TE Unplugged – Designing Fairness & Efficiency: Incentive Mechanisms in Staking and ZK Proof Markets

Transcript

this presentation, I'd like to ask Franklin on stage again. Um, yeah, and let's kick off our TE unplugged session. As mentioned, this session is meant to dive deeper to discuss dedicated aspects of the concepts we've just seen in a brief presentation and um better understand uh how they differ and also what they share. Uh let's first recap. So we have two principles or two cases where we apply mechanism design. One is in Franklin's case CK proof markets. So the basically an emerging scalability infrastructure

and uh layer uh where we use CK proofs to compress and verify computation particularly for scaling layer 2 networks and for pricing these proofs and matching the right prover with compute available to process the proofs uh with those who need such proofs. we need uh a way to determine prices and to match those two parties and Franklin's paper is about how to m match them in a double auction. Proof markets is like an additional piece of infrastructure for layer 2 typically. Now on the other hand in Tanisha's case we have

validators. This is proof ofstake consensus. uh we want to make sure that we have a wide diverse set of validators who are proposing blocks and submit um approvals. Now these cases are quite different. However, they share that we apply mechanisms designed to ensure we align incentives and we had this term aligning incentives um in the earlier conversations with what about autonomous agents in the systems and how can we uh make sure they don't gain the system in one way. Um so the economic approach to secure systems is to align the

incentives so that any actor any rational actor participates without hurting the total networks. This is the idea of aligning incentives. It's not paying users for doing stuff. It's like make sure that a rational actor acts in the best interest of the individual and the total network and applying or trying to aligning incentives is the let's say the shared principle we can observe in in both cases in DAR double-sided auctions improver markets and for the reputation score.

Now I'd like to start with looking through this security lens. Um and first this question to you Franklin CK proof pricing has been a longstanding challenge and it hasn't been solved or there were several approaches. So we have the transaction fee mechanism Tim Ruffgotten in 2020 this paper we have EIP1559 we have a number of papers who are proposing enhancements to EIP559. [Music] Then we have a different approach when how young atal mechanism designed for CK roller proofer markets where rather

focusing on the profit maximization of the matchmaker. This could also be an approach, not price finding, rather incentivizing the matchmaker. And then several networks try to establish static pricing, so better foreseeable stable pricing for those in need for a proof, subsidized models. We have gasbased cost uh costs of appro uh and also protocol owned provers, stocknet shop for example. So a a broad variety of approaches to solve this problem, how to price CK proofs. Now could you walk us through your double auction DAR solution and highlight some of the benefits?

Yeah, absolutely. uh first I would like to preface by saying that all of these uh alternative models are absolutely usable and that at this point we are more looking to optimize the value extracted for everyone involved rather than to say this doesn't work this is not the case this works it just doesn't work as well so for instance if we take the static pricing static pricing is easy it's business friendly because the suits like to talk to each other sign a big fat contract and bidden with But it kills a lot of flexibility. Like for instance, if an operator, I don't know, through some hardware optimization or whatever can decrease the cost uh the proof cost by 20%. Then the client will

never see the the value coming up for them until the contract is renegotiated for instance. So this is something that is working that is forecastable but something that is not really flexible and that basically kills in some ways the value that can be extracted for all of the people involved. Same thing on the other way. Let's say let's assume for instance that I don't know we have a heat dome over Europe and electricity prices are skyrocketing then the operators will pay 10 15% more for the electricity and they won't be able to make back their cost or to carry it over to the clients.

uh when you take the gas system for instance uh gas system are very dynamic they are very adaptable but the problem is that they encompass such a much larger domain that is in some way correlated but not actually the same thing as what you are doing because stuff that is cheap on gas will actually be very expensive on the ZK system and vice versa and more generally speaking uh the gas prices are fluctuating with the use of the whole mainet used right of the whole Ethereum blockchain use. So it will be hard to match it with the fluctuation of availability of comput computational power or with the prices of computational power and so on and so

on. So this is actually not a bad thing to use but it is I mean the concept is actually interesting but it real implementation is too decorated from what we are doing to actually be useful in our case. uh on F sorry what was the third one come again there are several models I just wonder um um what you just mentioned so both uh tying the price to gas uh or have stable prices fixed prices is like it doesn't push proof providers uh providers of proof compute to maximize efficiency and to innovate do you have a some insights on to what

extents Can they become cheaper? So, are we uh like there's a lot of efficiency gain still to make or um is it hard to you know optimize prices in this market already? That's actually a very interesting question with very deep ramification. I will try to avoid going into too many technical details not to bore the audience but uh I would say that there are multiple ways of dimin let's say that the goal is diminishing basically computational price right. Mhm. So you have a multiple levers to act on that. The first one is energy price. I mean it may sound stupid but actually in the end everything is electricity right? So if you can move your company to somewhere where

electricity cheap mechanically you will decrease the price. Uh that's the first thing. Then one layer above uh you have hardware uh even if we are not in the glory day of CPU evolution where we had twice the power every every six months you still have incremental uh improvement in CPUs. So a proof that was costing X one year ago is now costing maybe 0.8 or 085X instead. uh one layer above again you have the implementation proving system are programs unlike every programs they can most of the time may be uh can yeah can can be made faster right uh memory patterns CPU usage paralization porting from CPU to GPU and

so on and so on so yeah and finally you have the highest level which is basically operational cost right uh if you are a company and you can work with 10 guys instead of 15 then that five five less to pay and that's the the same amount of money will be carried over to the effect effective price of your proofs and finally at the very very top it's uh how do you want to run your prover like a lot of people currently are using machines that are based on AWS or GCP cloud instances but we found practically that these machines are really not costefficient at all for intensive computations on a lot and I mean like at least 50% of the money can

be can be made back just by switching to bare metal computers for instance. So a lot of levers to move and uh some of them are long-term some of them are shortterm and some of them are middle term. Okay, makes sense. One alternative method I mean with your double auction principle you are trying to establish an a completely open market uh for CK proofs. The exact opposite would be protocol owned provers. So could you elaborate a little bit? I mean we could also say well this is a piece of infra in infrastructure proofs should be in you know providing secure proofs reliable proofs should be in the best interest of the protocol. Why not make

it protocol owned infrastructure? Obviously you see some benefits in having a market. Absolutely. Uh I would say there are two main components to it. The first one is an ethical component in some ways. So we like to see information flowing. We like to see decentralization. We like to see basically the community and the ecosystem taking ownership of the tool that they will be using. And the second one is maybe a bit more cynical or at least pragmatical one and it is that company do not have infinite manpower.

Uh we are not all running on Google or Apple budgets. So at some point uh let's uh let people do what they do best uh to extract the most value in the whole process for everyone involved. So if some people are out of the out of league system administrators and can make their machine run 247 without any hiccup then let them do that. Let the protocol developer focus on the protocol and mix and match between these two to generate the more value and use utility for the community. Okay, cool. We can build on this with let's uh contributors do what they can do best. Um and take a look at the mechanism in the reputation score

you introduced Anisha. So reputation models obviously are built on and measured by value contributed and again this is a similarity and dera we have uh the resource requesters match with the proofers based on a cost performance bidding and the reputation core score looks at the per pass performance of validators and then um determines the score and also uh the access um and ranking and picking validators. Now you say influence and rewards should not be directly proportional to a validator's financial stakes.

Explain why. So it's the same principle, right? The more wealth I have and the more I'm able to kind of get my foot in the door and the more foot in the door I have, the more rewards I'm able to get because often it's determined by your stake weight. and that makes the rich richer while leaving the top the lower rungs scraping for like the remains of it. This obviously isn't a fair system and if we wanted to build something centralized we already have the web. Do we have traditional infrastructure that's been built over the years? So we're just recreating history and not really solving a problem. Um that's why I say stake should exist for security in the sense that if this malicious

behavior proven double signing slashing can take place. However, it should not exist for privilege where because I have more stake I get more opportunities to be a block producer thereby allowing me to make more wealth. Maybe it's worth taking a moment to uh dissect a little bit uh when a validator gets rewarded and how they are picked in proposing blocks. uh in submitting approvals. Yes. So different systems have different designs but fundamentally two actions are rewarded. Number one is block production. So this is where in polygon for example you'd have a primary block producer. So they would select me because I had the highest take for instance and they would select him as

the secondary block producer because he's second in line right? So if I don't sign and I don't produce the block he's immediately called upon to act and produce the block. um if he ends up producing the block, he gets the rewards if I get. However, let's say you're on the lower 50%. And you don't have enough stake weight, right? Um that puts you at a disadvantage because then because of your lower stake weight, you don't get called as a block producer, thereby not allowing you to make more economic rewards for yourself. Um yeah, and uh let's dive into the three components of your reputation score. Now, we have an attestation effectiveness. We have a proposal effectiveness and we have a social effectiveness. Let's take a

closer look. How are these calculated or how do you compute the score based on these three aspects and dimensions. Yeah. So very fundamentally attestation effectiveness score is how many checkpoints you signed versus how many checkpoints you missed. So in Ethereum they do it with inclusion delay. They see like you were supposed to include a transaction in block T, you did it in T plus one. If there's a higher inclusion delay, the lesser your rewards. Um, this is normalized over your active time because it's unfair. If I've just joined now, I've attested fewer. So, doesn't mean I should be getting a lower score.

So, we normalize this over your active epoch, which is how active you've been. The second one is block production. Simple. How many blocks did you produce versus how many blocks did you miss normalized over your active time? Also we use something called as probability here which means because I have a higher stake weight I have a high probability that should reflect in the score right so you also normalize it with the probability of producing blocks and the third one is very very modular so at avail social effectiveness score we've defined it as your cloud provider selection the location you come from because we want to kind of reduce single points of failure um in polygon social

effectiveness was more focused on governance participation. So it's it's very modular in two ways. One what's important to the network and two the weights. So I can say 70% of your score is block production. 20% is attestation 10% is social effectiveness. So you can customize that based on how the network reward distributes rewards. Yeah. And uh I'd like to um explore the social effectiveness for a moment a little bit more because I think this is a general observation when it comes to reputation and networks no matter if we're discussing validators or other roles. On one hand you have the key key role or key task to complete uh for

a certain role which is attestations and block production in the case of validators. um that's their their key responsibility here. We'd like to see innovation, efficiency, gains, um diversification. Of course, for the social effectiveness, it looks to me a little bit different any on one hand any ecosystem tries to have active engagement of their core contributors. No matter if we are talk about validators or developers um or applications on top of protocols whatever uh you'd like to have them

involved in governance and this is a new participation is totally a new challenge to solve which centralized organizations doesn't have to solve to the same extent. Now then there's a uh there's this attempt to say okay you'll you'll receive rewards for good engagement for positive engagement. Um and you mentioned a couple of criteria for positive engagement. Isn't there a risk to distract or to overburden validators with activities they should not you know uh in fact which are not core of their value add to protocols 100%. There is that risk where

if you poorly define if you're active on Discord send 100 messages you'll get X dollars ridiculous. Never been a fan of that system because you can game it. However, if you have a social effectiveness parameter like speed of upgrade, where it's critical that validators upgrade their systems, otherwise the network stalls, that now plays positive reinforcement loops in social effectiveness. So, it really all comes down to how critically you think about it. In polygon when we designed social effectiveness the reason governance was a part of it and when I say governance it was specifically how many improvement proposals have you authored not how many votes you've done because the system was

ready you couldn't simply author a proposal it was hard to gain it had a a very structured process so that fell into the social effectiveness bracket one of the suggestions which is how many votes that did not because that could be gamed and like you said I'm not warranted to vote on everything if it doesn't affect me. So why does my reputation depend on that? So that was removed from the social effective. Yeah, this makes perfect sense. Um, some more words on on the implementation process. So the polygon reputation score is in fact live and applied to the uh live system. So it's it's live only for validator admissions which means when you're applying to as a node in Polygon

you have an admission type form that you fill and it has a scoring logic that's open so you can go to a dashboard check it's a graphana board that's public you could see who applied without personal information so you see tokenized unique tokens and you see their scores and you as a community can say this person got a good reputation score why why weren't they selected into the system. Why weren't they in the node like slot? Why weren't they given a slot? It created very good feedback loop from the community for us but rather in the sense of the community has a say or it it gives more visibility to and transparency to activities and not so much to the uh rewards distribution itself. Right. Yes. So reward

distribution the reason we didn't jump is we wanted to do it progressively. So if we know it's working for admissions and these validators are in fact performing well we know the system works we know the score works right and then once we know it you can take more drastic steps like capital allocations taking rewards in fact there is this team which I highly recommend everyone here check it's called as validator.info info. So you can go on their like website, you get all the list of validators with all their reputation scores in Polygon, which is really cool. So now if you want to stake with a validator, you can see their scores and take an informed decision uh because you can see they're performing well and you can stake with them more. So this is

like a newer experiment that we're Okay, cool. validator.info. Validator.info sounds good. Which brings us to the implementation lens. So both design draw from economic theory. So you even are drawing from picod economic theory. Um we have genie coefficient who is applied to all sorts of economic um networks no matter if crypto or not and auction theory in Franklin's case. Um in crypto though we face some particular challenges on implementation and you touched on it Franklin with your diagram that uh you showed last. Um can you again give us an overview of the

particular implementation challenges you face because one thing is to write and verify the the auction mechanism and the other part is how to make it a reality for those who are involved. the three rows you mapped out in your system. Yeah. So the the main issue that we hit of course is the good old how can I say that hypothesis of the spherical core on an infinite frictionless plane. Right? uh theories are always developed within a given frame with hard hypothesis. And when you try to actually port them over to to the real world, well, you you realize that oh well that this

hypothesis actually doesn't hold or oh well we didn't think of that or oh some guy is trying to do some weird stuff or oh this stupid corner case that we saw didn't make any sense. It actually makes sense for that one guy somewhere in the middle of Wisconsin. So uh I would say that yes uh the the hardest part of course is to take the clean nice research paper and actually make it into something that is resilient to the the infinite human imagination I would say. And where are you in this journey? Uh should I really reveal that or am I at risk of being sniped by some people trying to pentest our system? Uh I would say we are close to the end but in some

way it's an ever uh It's really an open-ended problem, right? Whenever you I would say we have something working right now, but whenever you see some kind of undesired behavior emerging, then you have to squash it one way or the other. I wonder I mean Tanisha, you introduced how you implement step by step these algorithms at Polygon. Um are we ever at a stage where it's final and where we can say okay we we uh can make sure that we have found all the vulner vulnerabilities. Um there there is no way to game a system anymore. Franklin what do you think? Oh no I'm saying I'm a very I'm an engineer. I have seen so

many so many unexpected stuff that I'm always starting from the hypothesis that software is never finished. It's uh I mean at some point you just have to embrace the cows and it helps me sleep at night and uh do you see any difference because many people if I talk about economic security say well we have software audits what's the problem we have security audits but economic attacks are slightly different so um how do you approach this to get to this stage where say okay we can never fully guarantee we we are this is an like an open-ended exploration on how to secure such mechanisms. how to get what are

your criteria to get to a stage to say okay it's ready for launch uh when the deadline is coming in no more seriously we are uh I would say uh at some point you have to accept that uh there will be atm there will be I wouldn't even say necessarily attack but unexpected stuff happening bug being revealed or whatever so I would say we're kind of working like a counter intelligence service uh you know that there will be spies in any case so what you are really trying to do is not necessarily to to prevent them from coming in because it's completely impossible. But what you are doing is you are just uh taking the the problem one step further and you are setting up monitoring, you are setting up in inspectability of the system. uh you are

setting up all that kind of uh let's say facilities that are helping you to not block everything but be able to immediately detect when something goes wrong find what is going wrong and have some contingency plan to actually address the stuff going wrong and again this is about aligning incentives is more make sure that everybody is really incentivized um to do the right thing because the the idea is we have open systems So it's not about blacklisting or uh granting access rights like as long as you play according to the rules uh participate and this makes what this economic security a special let's say component special aspect uh of uh crypto security.

Tanisha, you also gave us an overview how you step by step um implement and roll out the reputation score. Your the first component you need is data. So what data do you use for defining a reputation score? So it's it's a multiffold process. The first data is just surveying validators. This is me getting a lay of the land, right? Um this is where we ask them do you think experience has how many years of experience you have and then checking whether there's a correlation between experience and performance. Uh so does that mean that validators with higher experience tend to have a better

performance. So this was the first study. The second was a company study which is what is the vision of the company? What are their goals? So this is where the social effectiveness is modular. Right? For some companies it's diversity of validators prioritized over governance. Um so the lay of the land is that and then the third one is for the reputation score itself. Uh we've had to these are fairly like onchain. So even when I say governance I'm not speaking about temperature checks that we do on forum. It's the actual onchain vote that they cast. Um all of these are fairly onchain and traceable. It's just that once you know where to pull them from you can pull them apply the formula check the epochs that they've been active on and and show. So validator.info info has already kind of

done that. They've taken that onchain data, applied the formula and kind of put it out on their website. Okay, that's cool. That sounds comparably easy to other reputation mechanisms. Cool. Uh then u next step is the algorithm itself. So how to compute the exact score. Um you said it's not implemented yet or is it uh available in other for other protocols? No, it's not implemented in other protocols. And I think the biggest challenge in implementation is not so much data, it's actually the mindset. How do you convince protocols to invest in this?

Because your reputation needs to sit on an identity, something that I can take and show you and say, "Look, this is my reputation in a way that's secure. Um, do protocols want to invest in this? Is it easier for them to run their own foundation nodes and keep the system secure? Do they even want to invite external node operators? I think that's the bigger challenge today. Um, which brings us to robustness and manipulation. Again, there are attack vectors. Uh, so you said, um, for example, simplest attack vectors, if you vote, you can increase your, uh, reputation score. Um, that that would be comparably easy. Um are there other

aspects maybe uh to highlight where it's particularly hard or where you worked hard to make it hard to gain the score? Um so whenever we design like reputation we make it a point to check for attack vectors. This is how we eliminate. We make a list of everything we want and then we start eliminating which is we cannot have GitHub contribution for example PR because that can be GD right um that's how we kind of down select our ideas there are always attack vectors but that doesn't mean you can't go back to change again so the given that you have a good governance system you can go back and propose that I want to add this

to social effectiveness I want to remove this and then you back it up and then the community arrives at consensus and you can change. If there's an emergency, you have an emergency governance track that allows you to change sooner. Um, so there are always attack vectors, but given that this is less subjective, we're already eliminating a lot. All of this is purely performance on chain metrics. It's as objective as they can be. I was just about to say um the attestation uh and block production side is probably pretty much reading activities, right?

Yeah. Exactly. And then when you tie incentive and capital allocation to it, it becomes a self regulating system. Uh because now higher reputation, you're more incentivized to perform for the health of the network. The more you do, the more incentivized you get rather than the most stake that's just sitting there. And um the the final step uh for optimizing such algorithms is always evaluation. So do you I mean there are networks who in fact uh ran AB testing on reputation. So uh for example definition of reputation based on expert uh assessment versus algorithmic assessment. Is this something you a direction you explore? I have I have

read a lot of research papers on subjective interpretation of praise transforming it into scores versus this. Personally, I am more of a objective person. This is a personal choice. I tend to like params that are fairly straightforward. Um, which is why I designed this, but I have read of papers where even subjective praise has been transformed into scores and kind of being looped back into the system. I just don't think we're there yet. And I don't think validators need that. They're very clear deliverables to do to secure the network. uh and sticking to that kind of objectivity just helps with this use case. If I was designing reputation for gamers, my approach would

be completely different. Perfect sense. Um all right, I stop for a moment. Do you guys have any questions or comments or insights what on what you've heard so far? Um let me yeah get a mic. Thank you. Oh yeah, there we are. Thank you for both very interesting talks. My question is for Tanisha. My name is Ellie by the way. Um so when it comes to the objective measures

and I mean so so my expertise is more in the Ethereum validator ecosystem for the mainet which you haven't really discussed but the um and cosmos which is completely different but the um in in this case the objective measures are performed through the code and there is a a a process by which you the clients, the validators are kind of in a process of uh interrelated observation of each other's behaviors around attestations and rewards.

um for people who have attempted to run a home validator solo stakers things the things that you're saying would increase reputation such as you know whether you've successfully produced a block I mean they're limited by factors such as your internet connection not not by your stake you know they're limited by actual things in the uh whether you're moving house and need to move your machine it just so you're suggesting actually in this case that those with more capital who can run them in the cloud via Amazon

you know in data centers or in you know their own on their own servers with massive capital would actually benefit from reputation from what I'm hearing when the kinds of things that we need to be rewarding and incentivizing if you break down the actual roles that validators do which I think you know people like Barnaby from what I have um uh listened to and read of his work are trying to do. you know, what are the actual roles at in a granular way that we should be rewarding and which actors are performing those roles and how do we actually reward security of the network which in this case is rewarding probably home validators much better than we are

now. And then we have factors like liquid staking which there's more incentive to be a liquid staker now than to actually be a home validator in terms of the rewards that you're going to receive. So for me it needs to happen through the protocol itself but the protocol itself needs to be adapted fundamentally to identify which contributions to the network are important and I'm not sure your your ranking system is doing that. That's why I'm curious of your response. So just to rephrase and understand your question better, are you saying that we need to get the protocols to adapt this right within their systems of how they allocate rewards or Yes. Yeah. Yeah. So I mean it is it is already it is already

a reputation system is what I'm saying. The code itself is in Ethereum's case the protocol is a reputation system. It's rewarding and penalizing based on the val the validator machine's behavior according to rules which are set by the client. So and and agreed to through various governance processes. So so that is the thing that needs fixing. I agree but like I don't think Ethereum looks into social effectiveness core does it? like it looks at block production, it looks at checkpoints, but it doesn't have more granularity into skin in the game parameters for what the validator, the speed of upgrades, the uh governance participation, their skin in

the game. Is that granularity existing in Ethereum? So I mean there's a there's it's not necessarily able I mean we know for instance which what the proportion of nodes are and validators are hel are being run by large companies for instance um I'm not I mean in many cases they're I mean it so they're not participating directly in governance unless they're sitting in on dev calls presumably.

um different with other blockchains. I agree. I mean we can have this conversation later but the the point is the the social and technical dimensions are interrelated to the extent that we should be able to deal with this through the protocol design and I I don't think you can say that you know the just because this large company has um is you know has a large proportion of the stake that they're not going to be good at governance. I mean my experience with Cosmos validators is in fact that they care a lot more about the longevity of the networks than um delegates people who are delegating to them or people who are voting

independently often. So I think that you know that it's it's much more complicated than that and if they're performing their functions well that's good. Security, however, is the core issue around someone like Coinbase running a vast majority of um validators for a network, right? And often they won't participate because they understand that conflict of interest in the Cosmos chains examples. So there there's also um Yeah. So I think it you can in in blockchains with Anyway, we can talk about this later. Yeah. So just I'm dominating. No, no, no worries. It's just it's it's good that you're kind of saying that to answer your first part that is the biggest challenge protocols

that's exactly what I was speaking about which is protocols you know embedding this reputation score uh into their systems. The second part about what you said um it's not meant to replace the bigger ones. It's meant reputation scores to give accessibility and visibility to the ones who performing well and also have skin in the game versus someone who's with a high stake but not having skin in the game. Uh not upgrading fast, not participating in important decision client upgrades. That's the difference. It's not meant to replace the bigger ones all of a sudden with the ones below. It's meant to create that accessibility where they can move from one quantine to other pro provided they're performing on all three

fronts. Well, that's the goal to create that accessibility pathway which doesn't exist today. But yes, we can definitely chat more and I think you were referring to Ethereum rainbow staking which is also an new approach on how can we diversify the landscape of validators and in this case it's about having a more granular menu of services this validators can choose to participate and contribute to so that we are not like they compete for very particular services where they are best. And this uh brings us back to Franklin where uh are your best and how to um maximize also the efficiency gains and security

gains in our case. All right, I'd like to close this round with a tiny thought experiment. Um let's cross apply each other's approaches. So, um, I'd like to ask you, Franklin, could reputation, would it make sense to add a reputation layer to define CK proof pricing? I definitely think so. Uh, when you take a look at the the existing real world, real life auction market, none of them are completely open, right? It can be the the bulky guy at the farmer auction, it can be your Swiss passport at SOB or it can be the SEC at Wall Street. But in any case, you always have some kind of more or less explicit barrier to entry.

Uh and this is an issue that we hit as well because uh when we are thinking of things that can incapacitate or damage a network, we are always thinking about some vicious hypers smart hackers from wherever in the world. But uh the most issues that we have were actually coming from what I would call what I would call uh involuntary attacks. So it can be people uh running the software wrong. It can be people mishandling their network stack. It can be people not updating stuff. I mean a lot of issues are actually coming from non problem nonoiley actions. And I think that uh I'm actually completely

convinced that uh reputation measure reputation based barrier to entry that shouldn't be like hermetic and be one for all right. It's more like uh go get out sort your sheet get your ducks in a row and then come back right and I definitely think that this is uh the next step for our own system because uh on the one hand we don't want to lose time and to lose money giving proof to people that cannot actually handle them but on the other hand we still want to give chances to these people that are just clumsy or not really hostile and we need to to give them a way to re their reputation with us so that they can reenter the system for the the greater benefit of everyone involved. So yeah,

I'm pretty sure there are I'm absolutely deeply convinced there is a lot of synergies between the two. Right. In Tanisha, would it make sense to implement a market design for validators? You mean the big bidding market design? Bidding market. Yeah. I mean if you look closely it is that either through your capital or through your stake or through your delegated stake you are technically bidding for a slot. So in a way we're already at that system increasing skin in the game in a sense.

Yeah. But it's also a problem in that sense that now the wealth about the economic angle of it. It's secure. Sure. You're putting in skin in the game. What about the rich being richer and continuing to dominate the network? Um, how can you have both skin in the game and the economic accessibility to rewards? So, that's the real challenge. So, not sure if uh it's a good idea to combine. It is a good idea to combine like stake should exist like reputation is never meant to replace stake. A minimum accessible stake should exist for slashing, for corrective behavior, for security. But reputation should exist for fair opportunities and what that reputation is for different

networks will will will wastely depend on their mechanism design. Any final comments from the audience? Okay, re let's wrap it up. Thank you. Oops. We have Yeah, we still have time. No problem. Thanks. So, I had a question for Tanisha uh also on sort of robustness and security. How do you think about civil attacks like ultimately if your goal is to you know have less giant operators and you know have like more little guys in the game how do you like do you worry about like giant operators pretending to be a bunch of little guys and like how do you defend against that that's a great question so in my earlier talks I'd always spoken about reputation

sitting on an identity system a ziki identity system um I didn't speak about this this time because it was focused more on this but every time I think of reputation I think of it sitting on on a on an identity and we did this with privado ID at polygon. you'd like you have a basic identity system whatever your admission score is would sit on that uh you do KYC of course to get that and then you could use that to enter the system as a ticket as a very valid ticket that's how we've been thinking about civil resistance right any more questions good oh Hi. So, I have a question about

reputation. Um, I want to bring a different angle. I'm more in the DAO management kind of thing. I believe that sometimes when we talk about coordination and we focus on in technology we forget that there is a lot of complexity in the human side of of this of reputation. Um I think there is like a a need to complement some of the consensus layer kind of objective reputation with more soft. I think you mentioned a little bit about how can we um translate more soft uh like perspective uh data into more

objective rigorous kind of measurement. Could you talk and I mean for whom whomever but like a little bit about that. So just to again understand your question right are you speaking about in a DAO situation where there are DAO members how could you potentially translate soft consensus into like tangible reputation scores is that yes I think that's mainly that's a very tricky question so I've worked initially when I was new in the space I used to work for this company called as Dow lens what we used to do is we used to do onboarding so you join a discord you'd go through a really nice onboarding flow say what's your purpose? Why are you here? And then

based on that, we would give you like a tag on Discord and that would become reputation. That's literally the simplistic way we started. Turns out it's gameable and everybody gamed it. So we had like grants tied to it. We had community incentives tied to it. Terrible idea. Learned it the hard way. It's I don't have an answer to your question. I need to understand more about your DAO and the the objectives to be able to say how can we tie soft consensus into like hard reputation that's not gameable. Um it really depends. My favorite way to do that often is vote escros. It's uh not soft, it's definitely like escro contract you lock your tokens in and based on how

long you lock uh how much skin in the game you show you get a multiplier. I tend to love game theories like that. Uh because one it's healthy for the tokconomics and two it's also skin in the game like the longer you lock it's also not necessarily wealth based because he can have a lot more wealth lock it for one year I might have lesser lock it for four years I get a multiplier that might make me make me equal to his voting power or maybe even more. Um, so I'm more into those kind of ways to see how do I bring in some game theory angles with both escros in Dowo systems, especially if you have a token or an NFT.

Okay, great. Thanks so much everyone for this free sessions and the TE unplugged discussion. Uh, I think we are ready for lunch. We'll see you again uh at two where we continue with another two talks and a TE unplugged session. So um happy to have you back and thanks for now. See you at 2 pm.