hi everyone my name is Tim ruffgarden I'm a professor at Columbia University and also the head of research at a16z crypto this is sort of a home recorded version of a talk that I gave at the crypto Economic Security conference on November 1st uh 2022. and let me see at the outset just thanks to both don san and um Arthur Gervais uh for inviting me to give this talk and what I wanted to tell you about today for lack of a better title is some results and challenges in crypto economics uh I have to be honest I don't actually use the word crypto economics uh all that often it just sort of means such different things to different people but when Don and Arthur asked me to give this talk I thought I'd use it as an excuse to kind of ruminate for a little while you know about what are the sort of unique characteristics of the types of problems that come up in this sort of nebulous area of crypto economics and actually I'm really glad that I did because it kind of reminded me of just how unique and intellectually fascinating um the problems in this area are I thought it would be useful to start by comparing and contrasting the types of problems that come up in crypto economics with the types of problems that have traditionally come up uh in mechanism design which is a subfield of microeconomics now mechanism design one of the descriptions I like is as inverse Game Theory so what do I mean by Game Theory when Game Theory usually some game falls from the sky okay and you're just tasked with studying it right maybe it's prisoner's dilemma you know maybe it's tragedy of the commons rock paper scissors whatever and then in Game Theory you kind of investigated from a strategic perspective you say ah did any of the players have for example dominant strategies what are the Nash equilibria of the game that you're looking at in other words given a game you try to identify what are the most plausible outcomes of that game meanwhile in mechanism design you start not with the game but with the outcome and often that it's an outcome you want to see happen like all of your protocol participants sort of following the protocol as you intended them to and then starting from the outcome you reverse engineer a game for which that desired outcome is in fact an equilibrium maybe ideally an equilibrium and dominant strategies but if nothing else uh a Nash equilibrium so mechanism design has a strong engineering flavor that's probably um one of the reasons why it's been one of the strongest points of contact between computer science and economics there's even kind of a meme originating in a paper of Al Roth about mechanism design as Economist as engineer we've already mentioned one example of a sort of desired outcome you know all protocol participants following the protocol as intended another very basic one also relevant for us uh will be imagine you just have some scarce resource that you want to allocate in the most efficient possible way you want to allocate it to those who who value it the most if you've ever heard of like a second price or Victory auction or the generalization to the victory Clark Groves or vcg mechanism those are all really just games in the sense that the way is upsetting up payments they're games in which the desired outcome okay welfare maximizing allocation that scarce resources arises as the dominant strategy uh equilibrium so these are all sort of games that incentivize participants to to report to their true value for what's being allocated and then with that accurate information the mechanisms in a position to to have an efficient allocation of the scarce resources now in many not all but in many applications of mechanism design including you know these Victory auction and vcg mechanism examples uh there are payments involved there will be a sort of transfer of money uh between the participants of the mechanism and the person running the mechanism and generally in mechanism design you don't really worry much about what this money is you know it's something like you know US Dollars and you don't think much more about it now many of the key problems in crypto economics concern the following question which is what if when we are designing our mechanism we have access to a cryptocurrency to a currency native to a blockchain protocol on which that mechanism is executing so this is just a really really fascinating question I don't really know of an analog of this um in the traditional economics literature from the pre-blockchain days and this question is really interesting for two reasons um the first reason is is just enlarges the design space right so I'm just letting you do more things with mechanisms than you could before you could access a sort of native cryptocurrency you might ask you know what could you do with that power well for example you might be able to print new money meaning new coins in the native cryptocurrency or conversely take those coins permanently out of circulation right right it's like imagine a traditional application of like a victory auction where the mechanism description includes includes oh by the way just print some new US dollars to sort of complete the description right it would look nuts but this is actually Supernatural in the context of blockchain protocols that host uh cryptocurrencies the second reason this question is so interesting is because with that power for example to Mint new coins or burn old coins with that power comes responsibility in the sense that they're then macro economic consequences of your mechanism it actually affects the supply of the Native currency that it's working with think about running like a second price auction in the real world right you are not thinking about the U.S money supply right it has nothing to do with you it's totally outside of your mechanism whereas in these types of problems in crypto economics the microeconomic design aspects of the mechanism uh are just intrinsically coupled to the macroeconomic consequences of the mechanism which again is something I really just never seen it before so let's proceed to two um I think really good examples of these types of problems in crypto economics and in particular this interplay between the micro and the macroeconomics uh in mechanism design so let's go back to the beginning let's go back to Nakamoto and the Bitcoin protocol and let's examine it through a mechanism design lens so let's start with the desired outcome okay so here we'd just like everybody to follow the protocol uh as intended which means any bit every Bitcoin miner we would like to see them dutifully trying to solve their their proof of work uh crypto puzzles uh and in doing so trying to add a block extending the end of the longest chain the question then is how do we reverse engineer a game like for example um payoffs uh to the participants that incentivizes them to act in this way so the so the behaving in this way constitutes an equilibrium of that game uh Nakamoto and Bitcoin approach this and what you know is probably the most natural um possible way which is like look you know whenever a miner gets a block on the longest chain let's give them a reward um we're not interested in blocks that are off of the longest chain so it blocks off the longest chain will get uh no reward and the intuition behind that is then sort of pretty clear right so as a minor your rewards come purely from blocks that get added to the longest chain so any block that you create you want to maximize the likelihood that it's going to wind up on the longest chain and intuitively like what could you do that would be better than just appending it to the end of the longest chain such as it is right now so just one one quick digression which is that you know in a famous paper almost a decade ago by ayal and surir uh they said that actually this intuition isn't a hundred percent correct um it does incentivize miners to coordinate on the longest chain in certain regimes there are other regimes where actually um while it is true nodes are incentivized uh to some extent um to add blocks to the longest chain in an orderly fashion uh in some regimes they're incentivized even more um to behave in unintended ways deviating um from the the longest chain protocol and this you can interpret as sort of a cautionary tale of just how hard it is to get the game theory right in the design of blockchain protocols now for this idea of rewarding blocks from the longest chain for that to be meaningful it must be that those rewards have real economic value and so the question is like where is that money where is that value going to be coming from who's going to pay for those rewards that are going to bitcoin miners so this question would be difficult if not impossible to answer without access to some native currency to some cryptocurrency but of course the Bitcoin protocol does have access to its own native currency and in particular the protocol is in a position whenever a block is added to the longest chain to Simply Mint new Bitcoins currently six and a quarter Bitcoins and then that is the reward that gets directed to the creator of that block this is in fact the only way that Bitcoins have ever been um brought into circulation uh and here already you can start seeing that there are in fact macroeconomic consequences of these mechanism design decisions right so printing new money that's a form of inflation uh so when we ask the question you know if meaningful value is going to sort of creators of these blocks where is that value coming from um it's basically coming from the holders of that of that cryptocurrency so from Bitcoin holders uh in the form of a type of dilution so from my description so far right it probably just sounds like money is going to be printed in perpetuity right blocks are going to keep getting created I stood with every block new coins get minted um so it would seem that the supply of Bitcoins is always going to be increasing um but famously that is also that is not the case you could imagine a version of Bitcoin like that uh but Nakamoto famously was very intentional about setting a hard cap on the number of Bitcoins that could ever possibly be in circulation that number being 21 million and I want to emphasize this design decision in Bitcoin is really orthogonal to all of the other stuff you might have heard about right longest chain consensus proof of work civil resistance you know utxos the Bitcoin scripting language you could have all of that stuff with a different macroeconomic policy but in this independent Dimension Nakamoto uh required that there would be this hard cap on the number of Bitcoins perhaps as a reaction to the to the bank bailouts that were happening at that time uh during the Great Recession so again Nakamoto was not like logically forced into this macroeconomic policy but then after deciding upon this macroeconomic policy it logically forced other aspects of the design right given that the Bitcoin protocol never Burns Bitcoins so Bitcoins never get deliberately removed from circulation so the money supply is only increasing so if you're gonna have a hard cap of 21 million that means at some point the black rewards uh have to be going to zero and that's currently expected to happen in the year something like 21 uh 40. all right so at this point you'd be totally right to Circle back to our original question which is like okay well then what's actually our plan to sort of motivate um nodes to carry out the the Bitcoin protocol if we're not going to use black rewards I mean the vision has always been you know since the Bitcoin white paper um that you know look if Bitcoin if Bitcoin protocol is still being used you know over 100 years from now presumably it's pretty valuable um and presumably people are going to be paying non-trivial transaction fees um for the privilege of their transactions being included so the Hope was always that decreasing block rewards would be supplanted uh with increasing transaction fees it's always been obvious that if that didn't happen if the black rewards went to zero and the transaction fee stayed small it's always been clear that you know the Bitcoin protocol would um cease to be useful at least for sort of valuable transactions with a poor Economic Security there's also a much uh sort of more subtle issue which was investigated in detail in a paper six or seven years ago by Carlton at all which is that another difference between block rewards and transaction fees is that you know black rewards they're always the same every single block So currently um you know it's six and a quarter Bitcoins for every block doesn't matter if it's today's block or tomorrow's block or whatever yes every four years it changes but sort of locally um all blocks have exactly the same block reward transaction fees meanwhile can vary by an order of magnitude or more from black to block so you have high variance in transaction fees you did not have with block rewards and that actually creates some additional game theoretic issues for example imagine you're a Bitcoin miner and the last Miner that sort of produced a block cleared out the mempool there's nothing left and all of the transactions that it included had like tons of transaction fees so the last Bitcoin miner just made a killing and there's literally no transactions for you remaining at all well then you got to start wondering especially if the block reward is like pretty small like maybe instead of trying to extend that last lucky Miner that their block like I'm supposed to do maybe I'll actually try to Fork their block so basically you know replay a version of almost the same block maybe with a few less transactions but basically stealing most of their transaction fees and more generally when you have very small black rewards and high variants and transaction fees uh miners are incentivized to keep undercutting each other as opposed to extending each other's blocks and that's that's actually a real that would be a real issue now with transaction fees you can imagine you know pretty simple tweaks to bitcoin that would fix this it would be a hard Fork but it wouldn't be a dramatic hard Fork um for example you can imagine smoothing transaction fees over many blocks so maybe when you mine a block you would get one percent of the transaction fees of that block plus you'd get one percent a one percent cut of the transaction fees of each of the next 99 blocks as well then you know it wouldn't be a perfectly constant reward from block to block but it would presumably be uh close enough so when this Carlton at all paper came out um ethereum had really sort of just started there certainly wasn't any um D5 or decentralized Finance so pretty much the only things they were to worry about were block rewards and transaction fees there was not really any other source of minor Revenue back then now and this is what so often happens with good research which is that if you nail the essence of some issue you just find that it just sort of reoccurs in Disguise you know in a way you never could have intended many years later that's exactly what's happened with that Carl's done it all study um because they're sort of their identification of these undercutting issues that that my neighbors are incentivized to carry out if there's High variance um Revenue coming from blocks that's now become more relevant than ever with the Advent of decentralized Finance on ethereum and other smart contract platforms there's something known as Mev I'm not going to even try to Define it but I roughly think of it as like you know miners or black producers kind of siphoning Revenue off of the off of the application layer thereby giving them a third sort of stream of rewards um and even more so than with transaction fees Mev opportunities can vary tremendously block to block by by orders of magnitude so again you sort of wanted to try to think about sort of smoothing them out now that's going to be harder with Mev than with transaction fees because transaction action fees are directly visible to the layer one so for example two ethereum mainnet and it's in a position to just smooth it out itself Mev lives at the application layer so it's not directly observable by the underlying consensus protocol by by the by layer one so to carry out this smoothing idea you need to have some other ingredient like maybe a sort of competitive market for Block building so that you know the the Mev value of the application layer winds up getting expressed down to the consensus layer so that it's in a position to to smooth out those payoffs if you want to see the latest along that I would um just do a search for Mev smoothing and in particular the ethereum community's latest work on that topic so let's move on to our second example um we'll be segwaying from Bitcoin to ethereum and from black rewards to uh transaction fees and let's look at the market for ethereum Block space through the lens of mechanism design so ethereum black space is typically a scarce resource so one desired outcome we might be interested in is allocating uh that scarce resource in the most efficient way possible that is to the transactions that gain the most value from being executed in the ethereum virtual machine one way to suss out the most valuable transactions would be to charge as a transaction fee the market clearing price here by market clearing price I mean the price at which Supply equals demand so if the supply of ethereum block space is like 15 million units of gas per block then I'm talking about the price at which there's precisely 15 million units of gas worth of transactions that are willing to pay that price that's what I mean by the market clearing price if the market clearing price just fell from the sky we'd be good to go the black space would be fully utilized and by precisely the transactions willing to pay that price so precisely the most uh valuable transactions of course market clearing price does not fall from the sky so we have to address where it might come from in the transaction fee mechanism that's that's always been used in Bitcoin and also was used in ethereum until a little over a year ago something known as a first price auction uh their users basically just have to figure it out for themselves so every user just attaches a bid to its transaction which it has to pay if that transaction winds up getting included in some block and with this mechanism you're kind of hoping that you know at equilibrium in some sense or the word users have basically figured out how to bid meaning they basically figured out the market clearing price that however is not so easy to do first price auctions this actually can be pretty tricky um to figure out what to bid especially if you have sort of wildly changing demand as you often do in a blockchain context uh so those concerns were the primary motivation behind the design of the mechanism known as eip1559 so this is the transaction fee mechanism that ethereum is using now and has been since August of 2021 and here the idea is rather than foist any work onto the users the protocol itself the transaction fee mechanism itself is going to try to keep track of what the current uh market clearing price is something known as a base fee which again is an estimate the mechanism's current estimate of what the market clearing price is so some additional details so first of all the mechanism is continually adjusting the base fee after all demand and therefore the market clearing market clearing price might be changing over time so it's continually updating the base fee using a form of local search um so if recent blocks kind of look too small smaller than you'd like that suggests the current base to be set to aggressively and then you should decrease it if recent blocks are sort of bigger than you'd like that suggests that the base fee is too low and that you should increase it so this mechanism turns out to have some really nice properties from a ux slash sort of game theoretic perspective so for example most of the time it's as easy for users as shopping on Amazon so it acts as something known as a posted price mechanism so basically as it from a user perspective it just kind of says there's a price posted hey 10 bucks to have your transaction included you can either pay it or not just as if you were shopping at Amazon there is sort of an edge case or if the base fee is like way too low then you kind of gracefully revert back to a first price auction but but you don't have those nice sort of ux ux properties the mechanism is also unusually resistant um to collusion of various forms so for example between a block producer I mean a bunch of users who'd like to have their transactions included in that block and a Twist to this whole analysis is that those collusion resistance properties actually hold e only if you do something counter-intuitive with the revenue generated by the base fee the obvious thing to do with the base fee revenue is to pass it on to the producer of The Block right that's what's always sort of historically been done with like transaction fees and and bids in these transaction fee mechanisms but if you do that then in fact the game theory breaks down okay then in fact a cartel of a block producer and users can collude off chain to circumvent any kind of Base fee you might be trying to enforce on chain however if you redirect the revenues generated by the base fee literally anywhere else it doesn't matter as long as it's not to the producer of that block if you direct them anywhere else then actually you do get all of these unusual and very attractive collusion resistance properties so you could imagine different versions of eip1559 that answer the question you know where to route to the base fee revenues in different ways just like you could imagine different versions of Bitcoin with different macroeconomic policies concerning the circulating Supply and again here a specific Choice was made in the version of eip1559 deployed in ethereum which is do maybe the sort of simplest possible thing with the base fee revenues which is simply to burn them by which I mean that the currency used to pay transaction to pay the base fee uh those coins are just removed permanently from circulation and it is in this design decision then within EIP 1559 where it's now crucial that the mechanism has access to a native currency that's why it's able to burn permanently um the coins used to pay those transaction fees I should say up till this point like everything on the last slide all of that everything could have been denominated in usdc and it all would have made sort of perfect sense this step would not make sense if everything was denominated in usdc uh you know the ethereum blockchain of course isn't a position to burn some digital representation okay of a US dollar that sits somewhere else it is not in a position to actually burn that U.S dollar right whereas it is in a position to burn the native currency to burn eth so the only reason this works the only reason you can get away with this Burning the base fee Revenue solution is because you have access to that native cryptocurrency but as usual with this power to manipulate the underlying cryptocurrency comes responsibility meaning macroeconomic implications right so the design decisions that have been taken in EIP 1559 all motivated primarily by Simplicity either Simplicity for the user and figuring out how to bid or Simplicity just in the in the protocol and its implementation by doing the simplest possible thing to dispose of Base fee revenues that combination it just has macroeconomic consequences right it's sort of the opposite of the block rewards that we had in Bitcoin rather than a form of inflation this is a form of deflation right so this is pushing toward decreasing um the circulating supply of eth there are still sort of inflationary aspects of ethereum so you get this tug of war between the inflationary parts and the deflationary parts but long term I think most of us expect eth to be a deflationary currency so we have this non-trivial macroeconomic consequence of design decisions that were really made more from microeconomic reasons and a question you could ask is like is this byproduct should we happy about this or is this just sort of a cost we have to pay you know how should we feel about this now if you ask anyone who holds a non-trivial amount um of eth of ethereum's cryptocurrency they're generally going to be you'll find they'll be really big fans of these design decisions and of the deflationary pressure on the currency that's caused by EIP 1559 the reason being is if you sort of you know posit that the overall value of the ethereum cryptocurrency should be independent of the circulating Supply right like if you double the number of coins well then each coin should just be worth half as much sort of and vice versa well then whenever coins are being taken out of circulation and the overall number of coins is decreasing that means the per coin price should be going up so holders of the currency are sort of hoping that this deflationary pressure makes their coins worth more going forward if on the other hand you ask any traditionally trained macro Economist about this they're going to tell you oh my God what a bad idea they're going to just have a knee-jerk reaction that deflationary currencies are doomed to fail they'll probably tell you stories about sort of stagnation uh in Japan back in the 1990s now I got to tell you I'm not super convinced by this analogy it's not obvious to me that like I don't know sort of the currency of a of a Sovereign Nation is such a great analogy for sort of a cryptocurrency like eth but I have to concede that like the gauntlet has been thrown I mean I do think it is up to researchers in the blockchain space to explain why right why it might be that deflation is not a disaster for a cryptocurrency in the way that it's believed to be for currencies of uh of sovereign Nations all right so the two examples we've covered so far you know I think kind of despite the fact that so many people Define crypto economics in so many different ways I think um pretty much everybody would agree that the two problems we just saw are indeed um crypto economics and in particular both of those illustrate this very sort of unique interplay between the micro and macroeconomics sort of the mechanism design problem which again really does not have analogs and traditional mechanism design for this third example you know I think some people would call this crypto economics and some people wouldn't I wanted to talk about it anyways number one because I wanted to say at least something at the application layer we've been talking much more about the consensus layer thus far plus I just wanted to you know tell you about sort of some of the work I've been doing uh most recently in sort of default in decentralized finance so specifically I want to talk a bit about automated market makers or amms through the lens of mechanism design now my guess is you know if you're watching this video there's a good chance you know at least kind of superficially what amms are about you know but just sort of a quick review but to the point of any exchange and in particular an amm is to allow two different assets to be exchanged for each other you know maybe eth for usdc and vice versa we've of course had exchanges sort of in the traditional markets forever I mean there's largely a convergence on a particular design namely a limit order book uh but in blockchains thus far the dominant Paradigm at least for purely on-chain Solutions has been different and that's because limit order books they're just sort of a poor match for blockchain protocols at least you know the ones that we have right now they're primarily just because they're too expensive they require too much storage and too much computation for a purely on chain exchange you really want to do something simpler than a limit order book automated market makers or amms are then that simpler format so an amm is not keeping track of some perhaps long list of sort of outstanding limit orders it's really just keeping track of generally sort of two numbers which is how many coins X and Y um it has of the two different of the two different assets that's enabling enabling the exchange of and as a function of those two supplies of X and Y programmatically um there's going to be some spot price which is then quoted at which Traders can are free to buy or sell so in an amm a little bit of computation is required to determine how much you get back of one asset in exchange for a given amount of the other asset but it's generally a very lightweight computation certainly much lighter than trying to do matching of orders in a limit order book interestingly right amms they've they've risen to prominence um kind of in web 3 in the blockchain world not so much for for economic reasons but really just for a computational considerations right just for their uh simplicity so they're really kind of the product of doing mechanism design under very severe uh computational constraints just because you know the amount of computation available in a blockchain like ethereum is just is really quite Limited this Echoes um sort of the theme in a well-known branch of algorithm the game theory called algorithmic mechanism design which focuses on mechanisms guaranteed to run in polynomial time but polynomial time is definitely not good enough for something you're going to implement purely on chain on a blockchain like ethereum you really need much tighter control over the computation required and it's under those severe constraints that amms really shine unlike the last two examples amms don't really fundamentally rely on the existence or access to cryptocurrencies right back when we talked about Bitcoin right big the mechanism needed to sort of print new Bitcoins in order to reward producers of new blocks that obviously isn't possible unless you actually can directly control and manipulate the currency similarly in EIP 1559 all of those base fee revenues got burned so that you'd have good game theoretic properties again not clear how you'd pull that off if you couldn't sort of directly work with a native currency amms you know they're typically used to trade cryptocurrencies granted but there's nothing fundamental about that you could use an amm to trade any two assets right X and Y on this slide could be units of anything and meanwhile the actual sort of operation of what an amm does I mean it's basically just some arithmetic right it really doesn't need some kind of native token or native currency in order to carry out the operations of an amm again unlike Bitcoin block rewards unlike EIP uh 1559. and indeed when uniswap right the most famous amm out there when they launched originally they did not have any kind of token because they didn't need one there were other tokens that you could trade in their amms and again the operation of the MMS didn't didn't really need that token so there is a uni token now but that was added only later it's really not fundamental to how those uniswap's amms operate all right so I promise to say just one minute about um the work we the research I've been doing lately concerning automated market makers um which is about a measure known as lever so that's lvr which stands for loss uh versus rebalancing now we've already talked about you know benefits of amms so principally they're just computationally uh very simple so suitable for on-chain implementation in today's blockchain protocols um they also guarantee liquidity even for sort of long tail um assets where there's a thin market so that's another nice property of them but they do have uh drawbacks too inevitably and so one of the main ones is what's known as adverse selection okay for the liquidity providers remember LPS they're the ones that actually deposit the coins in the first place the Traders are then trading um back and forth so by virtue of always being open for business always being willing to take either side of a trade at the at the current quoted spot price basically liquidity providers can be forced to carry out trades at worse than market prices so for example um if the spot price of an mmm ever become stale if it ever sort of drifts away from the price of the asset on the open market then you can expect some sort of opportunistic arbitrageur to make a trade that corrects the spot price of the amm brings it in line with the rest of the market those kinds of those Arbitrage trades they will um yield profit for the arbitrageur and that profit will come at the expense of the liquidity providers so that's what I mean by the LP's adverse selection costs being forced to take possibly the wrong side of every single trade so the work I've been doing with a number of co-authors recently strives to isolate and characterize these adverse election costs to really just give sort of a formula for exactly what's the cost to an LP for providing liquidity to one of these pools now remember it's not all bad news for the LPS they're also getting a share of the fee Revenue so if that's high enough that will outweigh the losses from from Lever but lever is sort of isolating the cost part of the cost benefit analysis that liquidity providers should be doing so before our work so over the last couple years adverse selection costs were typically measured by something known as uh impermanent loss and I won't Define that for you here but let me just say our critique of that is that it conflates two really fundamentally different things it conflates the adverse selection costs that we're trying to isolate with sort of price movements of the underlying assets so in particular large adverse selection costs can be occluded by favorable changes in in market conditions so one way to think about our concept lever is as basically taking in permanent loss and modding outs by all the parts of it that depend on Market movements so another way to think about is if you hedge out the hedgeable parts of impermanent loss you're left with this sort of unhedgeable residue okay which just stems from sort of Trades being executed at unfavorable prices and this unhedgeable residue that you're left with is exactly what we're calling uh lever or loss versus rebalancing I'm not going to go into technical detail here let me just say that if you want to compute historical lever so using past data that's quite straightforward to do using the typical data you'd have for an amm's performance if you want to make predictions about the lever going forward like if you're a potential liquidity provider assessing whether to deposit your coins in an amm or not you want to know what kind of lever you should expect it actually turns out you know for most of the popular amms to be surprisingly easy to work with analytically in kind of a standard black Shoal setup so one example of that would be the sigma squared over eight that's on the bottom of the slide that I'm not going to explain here but if you want to know more about lever you can check out YouTube videos by by any of the co-authors of this paper so myself Jason miliones who's a PhD student at Columbia uh CMAC molemmy who's a professor in the business school at Columbia and Anthony Lee Zhang who's a professor in the business school at University of Chicago um so we all have longer form talks on lever which you can find on YouTube all right so that wraps up the examples I wanted to discuss let me just conclude this talk with three slides of kind of grand challenges let me call them in crypto economics I've grouped them by theme so I'll sort of talk about the theme of each and mostly just leave it to you to read sort of the specific example questions um that are on the slide so Grand Challenge number one which is maybe the most ambitious of them all is to make macroeconomics our own so and here by we I mean you know the types of people that I Envision probably watching this video so computer scientists you know blockchain researchers uh Etc and this is something we've seen happen uh frankly with other parts of Economics over the past 20 to 25 years it started with Game Theory okay so that's you know it's been around forever since the since the 40s um and in the late 90s computer scientists got interested in Game Theory because they needed it to reason about the applications that they cared about uh and at the same time that they learned how to apply game theory in their applications they also brought sort of unique um things to the table okay so they also made Game Theory their own bringing in for example a concern for computational complexity you know the price of Anarchy of Nash equilibria and so on it happened next with mechanism design which again has been around forever it's been around for I don't know you know 60 plus years and again computer scientists you know it you know we sort of needed to learn mechanism design because again it was so relevant for the applications that we wanted to think about but over the last 20 years we've also in some ways made it our own again we've with an emphasis on sort of computational and communication constraints um using approximation to overcome problems that are historically sort of dismissed as as just unsolvable and weakening informational constraints which are sort of pretty common historically in the mechanism design literature and I gotta tell you like to date at least as far as I know there's been basically zero conversation between computer scientists and traditionally trained um macro economists like if you look over the conference proceedings of the EC conference so EC stands for economics and computation that's the flagship conference in algorithm the game theory I'll bet over those 25 you know conference proceedings the median number of papers okay in that conference in a given year that touch macroeconomics I'm guessing the median number is zero I'm not saying there's never been any in the history of the field but it's it's really very rare similarly you know I go give seminars at economics departments all the time I've done it dozens and dozens of times they always take me out to dinner with a selection of the faculty I've never met a macro Economist at one of those dinners ever so there's definitely like a big gap as far as right now as far as the interaction between the two like Game Theory like mechanism design you know we're gonna have to come to macroeconomics because we need it we need to understand fundamental questions that come up in the design of blockchain A Cause right Nakamoto a hard cap was a good idea is it a good idea and again I don't want an answer based on like politics or ideology I want to I want an answer based on Technical and economic analysis right similarly remember the Paradox of the deflationary cryptocurrency right you have etholders who love it you have macaroni macroeconomists who think it's a disastrous idea please give me some technical or economic analysis to help sort of resolve these these conflicting opinions so the second Grand Challenge is a topic on which I think we've actually made a lot of progress as a community over the last few years though I do think there's still a great you know huge opportunity to take the ideas that we've learned so the lessons we've learned so far and transform that into really a mature uh Theory at the level of Shay the theory for analyzing classical consensus protocols where you know people pin down sort of the optimal fault tolerance possible for like every imaginable distributed computing model you might you might ever think of I think we can get to to a similar level with this topic namely uh incentives uh at layer one in the consensus layer already with the Bitcoin and ethereum protocols there were some very nice ideas on this topic we've discussed some of those previously uh in this talk um and then there's been sort of more breakthroughs since right so for example um we've got post merge ethereum or eth 2.0 whatever you want to call it uh which introduced a number of further ideas um in order to incentivize the nodes running the protocol to behave as intended but while we've definitely had some good ideas thus far I do think there's a real opportunity to take kind of our I'd say a fairly ad hoc current understanding and transform it into a really kind of complete and mature Theory right so just to give one example so you know slashing is definitely part of the post merge ethereum Vision so post merge ethereum is a proof of stake blockchain protocol those those nodes responsible for carrying out the protocol are forced um to put up a stake uh in the native ethereum cryptocurrency and that state gets slashed I.E at least partially taken away from them if they misbehave in a detectable way and I should say it is not at all universally agreed upon that you should do slashing in a proof of stake blockchain there's plenty of other proof-stick blockchain protocols that do not do slashing they will still kind of withhold inflationary rewards from nodes if they misbehave but they won't actually sort of take away any of their stake whereas in post-merge ethereum um If you're sort of a validator you really are at risk of having your funds slashed so the obvious sort of danger with slashing is you might wind up slashing nodes that we're trying to be honest and just there was some kind of mistake uh so that might be a little bit of a harsh of a punishment I mean then moreover right you might deter nodes from running the protocol if they're concerned about being slashed um you know despite their uh best efforts on the positive side it seems like you know slashing enlarges the design space around proof of stake blockchain protocols in a way that seems to let you do things that you couldn't otherwise but again we do not have a formal proof of that so once again we just have these super super basic questions you know which for example I get asked all the time in my role running the research team at a16z crypto and frankly just the theory is not is not there I think it will be there but we've got a lot of work to do so the third circle of questions um concerns the layers of the blockchain stack and understanding the economics of each layer in isolation and then also the economic interactions between the different layers of that stack thus far most of the research effort has been focused on layer one the consensus layer like we've been discussing or on the application layer but there's other layers of the stack and they're actually really important for example I have to call out the peer-to-peer layer or layer zero right so you know when you think about a consensus protocol you presuppose something underneath it which allows the nodes to communicate with each other you know either maybe they're voting on blocks maybe they're sort of sharing information about pending transactions right and blockchain protocols usually do not directly incentivize nodes to participate correctly in the underlying sort of peer-to-peer communication protocol the status quo may or may not be economically sustainable it's not obvious perhaps you know nodes have enough external reasons to participate correctly in the peer-to-peer Network um to do it despite the lack of direct incentives but again like we just don't have a formal analysis of that we really just don't understand uh layer Zero from an economic perspective and we need to then there's also at least in sort of our present designs fairly complex interaction between the various layers like I think of Mev as BB being maybe sort of exhibit a here I think of that as an interaction between layer one and the and the application layer and you know for example people like to assert that Mev is sort of unavoidable like you you can't get rid of it maybe that's true you know I don't know but certainly no one's ever proved the theorem which says that it's unavoidable and to make that statement you know to my taste I would like to see it backed up with some with some formal analysis you know zooming out even more broadly you know at the moment we're seeing these very complex interactions between the different layers of the blockchain stack and what's completely unclear to me is whether there's something about Building decentralized Systems that makes that inevitable that you just can't have clean separations um between the different layers but for all I know that's really just kind of you know an artifact of how little of the blockchain design space we've studied to date so I hope that gives you a little bit of a taste of the kinds of questions that are on my mind I mean the kinds of questions that people ask me all the time both in my role as a Columbia Professor I'm in my role as head of research at a16z crypto the list of examples in this talk of course was not comprehensive this list of Challenge questions of course uh is not comprehensive it's just merely meant to indicate that while we have a few answers we have far far more questions we have a lot of work to do but that's also the most exciting time to be in a research area so that's all I've got thanks very much Back To Top