Agent-Native Coordination & DAOs

So, hi everyone. Thanks a lot for being here today. So, my name is Gwen, so I work at ConsenSys on the DevOps side. So today we will speak about the different topics, so I think it's really like AI, education, governance, DAO, DeFi, so like a widespread topic today I think we can start with a random intro. So, I'm actually Nidia.

I'm still being in front of a team member, so I'm also kind of Gabrielle from the LiveOS AI agent framework. Hi everyone, I'm Dylan. At our common, we're a community platform, so we chat with team members, so we have a lot of common topics. I'm Colton, and I work with Idena DAO, and we're using Idena DAO in various roles. Right now we're doing publicants, so we're concerned with visually public grant to the community.

I also work at our action team, which will probably go into that as well. Thanks a lot. So, I think we can start by speaking a little bit about AI agent. I think it's a topic we heard a lot about, I think over the last few years, we spoke a lot about Idena DAO, and there was all this hype around it, and a lot of tokens and everything around it. And I think now it's a little bit more like a stable growth, right, which is happening in the ecosystem.

And we speak a lot about it, but do you see now some concrete use cases of AI agents? Can we really use that in production? Yeah, so, we talk about them in this paper, we talk about DeFi, and there's a lot of issues that we have to talk about. But on the governance side, specifically, I know with DNS, one of the things that we discussed early on is this idea of how do you automate these proposals. So, there's a different phase in each DAO team, but different phases of proposals, and you're doing those from forums, off-chain, or on-chain.

And a lot of that, we kind of know how that's going to work. There's different timelines for each stage of proposals, there's kind of what you predict that they have to go through, and a lot of that could be automated. But right now, we're pretty reliant on usually a few people in the DAO, moving those on-chain, there's certain permissions that we have set, and I think that's probably a use case in the governance phase we can work on, like how do we get this actually in the DAO, this autonomous part of the DAO, actually working more smoothly, and agents, I think, would be a way to get in there. We haven't gotten there yet, we really haven't found a solution, that's what I was talking about, this for three or four years, and I think that we just aren't sure we trust it yet. So, there's that kind of mistake, and then we still have to rely on individuals to do the code, and then if that goes haywire, it's obviously going to backfire.

So, we're still relying on individuals. So, you're saying an agent or something can review, summarize, for example, a governance proposal? I think it was one of the first use cases that people were seeing, which is that sometimes you should check, like, yeah, I don't know if you have a governance proposal, like any DeFi protocol, or like the normies, it can be very, very hard to understand, right? And sometimes it's probably like, because we see a lot of banks and institutions coming on-chain, right? Seems like, for them, sometimes there's not much difference between like Bitcoin, Ethereum, AVAX, and all of the tokens, and the same multi-platform, right?

They don't really see the difference. So, like, do you see like, hey, yeah, like a thing to summarizing those proposals for them, or? I mean, short answer, yes, but if you're not going to do the checks, you can do whatever tool you prefer to summarize some of these governance discussions, I think you're going to waste a lot of time. I would agree there. Number one, just kind of, definitely, we can operate on that DeFi core, you see that DeFi core here in 3.

2 of our hearts is operating as a DeFi principle for so many different projects. I think the one thing, just like, trend-wise, like the two biggest cases of like, advancing something for like, on-chain, very equivalent to like, an NPRM, and obviously like, there's so much infrastructure on that, CRTD, that like, people can start to trust that, and have automated reviewers. So, it's starting to mature, right? The kind of simulations and stuff, kind of going on-chain, but like, definitely push the button, it's a big, big trust thing. And then, I think on the second side of like, just being able to, everything still rests in single-player mode, for kind of like, these agents.

So, yeah, we can all use chat, or it's kind of a deep research thread, but building and, you know, whether it's like, the Notion connection, the Discord connection, the Telegram connection, some of them just don't have like, really great MCP support, or whatever, to like, get that data ingested, but once you have all those types built, I feel like the kind of like, shared context with the summarization is really, is going to happen. And, for what it's worth, there's, just a shout-out to my company, there's General Intelligence Company of New York, not a plug at that, but they just released an agent called GoFounder, and it's like, Notion, Linear, whatever else, and those types kind of starting to be built on something, so I think it would be really cool. Yeah, um, on the, like, governance side of it, we're definitely already seeing a number of new things that we've been doing. It's actually interesting to see how these, uh, those things are kind of growing, uh, in my mind, and, like, we all want to be thinking about how we can make things work out, but it'll be really interesting to see, over a period of time, on top of, like, from whoever it is, like, how is it actually working, like, is it, you know, positions will have to be a little bit more organized, or they're not a bunch of other new things. Education is something that I'm really excited about, because education isn't just, like, hey, we're trying to tutor somebody into learning something, there might be other components, like, just reminders, and if we're thinking about, I'm going to go into a very niche space here, so just bear with me, everyone.

Um, imagine a high school student, like, you know, whether it's your cousin, or, like, your friend, or whoever it might be, um, for them, sometimes it's reminders of, like, hey, you need to make sure you're doing it timely. Like, if there was some basic automation, it doesn't even have to be an agent right now, but it's just the basic automation of those reminders that, in a traditional context, like, that one would have to do that, and if we think about scaling something like that, and enabling people to then, everybody would have that level of support for it. Um, so then, when we're thinking about, like, more specific support, like math support, or something like that, where I'm trying to get tutoring around calculus, or whatever it might be, um, not really, but we're just going to go back and talk about this, Wolfram, and then David, of course, that would be really interesting, because that's, like, it's not live yet, or public yet, but we are partially on hand. Like, we have integration with Wolfram, so LSOS and, uh, Wolfram integration, that can enable an AI agent to do so much more when it comes to tutoring, because the responses that's going to provide are going to be vastly different than an agent responding on mobile back and forth. Now, I think there is a lot on education.

I think it's reached a completely new education system. So, I'm a master's student in Paris at an engineering school, so computer science class, right? Um, and, like, last year was the first generation of the full, like, child GPT student, meaning they had, like, child GPT from, like, a bachelor one to master two, and you can see the level, so it's not the same, like, it's very, very different the way, like, people's minds are shaped. Um, there was this interesting study from MIT, I think, published, like, three months ago, which is a very, very small sample, right? Uh, but that's shown that, like, I think the amount of activity when you use child GPT is something like 80% down.

Um, so it's, I think there is a lot, there is some trials in the U.S., like, high school kind of thing to, like, change completely the way we educate children, because you see now you have all those, like, AI tasks for child, with, like, you have your 10-year-old, which is, like, taking a picture of the math problem and asking the AI to do it for you. So, yeah, I think, you see, like, we get the risk of AI, right? So this is probably one of the risks, is, like, if you don't build the proper structure and the proper, like, critical thinking before learning, like, how to use AI, you probably won't go too far.

So I think, like, there is a lot of demand now in senior devs to fix the code spaghetti of junior devs, right? And I think this is a trend we will see, like, growing in the next few years. The developer generation in three years, right? So, like, now we have, like, one year, which is, like, a long term. And so in three years, we will see probably a lot of demand for senior devs.

So if you're a senior devs, yeah, you probably make bank in three years. But, yeah, like, what are the risks? Like, because you also have, like, prompt injection, which is a risk. Like, if you think in the context, even of, like, governance, if you have a prompt injection in your governance, it can be a very huge problem, right? I mean, if you, like, yeah, what's the main risk you see, like, in using AI in, like, a day-to-day use, either for governance or for something else?

Like, even in DeFi. I think in DeFi for now, I would not be confident. I have some friends that, like, yeah, you should try to use an agent and put your money there. But, like, come on. There's no way to do this, right?

Yeah, DeFi is a bit more serious. In governance, when you have a time loss and things that are in place, you can always move out of that game card. But it doesn't actually worry me as much. I would say, yeah, that's really the answer to that. As people use DeFi, we're doing extremely, extremely fast on a block-by-block basis.

And we're talking about, you know, people's finances and really the confidence in the entire network. Yeah. I mean, you know, the target as a model, you can just repeat what I think about, and then maybe we can circle back on the promise of direction with that, is, like, if I were to trust the agent with my money, you know, I'd think about it as, like, less secure than a hot log. So, like, whatever or not I get, they're comfortable with the same. And I do think there's some interesting products that can come with that.

But I feel like, you know, the code, actually, governance being a risk factor, and also city trust there, but on the engineering side, just like the supply chain facts, and we were talking about this outside, like, a coding agent being able to, or just, like, leveraging an upstream package and maybe not doing HPT or something like that, that's another thing that's scary. I think the one thing, I feel like the research from a graphic around, like, universal jailbreaks actually feels useful in this case, where they just refuse a bunch of different prompts, which, you know, maybe you don't like as a user, but for these things that you can take charge of and you can create another thing, but definitely, yeah, that's worried. I don't want to talk about it as, like, a security risk factor. Like, it's more of a risk factor. I think each of us, like, some of us, I think, are probably on the spectrum, and for folks that will say, oh, it's already available, I think they're responding, yeah, they only have another end of the spectrum than some of us here.

So, there's that component, and then there's also an element that's, like, like, I'm thinking tactically, if I, you know, if I was having an agent manage my portfolio, for example, I'd probably, like, say, oh, yes, from my, you know, other things, because then that also allows me to have a tolerance, you know, for not being able to say, okay, you know, I'm okay for this diversified into going up to now the next versus, you know, I'm probably going to crash and burn, you know, also being a non-zero possibility to whatever else that is, and that's kind of how I think about it. Yeah, I think that's true, like, the risk tolerance part is playing a huge role there, yeah. There was also, like, a question, well, still about the agent, I think, like, it's, do you think you can treat efficiency, right, especially in terms of, like, governance, like, sometimes we see governance proposals, like, taking, like, one week, two weeks, one month, three months, being discussed on forums, and then, like, being applied and being pushed by the government. So, we saw recently, like, the example of Skrull, right, which was, like, a decentralized governance model to a centralized governance model that was actually made before, and this carries some risk for some actors, because I think I saw something about, like, an HCI tweet about, like, oh, I may be shut down if you do this, probably. So, yeah, like, do you think AI can increase efficiency in the decentralized governance model?

Yeah, I would say yes, but, you know, it still comes down to, like, what we call, like, organizational design. I think, like, that bubble reaction is, like, well, we all know, just interacting with, like, ChassisVP every single day, it can output instant amounts of X. So, just, like, more is kind of going back and forth on the government side, I think, probably would slow things down, certainly, and they need to be advanced, which is, I don't know, a workflow agent that can help with bringing the thing on-chain, whether it's drafting the contract. But I would say the thing that it does need to be complemented, and the prediction markets are, you know, hot right now, so to speak, but just coming back to the idea of the target, so that the agents themselves, if they're participating in governance, I think it was maybe an idea whose time has started to come. Just because, you know, the top-down voting process can take so long, it can drag out those blame wars, and then, you know, and then we have to bring the thing on-chain, but leveraging the markets, I think it's something that could actually bring the efficiency that we, maybe, have to promise to put down.

So, yeah, that's what I think. Yeah, and now, I mean, maybe she wants to add something, but now, maybe, yeah. Yeah, I think the fact that, like, when you look at, like, thousands outvoting, like, most of those, they don't even happen. Like, they don't even happen. They're usually, like, a few terms or a few things, but if we need to inject something before that, and say, like, hey, like, this is interesting, like, how do we, how do we do that?

And I think we're just basically, like, following this procedure, like, we have to have it, and then, oh, what if there was some other process, like, you have to think about it as well? So, like, for me, like, if you're talking about, like, how do we make agents, like, make companies do quicker and faster, is injecting before the people, and then just have people as safeguards and help them out. Yeah, because, I mean, I see, like, on the, even on the linear governance, right, so we teach you yesterday, like, we have a consortium that we see in place, right, and, like, I don't, I don't see my company, like, especially, like, we have some people in our consortium, like, Sharpie, which is, like, a very traditional company, right, this and that, like, so I don't see them, like, saying, oh, yeah, like, let's just, like, give everything, like, give everything to the AI, and, like, let it run, right? So, so, yeah, I completely agree on this thing, like, probably, like, having, like, yeah, like, as a first feature, right, for most of these cases, but I think a lot of people still, like, fancy having, like, everything automated with AI, and, like, we can, we can, like, I mean, I think even for Forder, like, this company, I also saw this company, like, it, this idea of, like, there was also, like, Manus, which was another kind of, like, Forder, who, like, can book your trick, and I, I tried it, right, because I need an executive assistant, and everyone in this industry could not make it work, right, because the thing would, like, print you, like, 20 pages of paper, and, like, saying, okay, so you have, like, 25 different slides that you can, like, get something, but, yeah, I, I think at some point we, we have some, like, more efficiency in this, and then it would be good, like, for the question, yeah, like, so where do we have, like, the one person, one billion company, right? Yeah, I, I really like this question.

If, we were talking about the top five, and so maybe this is kind of a, a finger in the wind type of question, but, I mean, it's, for me, it's just, like, all the pieces are, are really starting to get there, and if I think about, like, I really would, as an example, is, is a 10 person, 10 billion, or, you know, house, whatever, whatever the market cap is, and so if you think about just, like, the roles that are involved there, it could be founder, co-founder, engineer, marketing, go-to-market, it feels like all these workflows, whether it's, uh, Preclin, Agents, Codex, whoever, on the other side, that end in end workflows, that, I don't know, at least on, on my feed, like, I seem to see folks really love, maybe that's just marketing, um, these folks maybe working on me, to the product stuff, coming through linear, coming through all the channels on, on, uh, support tickets, it feels like individually they're there, um, I think, I'll stop there, because I know that other folks have other great ideas, but if I, but, if, let's say, next year for a, a token that reaches 1 million, maybe, I think, I think we could have it, um, yeah. I mean, anything's possible. I'll leave it up to you, anything's possible. Um, I, I think the key piece that I'm curious to see is, how much can you relate this effort to AI, and how much of that will allow us to reach out to that person's 1 million? Because I think those are, like, that's their virtual eye, kind of thing, right?

But, anything is possible, based on what we've seen so far. Yeah. Yeah, I think it's really, uh, I mean, we can also ask an AI, right, to, to answer this, because technically, like, fundamentally, an AI is just, like, a probability model, right? So, we can try to, to get some, some answer on this one. Um, but yeah, like, I think, uh, we, we probably see, like, a lot of values bring, like, I think, uh, I've reached, like, 70 billion TVM recently, so we have, like, very huge institutions coming there.

Um, so, but, but I don't see, like, because when we were discussing, like, AI and all those topics, like, one or two years ago, there was, like, a lot of hype, and we were just, like, okay, this would be the next thing for blockchain, and I just think, still, we keep struggling, like, um, to have, like, a proper, like, immediate, uh, use cases and growth, which is tied to this, right? So, all companies use, are using AI now. I think, um, 214, for example, and everything, like, uh. Okay. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.

Thank you. Thank you. Thank you. Thank you. Thank you.