Conversation on incentives with Tim Roughgarden, Scott Kominers, and Pranav Garimidi

Transcript

[Music] so my name is Tim Ru Garden I'm head of research very excited to be here uh you'll be hearing from uh introductions to all almost the entire research actually the entire research team is here uh we're all just very excited to participate uh you'll get introductions from all of them shortly so I thought I'd just spend a few minutes talking about the research team in general and sort of how we interact um with our Portos um because life on the research team it's really kind of two different worlds that are actually much more integrated and synergistic than you might expect on the one hand everyone in the research team is not just kind of a consumer and is not just knowledgeable about Cutting Edge research everyone on the research team is expected to contribute to Cutting Edge research

themselves so you will see like major papers in crypto with the names of a6c crypto researchers regularly you will see that multiple times a year that's part of part of um you know the reason why we hire the people that we do the other half of the job is really just to collaborate directly with our portfolio companies a lot of people actually think the job is about due diligence like outside and it is a little bit of that but honestly it's much more the fun stuff where you take the portfolio companies we've already invested in and right it's not just we don't just sort of cut the check we also want to help those portfolio companies to succeed and that's kind of the adrean horowits philosophy and the research team is very much um kind of in that same spirit so we are regularly you know there's dozens

and dozens of companies in the portfolio that we meet with regularly sometimes it's very lightweight sometimes it's kind of just like office hours they want to sort of hear about what are the relevant research papers to the problems they're working on or they just want kind of another set of eyes on a design sometimes we're really in there sort of working together collaboratively and sometimes the research team really actually sort of deeply influences um what the eventual what the eventual product is and these again these are they sound maybe like totally different activities but they're actually really related it's not just that our sort of deep knowledge of the research enables us to be that much better a collaboration partner in solving those problems but I got to tell you just like selfishly like in an immature field like this where you can't just kind of read

last year's conference papers and figure out what the important research problems are the competitive Advantage we get as researchers by being at a16z crypto and talking to people like you all the time is unbelievable I genuinely feel like I get about an 18-month advanced notice on what's going to be the important papers to write compared to if I was just uh sitting in the Ivory Tower so it's super exciting for us to talk to all of you you know one we just think a lot of you are bu amazing stuff secondly it's sort of our little way of seeing around the corners about where the science and technology is going so that's going to be my intro Let me head it off to to Eddie these are the members of the research team that are going to be here today Daniel renod uh who you may know

as karma who's in the back there uh who has a he has a background in formal methods but uh but is also a smart contract engineer so he's very deep in evm and that SK those that combination of skills is why Karma works very deeply on halmos with dejun who's also here which is our symbolic execution and uh testing testing tool so if you're writing evm code you're concerned about the evm code being correct and you really you really really want it to be correct in a way that's maybe more interesting or deeper than standard tests or even a little bit of fuzzing uh could do you should talk with Daniel and dejun they'll be uh they'll be uh in one of the breakouts that we recently added to the schedule based on security uh

deun was also at uh runtime verification in the past and has a lot of experience with security auditing uh and has you know relationships with people in security auditing so if you have questions about about that uh dejun would be a great person to pick Michael Jew is one of the co-authors of uh of jolt one of the co- implementors of jolt Along With Sam ragdale from the engineering team uh and he'll be joining the ZK uh breakout group later uh and so if you have questions about jolt which is our general purpose ckvm you should definitely ask uh my we would really love to support you if you want to use jolt like EXT extra special that's like a special moment so please if you want to do jolt I would love to talk to you Sager uh does a lot of things within the firm besides uh

being very deep in uh like for example farcaster forecasters background uh uh backend implementation Hub implementation uh Saga was also at salana a while back and is very very familiar with the salana virtual machine and so on but also generally security practices you know signatures uh signature schemes in different contexts uh just he's sort of a generalist in uh security as well uh but I think he'll be I think he's going to be in the incentives group because he thinks a lot about uh about such things I'll say more later uh Noah many of you should know since he's staying in London during the program Noah is also a bit of an applied

cryptographer recently so also a great generalist but if you have questions about l2s if you have questions about uh optimism if you have questions about uh like clients if you have questions about uh any of those things I think Noah is the right person to talk to so that's a a quick cut of the engineering team the research team sort of Falls pretty naturally into two halves one um which is the side focused on incentives Market design mechanism design kind of business strategy so that's we're going to start with an introduction to that half of the research team uh and then the other half is in various flavors of cryptography ranging from from ZK to MPC to all kinds of other things so that'll be the that'll be the second half similar format for both so I want you to get to know the team members a little bit um

and then also just hear a bit about the kinds of work they do both on the sort of pure research side and in the Porco collaboration side and then there'll be at least some time for just you know General Q&A in each of the two sessions um so let's start um by having so pranav and Scott I'd like um let's maybe we'll start with uh start with Scott then move on to PR of so just to kick things off just a little bit about sort of who you are your background the kinds of things you like to work on um and then I'd also like to hear specifically something you're really excited about on the research side right I mean I think this is an opportunity for the group to really hear from people that are literally at sort of The Cutting Edge of research um so it could be your own

research or just something else that you're sort of super excited about that's happening kind of right now on the research side uh maybe also just kind of at least one anecdote about the kinds of collaborations you've had um with our portfolio companies thus far so Scott we'll start with you great to great to meet many of you and see a few of you for the for the end time thanks so much for beaming me in um so I'm Scott commers I'm a research partner at a16z crypto I also teach at Harvard Business School uh and in the Harvard economics department and I'm a market designer by training so I study the design of markets and market clearing mechanisms and more broadly how to think about you sort of the the incentives in a system and how to use a mixture of like economic reasoning and computer

science and sometimes a little bit of uh applied math to shift the incentives or to change behavior in a way that improves outcomes right so it's about you know it's it's like economics is a perspective in the world it asks how does the world work you know how do people respond to incentives and then what does that do to outcomes Market design asks how should the world work uh you know sort of how can we use economics to change and improve outcomes and as such I've been in different capacities involved in in crypto for a number of years at this point um and of late have been thinking particularly about uh a lot of consumer crypto applications um but you know I work with teams both you sort of on the research side and in terms of the portfolio on

everything from sort of like the the fine grain details of tokenomics to like much higher level strategic questions like how do you you know maintain competitive advantage and capture value um so you know uh and in fact maybe I'll maybe I'll talk about a story that that links both and sort of his is feedback back and forth so one of the first things I noticed when I joined the research team was that I was having t and tons of conversations with teams about versions of the question how do we do our initial token allocation or how do we think about the right mechanism for for selling individual units or something of the sort uh I was talking to a lot of companies that in one fashion or another manage nft projects or have developer ecosystems where they're trying to deploy tokens as a way

of incentivizing developers um sound like anybody in here maybe Tim smiling good actually can't tell other people are laughing the the the room is not mik so if people are like falling it's got to make it to the front mics I would say Smiles but not gas so fair enough it's it's early in the morning here sorry my comedic timing is way off um so you know and there are some patterns there are some things that you've probably seen happen a lot right you know teams will uh do something to try and identify and label their you know sort of their expected top contributors um or they might have you know various forms of challenges or contests you know you've seen versions of uh you know retro pgf too right sort of situations where you sort of try and look and later assess

like what created a lot of value and how do we reward that um and so we found we were giving a consistent on the one hand we were giving a reasonably consistent set of recommendations to teams we were sort of spending a lot of time telling them okay so you you know tokens you know people's incentive you know first first order economics right people's incentives to invest in in building value around the token depends on the share of tokens that they hold and like sort of the way and and as a function of the value they internalize um and you want to bring in people who drive you know who have significant externalities significant benefits that they drive to the entire group uh or sort of to that they drive to the ecosystem uh but there actually was no formal framework for thinking

about how to actually do this correctly but it turns out the tools of mechanism design in fact like a branch of me a particular branch of mechanism design that that I had a number of collabor has been working on for years uh made it possible to make this really formal uh so normally when we you know a lot of the classics of mechanism design Theory um assume that sort of everyone in a market is the same from the perspective of the designer that you sort of you don't care who you know who receives you know given units or whatever other than whatever they're willing to pay you for them or something of the sort but that's not true when you're building an ecosystem when you're building an ecosystem there are externalities if you're launching an nft project you maybe care about having Marquee members or people who are going to invest a lot

in like developing their characters if what you are as like a character brand um if you're launching a developer ecosystem you care about you know bringing in the developers who are going to work the hardest and and produce products that are the most value to the broader ecosystem uh with users you maybe worry about congestion you know so if or or you know God forbid adverse selection where you have people who just acquire the token because they're purely speculating and have no desire to build again you know sort of not sure whether that happens but um and so with a with a team of collaborators we've built a mechanism design framework and we're now calibrating it to sort of give direct predictions as a function of sort of how you understand the externalities of your

of your user base and it makes very precise predictions right it says things like um how you think about substitutability and complimentarity like cross- subsidizing different groups of users um and you know or for example uh you know a common intuition that we have is that we want to is that we might want to induce some sort of or require some sort of token vesting to sort of reward people as a function of their ongoing contribution um but there there's a precise reason for this that the the math spits out which is that um people's behavior in a you know in an ecosystem people's desire to select in

isn't a function of how much they think they can contribute it's a function of how optimistic they are about the future value of the ecosystem and those people are sometimes the best Builders but they're not always the best builders and so by requiring you know if you just distribute a token immediately and most of the value is to be developed Downstream right it doesn't exist yet you know maybe some people who are very optimistic about the ecosystem will acquire but everyone else will just sell it um whereas if you induce festing and work and sort of especially if you can tie that vesting to project Milestones um then you actually have a more virtuous selection pattern right it isn't just that you give people a reason to stick around but you actually endogenously sort of as a fun fun like the the incentive select for the users

who are actually building the value because those are the people who you know capitalize the most in this vesting MoDOT where those the people who like get the most value by D you know sort of in the long run because they know that they can invest to grow the ecosystem and so on the one hand it delivers a lot of economic intuition on the other hand it delivers literally like there things that look like formulas for how you sort of think about calibrating these allocations and then since you know since writing a paper on that and and and thinking about it as a researcher we now take that to teams and I spent a lot of teams helping them think about a lot of time rather helping teams um think about this in context and more broadly like sort of what these ideas about like

how the cross externalities between different types of users affect questions about how you should scale your platform like who you should try and bring in when and so forth and and critically again right it's like as a researcher we both provide intuition we're both explaining things like look what you're look what you're seeing as a selection effect right so you move up or down depending on which which direction you're going the demand curve um you're actually drawing in different types of users and you have to think really hard about like how you identify the different categories of users who are who are most valuable and try and select for the ones that are that are actually going to build along with you um and then you know we spent a lot of time saying looking at protocols like

literally yesterday I was on the phone for an hour with a team where they were running me through the details of their actual like primary sale protocol and more and sort of then what they think the feedback loop is going to be and we we sort of deconstruct them and try and help them build them better anyhow that uh sorry it was probably slightly longer than originally expected but uh hopefully give a gave a sense oh Scott that was great so panov yeah so hi uh my name's Prav and so oh yeah I'm a research analyst in the team um and have been doing kind of have a similar background to Scott and various kind of mechanism design research problems um and I've been doing that for about five or six years now and then for the past kind of two or three years I've

really been focused um on blockchain problems and I think kind of while Scott focuses um well he does a lot of things kind of focuses a lot on the application layer um I kind of look at the infrastructure layer a lot more closely and so a lot of my research is kind of understanding what's been coin the transaction supply chain uh and basically in different L ones and different ecosystems basically you know how does a transaction go from users and then there's like Searchers and Builders and kind of this whole pipeline of before transaction fully might land on chain it's part of some block of proposer kind of puts out there and so I kind of see a lot of the mechanism design questions that go along with that um and so if people might have heard about things like TBS or orderflow auctions or anything along those lines

um a lot of kind of what I'm working on is trying to understand how do those things work today how can we kind of formalized guarantees those things give you kind of asking questions around you know what properties would you want on uh the supply chain to give you how do you kind of measure uh are like users getting good execution or users having good like ux experiences you know are these kind of system decentralized kind of what the different economic forces in play when they kind of add all these different parties and so I kind of try to answer uh general questions around the whole supply chain in general I think one thing that's been really interesting for me as of recently is trying to understand uh kind of these newer kind of like high performance l1s uh so kind of in a portfolio we have

apos since we um kind of SL and really understand kind of like how does the supply chain differ in these environments compared to something like ethereum and so ethereum things have been kind of have evolved a lot and there's lots of kind of research attention from lots of people in the space trying to understand these different mechanisms understand kind of the properties you want and i' say that's kind of way less developed than some of these different uh l1s and so maybe one concrete example is one thing I study a lot are these fee mechanisms so basically you know on ethereum this is things like gas prices just like how are those prices set how how much you just have to pay I think on ethereum you know there's been a lot of work kind of you know starting from what Tim did um a

few years ago and I think at this point we have a like a pretty good grasp basically of there's some ways we can say the current fee mechanism is somewhat optimal and we kind of know the exact crade offs anybody can make but I'd say the question gets a lot murkier when you move to different l1s and one key reason is because you know ethereum has this basically a single dimensional environment you know everything happens sequentially whereas a lot of these kind of newer ones are trying to use these parallel extion Environ ments and that means kind of the resource constraints that transactions have changes a lot it's a lot harder to measure kind of like how much uh Compu do take ex transaction when not only you have to worry about kind of exting the transaction itself but you know what other transactions is executing one

transaction block for you know various like parallelism uh concerns and so that in turn leads to it's unclear how to charge fees in transactions in these kind of neural ones where everything's kind of parallel and so a lot of uh one at least one project I'm working right now is working with uh you know the Aptos team pretty closely and some people on the salana side of trying to figure out you know are there new fee mechanisms we can use to really kind of in some sense kind of unleash the full potential uh of like the block space that these kind of higher performance l1s offer and you know that's partly you algorithmic challenges of just the fact that you have to solve in kind of technical terms is like multi-dimensional napack problem versus single dimensional napack problem is

actually way way way more complex we don't understand it quite as well and there also just various incentive concerns of just kind of having all these different resources means the kind of information you might want to extract from user submitting transactions is a lot more complex right in terms something like ethereum maybe you're happy just having them report like a total gas cost and that's actually a pretty good proxy for how long a transaction might take but when you're in these parallel environments and there's kind of all these different parameters that might affect you know how performing a transaction can be there's just trade-off of how much you want uh to allow users to express right do you really kind of go this long rout and make you know us just Express these like full access lists and tons of details that can be really helpful on

the pricing side but also kind of open you up to different types of grouping attacks of people manipulating how they submit these parameters to kind of get favorable pricing you know the other side something like an aptose makes very few requirements that users to submit um information about their transactions and so that puts more strain on the apos side when they're like kind of charging fees they have less information to work with but it's better ux potentially for you know developers and so a lot of what I work on is trying to understand those tradeoffs FR grain level and kind of figure out um you know can something about what's optimal given kind of the use case you want let me ask ask a a follow-up question PR off so because you know you talked about the separation of the infrastructure and the application layers and you sort of pitched yourself

as sort of focusing on the infrastructure layer right but like you have this one phrase I really like actually which is you can't blackbox the blockchain right which really is kind of about how you're actually sort of working on the application layer in a way or you're informing the application layer design also could you say a little bit about what you mean by that phrase yeah so you know one of the main benefits of blockchains as Tim says lot you know this computer in the sky that kind of anybody can use and share uh but it doesn't mean you do have to share it and so you know what else is happening from other applications uh can actually really impact what your application is doing so I think it's you know it'd be really nice if you're at a point where you basically assume this computer was

perfect and you could submit things to it and execute it and you get back some great response uh unfortunately we're not quite there yet today with how the kind of inferus set up and so I think you know as somebody building applications it's really important to kind of understand how this transaction supply chain will affect you you know so if you're building a defi protocol it's not a good idea to assume you know I'll have my transaction and then that'll just land on chain the exact way I want it with nobody else seeing it you know you have to be quite careful of knowing okay I'm going to submit my transaction that's going to go to some type of me pool there's going to be some set of Searchers kind of interested Financial parties kind of looking at that transaction they're going to maybe insert the things before or after they

may try to manipulate that somehow by manipulating some other type of Market uh and then you there's going be some kind of long process of some Builder something have to package my transaction and get it in the block and then submit some bid kind of blah blah blah you know there's kind of a lot of contingent reasoning basically and and so the main point is you should definitely be kind of mindful uh that you are in some sense in the dark Forest uh and that like you you really kind of should be I think adversarially thinking uh especially kind of on the def side kind of any kind of side of uh any applications that really do interact with state in complex ways of there's lots of parties who might try to mess with you and it's important to kind of design your protol that in mind you it can also be helpful

sometimes right by being in this kind of in everybody's kind of looking at what you're doing you can use me style things to your advantage of you know sometimes having auctions with having like arbitragers uh can be something that takes something off your plate if you can assume that you know there is some type of your Market um that'll be corrected by just the market deciding it versus you kind of explicitly having to control for something uh you know can also help you out and so it's just I think being mindful of that whole process is quite important and even if you're not doing something defi related if you're doing a more kind of traditional app layer thing it's still important to think about you know how these kind of transaction fees and you know blocks getting full uh kind of full can affect you you you should be mindful

like if if kind of my application is assuming every single block there will be some transaction included and that's actually very crucial to some ux flow that's happening offchain you know it it might not be the case you get a transaction L at every single block you there might be some random thing happening on chain and you know infra things get weird and you know n world and there's some blocks that like can look very weird and so if there's a span of you know one two or three blocks where your transaction doesn't land uh you should be mindful of like what's happening offchain is kind of prepared for that and so I think just uh having a good at least a high level understanding of this whole kind of transaction supply chain and exactly how it gets how blocks get built from a high level it's

important just to have kind of the right contingencies in place and kind of be mindful when you're designing different applications Scott let me ask one more question to you and then I'll open it up to the group for for Q&A so Scott you live I mean you kind of have multiple different universes you're in at the same time um and but just purely on the academic side right in addition to kind of the the market design expertise that you bring right you also um uh teach at Harvard Business School right and so and you have for many years you've been a professor there and uh like early in your career you would be teaching sort of basically you know web2 business strategy and then as things have evolved right what became more and more interesting to teach was web three business strategy which you have

basically had to invent in order to even teach it um and so I'd love to hear just a little bit about your thoughts on that side so what changes on the business strategy and value capture size going from web to uh to web 3 I know there's a million things you could say here but let's if you could keep it to maybe five minutes ideally that would be great yeah ex I'll I'll pick I'll pick like two or something okay um yeah so just to just to illustrate and as Tim says um I've been teaching various forms of marketplace and platform design um both in our MBA and our graduate program for a number of years now this year we just launched a new course called building web 3 business where we were literally teaching web 3 business strategy and so we had to do you know both all the think

again Tim says we're you know talking to Port Co brings us to the Forefront luckily we you know I've been doing a lot of this thinking with teams for a year and a half leading into it um major changes so first of all the sources of competitive Advantage have really changed substantially right if you think about the web 2 business model you're used you're probably all used to thinking of web 2 as you know there are these platforms who lock everybody in and they lock in data and this they build these walled Gardens and web 3 in a large part is about breaking out of that because we think it's value creating to enable data to be individually owned and and to transition sort of seamlessly across platforms but from a business model perspective that's actually a huge shift right because these web 2 platforms are all have

business models that are built around their ability to hold everything in this enclosure um and to use like sort of their Network effects as a source of gravity and usually in some way to become extractive over time right um you know sort of you you have users and then eventually like they're so locked in that you can show them lots more ads or charge them more money to use your services or something of the sort and web 3 breaks a lot that right sort of there there's this classic framework we have an article that's that's soon to be circulated on What's called the five forces there's this classic framework of like the different sources of comp comption uh competition from existing competitors uh buyer buyer and seller bargaining power and then um you know sort of potential new entrance and

potential substitute products all these different sources of of competition in Industry gets stronger once you have interoperability that comes from you know an open platform like a blockchain right now if a competitor wants to recruit the you know sort of your top customers they can see who they are and they can directly incentivize them to switch like in a vampire attack uh you know users can vote with their feed which raises their bargaining power and all of this makes value capture in the traditional forms more difficult right you don't have these locked in you know sort of network effect wall Gardens and at the same time um you know sort of competitive Force just get stronger however uh first of all critically like web 3 doesn't eliminate Network effects

it just changes where they Lodge right so instead of the network effects accre to the platform they acre to the protocol or to the digital asss right the the protocols that are used most frequently the digital assets that are you that people want to use in as many places as possible have a tremendous Network effect built around them um this sort of embeddedness can become a tremendous source of competitive Advantage right if your protocol you think it's a little bit like um like the security that comes in a blockchain right if your protocol is the base layer of another protocol that is the base layer of a third protocol that is the base layer of a fourth it's really difficult for this for anyone to disintermediate you even if they would like you know potentially like to for work because you run the risk of like

everything else breaking above them and then their business falls apart and so we've seen this in open source software pre blockchain and and it's much stronger here where you could additionally add an incentive Loop embeddedness is a core source of competitive Advantage uh and then can eventually lead to Value capture and we're starting to see Brits at that too with with protocols starting to turn on small fees for interactions uh and then another which is to my knowledge basically special to web 3 I don't think I there's sort of a a real instance of this that we've seen at scale before is you can also Drive incentives through Community cohesion uh sorry you can also Drive competitive advantage through Community cohesion by by either incentivizing or or just driving by having people want to engage in your ecosystem rather than a

competitors or something of the sort and shared ownership is a really powerful way to kick off that Loop to get someone invested in your ecosystem and want to be a part of it so anyhow we spent a lot of time I've spent a lot of time thinking about bits and pieces of these different class classic platform competitive Frameworks and then how we you sort of how we transition how we compete in web 3 and then again critically what that means for Value capture and again it's the type of thing that both I study as a scholar and you know you have lots of conversations with teams who are asking versions of this right how do we stop ourselves from getting you know forked or disintermediated um how do we sort of Drive Community engagement in a way that promotes sort of long-term platform growth and sustainability in a world

where users could leave and how do we capture value how do we when it's an appropriate time to start capturing value all of these types of questions as well all right so we still have a few minutes left in the session so I wanted to open it up to the group I mean the research team is around all day today all day tomorrow and and you have access to us beyond that as well of course but if you have a question you think would be you'd like to hear about now from the incentives part of the research group Now's the Time to ask it so yeah uh yeah I'd like to ask you Scott what do you think it seems that with facter for example example you know the incentive is primarily targeting the Creator the content creator and not so much the end

user which you know I think their argument is that the end user doesn't really care maybe I'm Mis misrepresenting what they they think or said you can correct me if I'm wrong but you know you can think that the the user already has as an alternative Twitter and Instagram which they get nothing none of the uh value of the platform or ownership right but at the same time it seems that if you're moving away from giving the end user ownership it seems like you know that's moving a little bit away from the spirit of web 3 which is to you know

share the ownership of the platform or the protocol with everyone so I was just wondering what do you think about that I haven't talked with the founders of foraster about how they conceptualize this so I'm going to speak from my own sort of experience and Analysis of the platform rather than from an understanding of how they think about these different components but I actually think there are if so forecaster does give users a very powerful form of ownership uh and in fact it's a form of ownership that we're actually seeing users with some willingness to pay for which is that they own their their social media platform identity or soci media protocol identity rather and can use it interoperability across different PL inter operably across different platforms and this is like you know very absent from the twitters of the world

right sort of a lot of people tried to leave Twitter to go and join threads or blue sky or whatever and they found they couldn't bring their social graph they couldn't Port it over their content they couldn't Port over their reputation this was very difficult uh and so you know Twitter maintains its gravity even a world in a world where at least many people think the sort of parts of the service have actually degraded whereas forecaster um because people can switch clients freely uh first of all any value they build up on one client sort of is instantaneously viewable elsewhere right if you're spending a lot of time on far.

Quest it also at the same time is like you know growing your warcast you know following an engagement along the way uh but moreover um it actually places limits on how extractive any individual client can be right if warcast is you know if if the team that runs warcast is thinking about you know sort of U showing ads or or or charging fees or something they have to be aware of the fact that people could just switch to supercast or whatever else supercast supercaster I for which way that which which one's the actual name uh and so as a result the user actually has much more power than they do in a typical um social media ecosystem or in a web 2 social media ecosystem uh and so in that sense I

actually think they do have a critical core component of ownership that aligns very nicely with the user purpose of being on social media the reason to do additional work to potentially in a context like that to incentivize creators is uh that creators often have to make a lot of investments before a social media platform is useful to them right you have to invest in building a following you know sort of refactoring your content for this particular platform and so forth and so they're the ones who actually face the higher activation energy at platform launch eventually by the way that starts to flip so so once you have a large enough user pool now you're the user gravity often sucks in creators this is why like now if you're starting off as a as a

fashion you're sort of like a you know bespoke fashion um influencer or something you just go to Instagram because there's a huge user base of people consuming that content but early on social media platforms often have to subsidize creators because of those initial Investments That creators have to make before the platform is useful to them and so I think of it more it's not like both sides actually get ownership of their identity ownership of their user gra like up their follower graph ownership of all their content and and parts of their reputation but when you're deciding how do you launch the platform you start by subsidizing the side that's going to create value on the margin for the other side so that like sucks in users of the other type and where the activation

energy is higher right when you if you if you jump start a social media platform by saying to a bunch of users hey just come um you know there's nothing for them to do there's nothing for them to see so it's it's hard to get them to come if you start off by sort of like seating it with content by incentivizing some high quality content creators then there's a reason for users to show up and that kicks off your flywheel by the way something I think Dan and the team of forecaster did really well with the early scaling and like sort of like very curated early supply of creators um you know specifically geared towards making it a like really positive and high value environment to draw users who are going to contribute to that all right so I think we have time for one question if

it's if it's quick so okay so this is uh just pron of I I I loved what you're saying about adversarial thinking and I just wanted to kind of drill down on one thing you said where you're exploring fee-based mechanisms in the more parallelized computation efforts right where where it's not just ethereum onedimensional thinking I was just wondering if you could speak to just like some examples or like what you're actively researching on different fee mechanism strategies yeah so I I think the one that's kind of most in production now is kind of what Sal is using so TR use in 30 seconds um basically Sona has this model um where instead of making the block constraints just total gas limits they have this account based model and they also have basically gas limits um on the total

number of transactions that can access One account which is a much smaller fraction than the total block limit itself and so this basically implies that kind of accounts that have kind of very high contention basically implicitly receive a higher price from the person who's building the block because they can't actually they have a uh they're more Supply constrainted on how much how many transactions they can include that touch that one transaction basically and so I think this is basically like the like a V1 attempt of basically trying to incorporate um kind of state-based pricing that gets the parallelism you know basically you fill up a block with tons of transactions or all touching different accounts you can paralyze this thing way faster then if you know you have these adversarial things if you pack a block full of

transactions that have to be serly executed then you kind of lost the whole point of paralyzation in the first place uh and so yeah I think this basically increasing the block constraints kind of like implicitly uh kind of affects the fees you need to access certain parts of state and so I think that's basically like the first thing we've seen but there's a lot of kind of ways to go and improving on that idea [Music]