participate in a new technology so early that you can really fundamentally change its trajectory and that is exactly the opportunity available I think around blockchains and and web 3 right I mean one thing I to tell people it's kind of like if you want to do work that going to show up in a textbook maybe work in an area that doesn't have any textbooks so by definition like the work in those textbooks has not been done yet um so one I think is just to recognize what a golden opportunity it is to to and you

know and we see for example on the mechanism design side like new mechanisms sometimes get incorporated into live projects with high visibility like literally within a small number of months I mean the time to impact is just unbelievably short you know because there's no textbooks like the on-ramp to sort of getting the the shared knowledge is not as simple as it would be in like machine learning or a field like that but I will say at least the the web3 community is very good I'd say at self-documenting and in particular you know just on YouTube you sort of have to know how to look a little bit but there's a very rich collection of resources out there like people wanted to know more about ZK there's entire mukes about CK you know Dan has many lectures about that he and I have lectures out there that are about that

are about me obviously you know many other people you know not just academics but you know researches from the ethereum foundation and many other projects all have talks out there that's probably for me the the best way to get a sense for exactly where The Cutting Edge is right now that's great Dan would you say anything as well yeah so if you kind of want to get into actual doing things I think the community is actually really friendly you can really find a project that you find interesting just jump under Discord and people would be quite happy to kind of comment on your ideas if you have code you want to contribute there there'll be a lot of discussion the community is actually quite open and friendly at least that's been my experience and the other thing I'll say is that there's a new Z prise

forming if you're interested in zero knowledge for example and you want to try try your skills yeah try to try to kind of go and and participate in the Z prise competition there's a whole bunch of challenges that are part of this competition go ahead and just try to build write some code and see how well your quote performs you can submit it and see where you end up on the leaderboard who knows maybe you'll win and then you know that'll be a really great contribution that's fantastic you guys well I think that's a great note to end on thank you so much for joining this episode I really appreciate it and taking the time time thanks so much Don thanks and all this was fun thank you for listening to web 3 with A6 and Z you can find show notes with links to resources books or papers

discussed transcripts and more at a6c crypto.com this episode was produced and edited by sonal toxy that's me this episode was technically edited by our audio editor s Morris credit also to moonshot design for the art and all thanks to support from as6 andz crypto to follow more of our work and get updates resources from us and from others be sure to subscribe to our web 3 Weekly Newsletter you can find it on our website at a6c crypto.com thank you for listening and for subscribing let's go [Music]