Panel I Putting the A Back in DAOs I Agentic Ethereum 2025 Summit

[Music] so uh I want to introduce a few more panelists here I want to bring on Sha Kevin auke Linda and Ken um as as our panelist and moderating this panel will be Tina and uh we're gonna get into how do we actually bring back the a& Dallas take it away all right can you guys hear me yep yep congratulation e Global's account has been encumbered by the pirate ship and we have many friends who where we collected a set of questions and where as a dictator I arbitrarily decided there is one and the only one best question will be discussed and debated in this 30 minutes but before we rush right into it you know gotta keep some the question is so good so we're going to start with some uh basic calibration questions um so rabid fire what can today's AI agent capability uniquely unlock in the Dow context and what can it not solve Uh Kevin I'm looking at you starting from you yeah so one of the things that I'm interested in seeing AI agents solve is Dows have a lot of context passing around in their discords and their governance forums and uh my stats on the getcoin governance Forum are pretty insane I have 250 posts in three days of read time uh but you can't expect that of everyone that's involved in getcoin Dow and so what I'm really interested in AI agents uh able to do is is basically pull out the information that's relevant to each participant and give them a customized summary of what's going on in the Dow so that they can have an overview of everything that's happening all across the ecosystem so that's where would like to start using AI in the Dows that I'm involved with I think I think we can kind of break this down into two parts um one is like generally um all organizations I think have certain communication problems and Dow's just kind of extrapolate that because they're like we're all implicitly remote there's lots and lots more people in like the slack than just your small team so to speak um and you have like this lot of like one to many communication relationships so generally like like could imagine if you're part of a of a really big org you're kind of shielded from a lot of the the um communication that's happening by hierarchy and when you get rid of that hierarchy and everyone just can be in any chat then you start to get a lot lot more communication um and I think it just becomes untenable at some scale to be able to sift through all of the conversations that are coming at you especially people can just like start a conversation with you at any point um and I don't know about you guys but like I get way too many messages and I just cannot process all of them every day um it could be my full-time job and it still wouldn't be enough so I think that that's like definitely uh part of it is that like AI are not necessarily right now like it's not the best thing for every single level of the stack but definitely like like um to auk's point like if you could summarize the important stuff that's happening hey here's the things I really care about anybody who's like talking about things relative relevant to me or or like to these things that I need to know like please just make a summary of that for me every day and give that to me like that would be like the bare minimum but I don't this is like really a dow problem this is just like a you know General like data information flow problem of organizations um and I think that probably traditional corporations have like found a way to work together that isolates this problem but that leads to a lot of like different kind of problem in hierarchy where like people don't see what the other teams are working on and stuff so it's like you know like any kind of thing where you have like a work group model you probably want to know what the other work groups are actually working on so you're not double working um and so I think that's just like the general organizational s and because we have this unique feature of tending to like congregate in telegram or Discord or these like like easy to throw a bot in there and track what's going on kind of spaces um there's like a it's like a nice sandbox for solving a lot of these problems that you know might not be so easy to solve if you work in like a cubicle in a large company um but the other part of this is like like what's unique to Dows um well you like the stuff that's really um there is more like true autonomy for you know for like well if we're starting an unincorporated nonprofit Association and we don't want to any of us to take the responsibility for that saying like Okay cool code is Law and and here's this system that's actually extremely Dynamic and can do all kinds of different stuff and I I it's not me it's not any of us it's just performing administrative function has like obviously a lot of appeal um and I think that you know a smart contract by itself just it's very hard to jam all of the different externalities like all the realities of our of our situation into that thing um but you could you know like uh if you can have an AI system that can sort of smooth over some of that stuff especially turning like unstructured just people chatting into like a structured decision um then obviously there's a lot of appeal there so you know I it's hard to say like when will AIS make decisions for us but certainly like the data information like processing problems and as well as like kind of collecting and normalizing things into a usable state for decision- making are like you know very obvious um so I think a lot about the intersection of AI agents and work um because we built a bounties platform on farcaster and so um something that's really interesting is uh you can have this agent proactively coming up with what are things that are needed to get done in this Dow it has all of the context as everyone mentioned um and it can say like here's what's missing who here's the best people to work on it so it can start coordinating people can then start raising funds and saying like oh wait we actually need a budget to create this and it can like Speck out the entire process and like break it up into different pieces so it's kind of like a project manager and then it can just start posting these bounties and people can now build up reputation for this work so it's just this like all- knowing project manager and it can just Pro be really proactive in getting things done so we're pretty excited by that and think that like basically every Dow is going to eventually have some version of this that is going to be um like a trusty little sidekick and people can also be like oh um I'm I'm looking to earn in this D like where's a place I can get started and it just talks to this agent and can be like here's the best place where you can get started these are the things that are relevant to your own skill set and things you've posted in the past so super excited by agents and Dows I think uh to build off from this point I think it's really cool to even think about the other way around of like what are the what are the outcomes that we want to see and how do we actually build backwards from there right so I think one thing that's been maybe this is uh maybe like jumping too far ahead but I think like something that's been really cool is thinking about like retr funding and how um you know people want to be rewarded for the work that they do uh but who evaluates that and I think when we think about like um sometimes politics of play like who's there how we actually Define like what proper outcomes are right I think if we're able to sort of um more objectively look at like what are the ideal outcomes for the organization then we can kind of think like okay well who has actually made significant strides towards this right if there's a spec and we're all building towards spec right like who has actually contributed to it and how can we actually work backwards from there so a lot of really fun things I think we can do with uh incentives one of the things that oh go ahead valalik one of the things that Shaw you and I had had talked a little a little bit about was the idea of taking retr funding which is currently happening in some ecosystems on a quarterly or every other quarter basis and starting to do it more on like a weekly basis and the idea is that would provide a tighter feedback loop where contributors would get record rewarded more often than they would uh on a quarterly or every half year round which allows them to justify uh spending more time on the Dow and gets them more feedback about what they've been working on and so just tighter feedback loops on uh rounds that are retr funding is an idea that we've that we've sort of kicked around and I think it'd be curious to see if anyone's going to be piloting that any anytime soon I mean certainly we've been exploring that I know valal you were just presenting on that yeah no I mean I think like I mean I've suggested to optimism I think Le at least a couple of times that I think there should be like part of the uh retro process that is just automated and guaranteed and according to kind of longstanding rules um and the reason why right is because the whole underlying theory of retro fund is that it's supposed to motivate people to work or to invest by the expectation that work is going to get is going to get rewarded later right but then if that you're basically just like randomly doing rounds whenever you feel like it and like sometimes um you know the team um messes up and they like blow up their scope and like you have the equivalent of Petra and then it turns out the round is like seven months late right then like it doesn't give people a stability that they need right so uh I think making these things more automated more institutionalized making the commitments to them uh be more of like ongoing commitments ASO as opposed to one-time donations like I think these are all steps in a good direction and that I think like yeah we should be moving towards real time so um Kevin you'll be glad to hear that uh our entire incumbered hacker house is being will be retroactively evaluated from code including private dependencies to um social contribution to buyers so we'll see how this experiment can out in the next month well moving on to a uh well another calibration question I didn't expect that one to take uh quite an interesting turn um what do you absolutely not want to see in Dows today using automation uh using what technology we have that you consider okay this would be a bad idea um uh limit a Time scope to one year this year 2025 what you probably would not advise uh us trying um I could just talk about in terms of uh we had a bunch of AI agents posting bounties on our platform um and then immediately a bunch of people are farming the agent for money so I would not advise like large sums of money being given to the AI agent to then be uh given out maybe we could do like really small experiments but we've kind of seen with some of the um cases in the past where it's like hey like under no circumstance can this agent give out money and then people are able to kind of like Farm it or U make attempts so I would just be really cautious around like giving this one agent like millions of dollars to be giving out to members of the Dow I'll go next go ahead Tina um I'll I'll just start with something that I'm already seeing which is only a mild annoyance but um uh I'm already starting to see AI slop in my replies and on my governance forums and um I just think it's such a sign of like disrespect for my time that you're sending me these shitty spammy like AI induced uh uh messages I'm talking about like the Royal Wii like sending the the them out to people and um I feel like we had this fantasy with AI that it was going to help us CX collapse context and pull more signal from the noise but what I'm seeing in the meantime is just like more crappy takes and more like low effort Bots um and I'm feeling more overwhelmed by them and so I'm I'm hoping to see more High signal Bots uh and a move towards um Bots that are mindfully and thoughtfully deployed to add value as opposed to just like trained on the worst use cases and and increasing the amount of spam going across the internet so um that's something that I don't want to see I sort of think that the uh the uh in a world of slot Bots maybe the most valuable product is a slotbot filter and um definitely like AI can kind is kind of like it's a fire hose but it can also be like a funnel um and we've seen just more of the value in like especially when nobody has a funnel just being able to fire hose it um but I hope that like by kind of accelerating the um annoyance of that are not so smart that they're indistinguishable from humans yet it kind of causes us to start thinking about the oppos like like you know they're sort of like red teaming The Social Network um how would we how would we legitimately stop that um and I do think that there are solutions to that but they sort of require us like starting to build those tools and like maybe even figure out how to make money off of that as a product definitely I want like an an AI version of Twitter that just filters all the crap and just shows you what I want to see you know um so the one thing I've really enjoyed about Dows and like sort of in in in the same vein as like everything you guys just said there's like a there's like such a a human um uh part of like organizations and and how you know we come together to make decisions and how we work with one another there's like social contracts as to like responsibility accountability and I feel like um we I can see like trends like starting to kind of head into this idea that like you know everything can be objective everything can sort of just be like abstracted away and like decisions or decisions um and it's really tough when you sort of are a part of a dow and you're you know uh you're engaging in these like fruitful conversations about what you want to see and like what as we Vision setting and North starring um how we can make decisions together uh it's really and to to Kevin's point and to Shaw's point right like it's really hard to to like go through like a chunk of text of just like I I know you just sort of like uh transcribed everything and then sort of like asked chat gbt like hey like how do I best respond to this to like get to this answer and and it's and I I get it right it is it is um it's a lot it's a lot of information but I think it's really I guess like we're not I at least I don't feel like we're we're at a place yet where within a DA to um just like governance in general like being able to have uh fruitful conversations um without like I guess like empathy uh I don't know maybe that's the wrong word here but it is it is tough at least on on my end to to have engaged with uh uh without with a without like a some like human input and insight and like responsiveness yeah well thanks for sharing and understanding the limitation I was going to ask one more uh calibration question than I realized I may be running out of time uh so that'll be um uh we can always uh uh do it next time so the main question that actually we selected from our Hacker House is what is the a in Dow is it um is the desired ideal of a doubt a autonomous or just automated are we truly striving for autonomy or are we simply automating the existing power structures so I think I personally am more interested in autonomous than I am in automated right and uh I think I mean this is like one of those interesting divides right basically yeah where like if you think about what the word autonomous means in any other context like if you think about say a territory of a country being autonomous right like that means like the control over the thing is is inside the thing right automated um generally means um like basically that there is no humans involved in the in the operation of the thing right and like if something is fully automated then there's a question of like you know is it even an organization right and you know to me the ultimate goal of this all is to um like satis satisfy the goals of the people who are uh who are creating these things and I think in any realistic case that's going to involve some form of constant involvement right and so yeah like to me the like the ultimate driver of all of this ultimately is you know creating power structures that allow power to be distributed and that that allow power to be held by know groups of people and different communities that care about about different things and for that to be compatible with efficient governance and I think AI is an excellent tool to enable that any other takers which a do you prefer or do you agree or disagree I would like to see yeah I mean I I think it really depends on the organization itself I mean there's I've been part of Dallas where it's like a few people and then I've been part of optimism where there's you know two different houses and there's so many people so um I think like having things be automated when you can't keep up with all the information um that's really helpful um but if you're just a really small doubt I don't necessarily think you need to like automate away a lot of interactions with other people um in terms in terms of the um autonomous part um I've kind of thought a lot about the like oh does this agent vote on people's behalf um based off their past voting history and what they care about and I think that gets to like a really scary point so I'm I'm a fan of like Hybrid models and like Progressive autonomization um just to make sure things are going well because I could see it being very dystopian when it's just like okay it's all these humans and their agents that are voting on their behalfs and we've now like trended this D towards some really strange outcome so hybrid's pretty cool I've got a take um I think it's hard to talk about Dows in the abstract because uh we're talk about such a a a big design space and so when you're talking about autonomy and Dows I look at that as sort of a progression more than uh a zero or a one more than a binary and same thing with decentralization these things are a spectrum and so there are people who disagree with me and say Kevin those aren't Dows if it's not fully decentralized and auton autonomous that's not my definition but like imagine asking that question about organizations you would you would say like oh are we talking about a church or a family or a company or a government like those are all different types of organizations and they all have different needs um for them and I think the same is true of Dao of DS like I believe that the ethereum layer one should be very decentralized incredibly neutral and decentralized because the issuance policy should not be able to be changed eth's money but you know something like blunt Dow I don't really care if that's autonomous it's just about like you know meeting up in the park and in like rolling blunts So like um there's just like a spectrum and I think it depends on the use case of the Dow whether it needs to be fully de centralized or autonomous and whether we should progressively uh automize or decentralize these things over time but now because we have a theum we can we can progressively uh decentralize and autonomies autonomization well uh then let's dive a little bit deeper because I like um how we're bringing up different type of organizations and maybe there's a dial between the different A's um the automated um versus autonomous um let's just take a concrete example let's take ethereum do you think that ethereum is a doubt raise your hand thumbs up or raise your hand if you think ethereum is a dow in your mind conceptually there is no taker okay let's go all right so I am a taker of of ethereum is AOW um great so uh and uh in this thought experim let's say that three years from now I'm not going above three years because I think we have different uh timeline estimate from a a safest versus a accelerationist approach so we're not stepping into that territory I think three years is pretty safe as a thought experiment okay what would um what would ethereum look like and this is does not apply to this is not limited to just the technology also in terms of the society the community and the impact that we can bring so given that we're talking about uh putting the a back into the Dows it can be putting the a back into all of the dehydrated Dows because uh the Dow toolings didn't work uh in the previous generation and rehydrating them so that there's blossoming Dows you know uh uh all over the place or it can be just like um you know a lady uh uh uh um Mady AI CEO RAB coordinator um that was a hackathon uh project that was really great um right so it can be anything so how would you apply the uh uh the Dow technology events onto the ethereum and what would you like to see happen in three years not one year three years okay I can start uh oh no vitalic should have start no and I'll just say I would love to see know like large scale DS that are uh funding big portions of like all open source software and Hardware um yeah for me the reason why I didn't necessarily think that it's a dow like one piece is um like there needs to also just be like some pull of funding that people can tap into to then self-organize so it just feels very decentralized right now right you have the EF but then everyone is building their own ecosystem too um so in some Ideal World um ethereum community has collectively decided to deposit funds whether it be like a percentage of fees received by each of these protocols or whatever there's like a pool of funds that some um group can kind of act as delegates for there's maybe agents involved that can help um automate some of it um and people are able to propose different initiatives and kind of like what's happening in Dallas right now and they can just like hey I want want to run like a marketing campaign right now for ethereum um here's my plan instead of going to the EF where there's you know um it's like a a specific organization this is like part of the Dow where a bunch of community members can sign off or agents can sign off and they can like keep track of all the work um they're more fair they're not just giving out stuff to their friends necessarily if they're agents so I would like to see some version of that because right now like as a small project we've had to like go out and just like apply to a bunch of different grants and we're kind of navigating the politics of it and all these things but if it's just like the Dow and it's like agents like I think that's a pretty cool world to be part of I think generally um personally I just like the idea that oh I have an idea I have an open source project I'd like to figure out how to administrate it and manage a community around it and if I could like oneclick all of the tools so that I don't have to then go and like hire a whole team and focus on that I could just continue to focus on building um that would that would just make me happy personally um and I think that it would lower the for people to become like for Builders to um you know run these things like like profit Mak organizations there's just like a lot of work that goes into making an organization that makes money um and and you know so like a lot of that is just operational overhead um and if you could minimize that as much as possible I think that you would just enable a lot more people to kind of jump into building stuff um you know that they that they want without having to get totally sucked into like you know being the CEO not being having time to build or whatever the the use case is you could sensibly have much smaller companies doing a lot more output um so you know maybe I'm selfish but that's definitely the world I'd like to see and so a lot of the governance tooling to me is just like General organizational Community Management tooling Etc um and it doesn't have to pertain to a dow specifically but I think that the Dow is kind of just a representation of this like lightweight you know permission list press a button and get it and if you get the whole tool stack with that to to manage that effect L um that would be you know that'd be the product I want to see yeah sha I I feel your pain on that uh having been down this road for a couple years and um I think you're taking all the pressure uh all around you really well so um but I but I have some of the same needs and um you know one of the things that I think a lot about is how can we R uh rally resources around ethereum public goods and make sure that the dependencies of the ethereum ecosystem are funded and the public good in the ethereum ecosystem is funded and so um I don't think we I think we all think ethereum is not a dow but if it was a dow you know I think that like funding the ecosystems public goods would be uh part of the major Charter of the Dow and um uh I'll just show really quickly the getcoin grants 23 is coming up in March and that's the 23rd quarterly uh public goods funding round for the ethereum ecosystem that getcoin runs so not the ethereum official Dow but but a dow that does organize ethereum public goods funding and uh for the first time ever we're moving from quadratic funding to also doing retr funding and conviction voting and futarchy and uh I think there's even a proposal to do another deep funding uh during G23 so we're going multi mechanism and hopefully we'll be funding ethereum public goods and demoing what some of this ecosystem support software could look like in a dow era but doing it for ethereum because we all rely on ethereum public goods to to make it happen so um public goods are good and we want to make sure that ethereum public goods are solved and I think that if ethereum was a dow that would be one of the things the ethereum Dow would do I uh I agree with what everything everyone said I mean I feel like I don't have much to contribute here I I guess like funding is is obviously a big part of it I okay can I if we have time can I throw a hot take in here I'm gonna play The Other Side okay so what if what if in three years we actually see like block building and like M optimizations like just like solvers just all go uh like full AI agents right and and um I haven't fully thought this through so please uh please don't roast or actually roast this so if if we think about this right what if um like you know one of the things that I think has been really interesting within ethereum is that like like unfortunately or fortunately you know For the Love of the Game like it is highly politicized right there is a lot of social contracts and L social construct of like you know who is um you know who is uh who's eth aligned right like what what does alignment actually mean here especially within like the the PBS space But at the end of the day right like it is all optimizations for how much like Revenue can be like seeped out of like you know different parts of the stack and I think if we are thinking about optimizations from like a financial perspective we move everything into this like almost asymptotic like uh extraction point where you know you can only optimize for so much you can only extract so much um to that like you know it it just has Market pressures away from wanting to even just like have uh greater uh higher value transactions um so I guess as a thought experiment like what if we actually have ai agents kind of like take over all of the block building relays stack and uh then you know the optimizations kind of go to zero and then we are you know to to as little as possible and then you know all the all the social sort of like oh well you know I'm ethline so I won't do XYZ uh kind go to zero because you know everyone is just really competing for like as little our profits as possible and you know that kind of moves away that would be fun to see stop I love it I was uh uh about to ask the same question if you didn't say it I'm gonna direct this question to First metallic is this feasible um I mean I think lots of things are feasible um you know I me okay I think like okay so for the public goods funding side of things like the like public goods funding in general is like always two problems in one right one problem is uh how to direct the money and then the other problem is where to get the money and like that you are in interconnected right because people are going to be willing more willing to give you money if they think that it's going to be used SA right but like the where the money is coming from is you know like one of those other pieces that I think is also really valuable to work on and that I think we really you know like haven't gone into yet right and I mean obviously you know optimism has uh done a lot with their rpgfan has heard of is octant it's basically a group of ethereum companies that are you know staking their e and giving a portion of the revenues to public goods funding and like I'm seeing different groups experiment and like talk about these uh like different structures where you basically have you know consortiums of people or projects and then there kind of pledge to fund public goods through a common mechanism as part of that uh so I think uh like figuring that part out is important and then like one of the things right is that if you want to get like small scale then you can just get like individual donations and like individual projects small projects right but that if you want to go large scale um and if he wants to get to like larger and larger tight scale uh protocols and especially if you want something to be able to get to sort of get you know credible neutrality points while at the same time as funding public goods like you really want the mechanism to be know un exploitable to have a good track record I shouldn't say unexploitable like clearly Limited in exploit exploitability right and so like this is part of why I think like I don't really believe in one master mechanism and I generally believe in multiple mechanisms where each of the mechanisms itself sort of has a lot of heavy smoothing in terms of uh Distributing funding and distributing power and I think a lot of the yeah mechanisms that we've talked about are good at doing that uh so the more like I think that's uh the sort of thing where we'll be able to climb up the ladder over time I mean or who knows like maybe someone will just like make a new animal coin and it'll go to $20 million and they'll just like stick 60% of the supply into a public goods Gadget you know anything can happen right I mean is that what I did um let's see is it up to 20 billion doll [Music] and well 20 is a high bar um oh wow even yeah Tron is not at 20 um Lio Stak each is um no no come on like you know you need to have like scale especially if you want to like compete with the big players and I you know especially on like the big budget stuff whether it's you know like Hardware or know like replacing World Health Health Organization or whatever whatever right so but yeah I I love the fact that uh you actually broke down the two side uh which is like well uh uh the money generation side oftentimes is less has been less effective in the past um now it seems like with the uh at least with the automation not even talking about autonomy yet that the Surplus could be Unleashed uh like uh technically even if we um keep do the same things um seems to be greater um and that we can we we can actually experiment um um at a greater scale um with our mechanisms to fund what needs to be funded such as uh free open source software but is there a seiling that we haven't really Sol for in the circulation or the Regeneration um of uh where's the real yield when is that going to come right like I mean like a printing um animal or um a a cartoon Avatar coins are not really sustainable there still needs to be real you so where do you see the a help us unlock in terms of a real yeld yeah and for that ultimately yeah you have to provide some kind of you know real service that brings real value to people right and that's obviously you know a challenge to do with these purely virtual Contraptions that are basically yeah like operating over these Network effects um I yeah like I definitely think that know whatever the solution is it will be something that just comes out of complete left field and will be unexpected right um you know in in 2021 nobody was predicting dog coins and so we got dog coins um in 2025 you know or 27 you know we're not going to get dog coins um what is the thing that we're going to get and I think it's uh you know in an open field and uh like what still a lot of possibilities I think we should give credit sorry oh no I was going to say Like Mine mine's maybe like slightly more hot take on this but like I think it's also a mix of like changing some like social expectations and validations so um people love to do things um and feel good about their social status and how that um impacted their um view among their peers so like people buy designer Goods like lvmh is one of the you know biggest companies in the world people care about social status so I actually think that like um instead of Leaning away from this kind I actually think there could be like you know like awards for like oh which companies like contributed most to the public goods and like these are actually like widely like it's like watching the Oscars you know or um it's like winning a Nobel Prize and so like just or if you donate a certain amount you get this like badge that's like you know on that's widely displayed in like very public places so I am actually like a big fan of like how can we like also add some like social pressure and validation elements to it like um there's like the giving pledge where it's like oh you're a billionaire and you donate certain percentage you get to attend this like Conference of all the other billionaires that donated and they actually like really like doing this and that's why some of them actually do it I've I've talked to some of them before and that was a reason they get to meet all the other billionaires so it's like I I think that um we can't like humans are just going to naturally like this kind of stuff and I'm I'm okay like leaning into that in the um event that it helps public goods I I also think that the other interesting phenomenon we've seen is um like uh and we were kind of part of this is like the bootstrap thing of public goods with meme coins where there's kind of a mix of like it's fun and it's kind of edgy and then there's also like a mind share component like oh other people are also using this and I think there's definely like a lot of price Discovery and you have people who like don't know anything about open source um like speculating on these projects at scale like had people like take my code and put it into Claude and be like well Claude said it wasn't good like okay um you know uh there's been a lot of funny things that happened but we shouldn't discount how much fun is part of this and that like um in a world where attention like is really a big fundamental to getting people to know you exist and to come work with you on stuff and to even like be a contributor to your project like like building in some of the fun and virality to that is definitely like valuable and I think maybe that's kind of priced into meme coins a bit like okay well you know maybe there is like the the how this generates revenue is unclear today but if there is enough attention on it and enough like smart people working on it then it kind of like you know then maybe they'll figure it out I don't know um but definitely there's a lot of speculation and what I did not expect this year was how much speculation on open source projects specifically there has been and maybe there's more public goods could go in that direction and maybe agents could kind of drive a lot of that if that makes sense yeah um I could give one um specific idea which is uh so you basically take things like the protocol Guild pledge and things like the GPL and you stick them together right so what I mean by things like the protocol Guild pledge right is basically you know there's a bunch of projects that are pledging to give 1% of the revenue 1% of their token to protocol Guild and and I actually think that and it's it might be valuable also to create pledges that are kind of larger but also more generically scoped in the sense of uh like the category should be public goods funding mechanisms where um that are rules based um and where both the power and the recipient base are highly distributed right and I actually think like protocol Guild satisfies those properties reasonably well and uh like even quadratic funding rual funding deep funding and all of those also do but um you know a five person Grant making body where like that just decides everything does not right um and then so like ideally we should have multiple mechanisms right but then you the nice thing about pledges is that I mean I think as we've seen with open AI it's very easy for projects to and have good intentions when they start out and when they're very small right and you know it's still very idealist Kumbaya we're going to improve the world but then as we saw with open AI you know first they gave up openness for safety and then they gave up safety for world domination and now we haven't either right um and so we you know we want to avoid the like you want something that like sticks right and obviously open AI board structure ended up failing that test but like the nice the more you can create things where like you make commitments early on and I like the fine fine I I the opening eye yeah like board thing ultimately failed but it did add some friction right but you know the more you can add commitments that are kind of early on and the more those commitments can be things that like stick and affect your behavior to some uh even as a thing becomes uh large uh the more I think like it you actually can create these like long-term sustainable streams and then on top of that it's so this is where the GPL part comes in right is that basically I think we should figure out a way to design the pledge in such a way that one of the things in your pledge is to somehow preferentially interact with other organizations that make the same pledge so and then if you do that then you know you're basically creating this like ecosystem of projects that are actually sort of creating marginal incentives that try to reinforce each other right and so you're you're not creating something that like feels very coercive where there's like a hard you know like 100% tariff sort of boundary between the inside and the outside right like you want like a these overlapping sort of recursive bubbles that all point in different directions to different things and the more your thing commits to having long-term incentives that are aligned with other peoples the more um you you know you get benefits from those other projects as well right so that's like if we can set up an ecosystem that works like that then I think that's also another way to fund lots of things very sustainably and then on top of that the pledges don't just have to be about funding the pledges can also be about things like hey things you do should be open source I just love how I heard some familiar terms that uh you know may have been a uh uh popular discuss in this Hacker House recursive value attribution and also cross holding of a uh uh of upside well sharing up the upside of multiple groups of people coming together uh and uh yeah I think there will be a lot of uh experiment in this direction so that's exciting that said I think that uh cark reminded us there is an excellent um audience question um which is um what if increased aut as in the a in the Dow context will lead to um certain negative externalities from disgruntled agents but agents are not just robots like you know we could be agents too like you know agency um there's a long Theory study in terms of like agency from uh the line of work in literature and economics but like this question what is motivating is that there's many things we can we have seen happen in the past by other humans in other failed Dow um Journeys burning token taking over Dows polluting communication channels all of the stuff we already know but so who holds the responsibility when self-destructing actions are made endows by nonhumans who may not be sensitive to financial incentives inside Orcs I know ital you wrote on this kind of like who's responsible for running the models essentially right right yeah oh wow this gets into one of those really big cans of worms I mean okay so there's like the whole kind of legal engineering question of like how to create liability rules for this and ideally like make those rules in a way that doesn't like basically incentivize like centralization as the most the easiest form of risk management um and then there's also the question of like how that feeds into social norms because I think in general you know the more that we can lean on uh on social norms rather than liability I you think it's like healthier up to some point um the yeah so like for if for example um you know you have a rule that says that like you're responsible for something that was done by a model that's a like where that's a downstream of your actions where you're kind of the immediately Upstream actor then that inherently makes it more risky to do like very extreme autonomy things right and it makes it inherently less risky to do things where you continue to be in um like in regular control right so that's like one of those types of POS options I think another access that we can lean on is the is uh as software developers uh we can uh also try to work on software that makes kind of human in the loop approaches to AI that like actually work more efficiently right because I think a lot of the time like the reason why the you know like full autonomy meme is very appealing is because like like it appeals to laziness and like we have very crappy ways of like actually providing input whether individually or collectively to a lot of these kinds of constructions right and that's like not really guess something that you generally see in in any of these AI toolkits as so you like better work on just any kind of human interface technology is uh one of those things that I consider super high value um I mean otherwise like there's a lot of a lot of different levers you know it's a it's a hackathon we can figure things out well uh just following up what pallic said sha uh as someone who's in this panel working having been working on the agent Stu what can we do today to actually better defend our Dows from uh the increased uh autonomy potential autonomy of uh of agents and these kind of threat I think we're still kind of in theoretical land um where like we kind of vaguely know that these these tools are going to change everything and we just have to do a lot of engineering So like um I I kind of would say that like accelerating the development and like just getting to the point where there are vulnerabilities we have to explore like nobody's really gotten hurt from an AI yet I'm sure that will happen at some point and I'd argue that maybe happening sooner than later would reduce the risk of catastrophe by sort of like pushing people in the direction of like oh maybe there's like problems here um you know you can say okay well somebody like downloaded replica and had a weird parasocial relationship but I just don't think it's gotten that weird um so in some ways I think like we need to like actually explore the the the space uh and like try to solve some of these problems practically and I'm sure that a lot of risks will reveal themselves and I think there's a lot of like vague risks from the kind of safety is community of like You Know instrumental convergence or these like very big long- termist kind of ideas um but the but the real problems are like super practical you know and um and there's this kind of ongoing like well can we actually prevent an AI from getting red teed or from being prompt and injected or whatever and um you you know I think that that there's probably like the more prompt injection becomes a thing the more incentive there is to solve the problem and so it's like kind of creating incentives for the right solutions to emerge is probably what we want to see next I'm an accelerationist you know I guess in that way uh what about Kevin uh you have so many mechanisms on your U Alo mechanism Alo Capital it's fascinating I it was a rabbit hole so like What mechanisms do you think we can use uh to better defend ourselves yeah well uh I was going to say maybe we should only accelerate the defensive things and the decentralized things and uh decelerate the offensive things and over time the defensive things will win someone should come up with that philosophy um so yeah uh I'm just posting vitalic came up with this philosophy of diak which I think is like a great way of thinking about these things like accelerating the defensive techn Technologies and in Democratic and decentralized Technologies and I think that um that's kind of the way that we want to go we want to accelerate uh defensive Tech and the stuff that has a huge blast radius and could cause existential risk to humanity um is the stuff that we probably don't want to accelerate as much and um that's easier said than done I think though because it's hard to know when you're in R&D stage whether a tech you're developing will be uh defensive or offensive but um I'm happy to talk next at the object level about Capital allocation mechanisms since that's what you asked about Tina or happy to volley this over to italic to talk about Dak if it's that's if that's interesting to him either way works for me first capital allocation okay cool um yeah so uh um uh we're spinning out a new Venture from gitcoin called allow. capital and it's just all about different Capital allocation mechanisms gitcoin got started with quadratic funding and I'm super thankful to Glenn W and Zoe hits and vitalic for inventing that mechanism but we've just learned that uh there's a ton of different mechanisms out there for funding your ecosystems public goods and so um I've been spending the last year more or less cartography the different mechanisms uh from quadratic funding to retr funding to deep funding to conviction voting to cookie jars to AI uh Capital allocation to social media Capital allocation there's just a smorgus board of different ways that we can fund what matters in the 21st century and because of The Primitives that blockchain gives us us uh decentralization credible neutrality transparency uh corruption resistance we can build better systems for funding public goods in the 21st century and to me that's one of the most exciting design spaces to be working on right now and I think that um not only does it make us more effective as an ethereum ecosystem to have funding of what matters but also um I think it's going to bring the autonomous and put it back into Dows because um I think funding the public goods of your Dow uh and DOW governance are kind of the same problem they conver Verge in in many important ways so um you know and and I think that like once we prove that it works in ethereum then we can go to the Department of government efficiency and say hey can we put some of the state functions on chain and do them more effic efficiently and effectively than ever before and so um yeah this is a really a beautiful Frontier for me on chain Capital allocation and we're going to explore it over at allow. Capital uh which as I said just spun out of gitcoin so um if if we if things go well you will see us building a Swiss army knife of many different Capital allocation mechanisms and then uh people who are building ecosystems uh will be able to use those tools in order to uh autonomously fund what matters in those ecosystems so it's a very exciting design space for sure [Music] Ken coming to play us off the stage with on so much money Capital uh right with before launching mechanisms what are you going to enable you have one minute um okay heavy optimizations for uh financialization of all Market micr structures I think when we think about um Unis swap you tend to think of you know this this old uh not old but like you know Unis swop is is a DEX but with V4 you really have the opportunity to build much more uh robust and comprehensive micro structures or Market structures for uh each each hook pool can itself be its own amm and so when we think about that as um as a financialized economy you can think about how uh each one lives in its own ecosystem but how they all compose we lost him it looks like a this is a let meure it was not I'm not I'm not that cool a it's a but uh yeah well I think that the oh oh back finish the sentence shut me down uh Mark micr structures but okay so so what I wanted to really touch into is is uh you know I think hooks are really great uh within unisoft V4 to build um build uh new Mark micr structures like how you want to customize a certain pool how you really want to customize how people Swap and then more to that is really sort of the larger interop picture for all of ethereum right how does Unis swap as you think about liquidity existing across chains um how do solvers plug into this and how do routers really kind of uh fill orders as you go throughout uh the different structures across ecosystems and how you execute across chain uh so that is where I think heavy optimization come from from um from an agentic perspective and I think it it's really cool to really think about unisoft not just as a uh as a DEX but really as like core infrastructure for um executing transactions all across chain stop there yeah lots of experiment and uh since uh I think that balic has written a really excellent multiple excellent blocks on Dak uh and uh so I think that it's probably best to lead that async that said uh I do want to ask Shaw one sentence what is the one thing this month this quarter ethereum can help you concretely say it and we'll do it I really think that a plug-in registry for us and then just more of generally an agent registry would be like the really compelling thing to start doing um otherwise you know it it would make uh managing our ecosystem a lot easier um and yeah so I think that that' be the first thing I think putting that on contracts awesome well I think that the beauty of ethereum and while we're here is because we're here for each other so hopefully you come back and you stay and let's make something beautiful happen great thank you so much everyone and thank you kic for letting us go over time for 30 minutes of course uh this was awesome that's why we uh were happy to continue and thank you so much for uh sharing all of your thoughts and thanks Tina for moderating w [Music]