The AI Agent Economy | Jake Brukhman - CoinFund | David Sneider - Lit Protocol |

[Music] all right I love it let's keep the conversations going keep it flowing next up we have a conversation uh about the AI agent economy it's going to be between Jake Brookman the coin fund CEO and founder and David Snyder co-founder of lit protocol please welcome them to the stage so much all right hey everybody hey everybody Hey Jake hey David how are you so good uh well let's I guess let's jump right into it AI agent economy or agenta Commerce I guess maybe we should start with what our AI agents and what do they do yeah um I'll introduce myself a little bit my name is David Snyder I'm the co-founder of a project called lit protocol it's a distributed programmable Key Management Network essentially what ethereum does for state by making it decentralized and programmable lit does for signing and encryption and so this is relevant uh because the general view that we have of Agents is agents are just functionally intelligent Services um like machine learning and artificial intelligence has been a part of the applications we use like behind the Facebook feed for more than 10 years and what we're seeing now with agents is these Services uh which are essentially just applications being able to perform more sophisticated tasks does that map to your definition of Agents I think so I think like when I think of Agents I really want to hone in on the agentic part which means these are really these are not chat Bots that I'm kind of going back and forth with these are um kind of autonomous programs that are acting on my behalf and doing useful things and like saving me time so when I look at maybe like the landscape of Agents the way that they are now I see uh like influencer agents which are kind of cool but they're not very agentic actually like for me they're just sort of sitting there tweeting and then I see um like domain expert agents like if you guys know axt this is an agent that supposed to surface crypto Alpha on uh Twitter orx and that's a little bit more useful um but again it's kind of a chatbot like I can go back and forth with it and ask it questions but to me like the really useful ones are maybe what Giza is doing which is onchain uh agents that help you optimize your yield roll your perss trade stuff um and I think it's going to keep going in that direction like agents will be able to do more and more and more over time yeah that that definitely maps to my mental model uh some people have described it as wave one of agents are agents that deliver information so like chat GPT would be the prime example of a thing that is in Wave 1 you ask at something it delivers information whereas wave two are agents that do things on behalf of the user or a company so like in the B2B World um outside of crypto the main agent use cases are like agents for small businesses that do things like customer support sales Outreach things of that nature in crypto like what Giza is doing with yield optimization or a number of the agents that are running on the lit agent wallet for authentication are doing things like creating tokens um optimizing yield impermanent loss prevention automatic solving across chains and there's definitely like a pretty interesting aspect where um all the technological components are here but the road ahead is around integration which the agents are also helping with with the coding agents and we're probably only like a month or two away where you can functionally start a business with a prompt basically you say to the agent you go mint a token for me set up a newsletter set up a Twitter account make the newsletter only accessible if somebody Stakes a certain amount of my tokens every morning send me a text message that has three clarifying questions about what the news was from the day before and because you're familiar with my worldview because I've been communicating with you for 6 months publish a newsletter from my point of view and now I've just created a media company that has on Trin components that I spend 5 minutes on every day and it's just like a totally new world in terms of Novel economic activity and really just kind of creativity and having a unique view on the world or what's happening as a way to generate value yeah and maybe to jump in um with like a lit protocol Centric view just to talk a little bit about you know what are the infrastructures required for agents I'll say uh I've always thought of it as like verifiability you know if your agent is managing a lot of money for you on your behalf you kind of want to have some pretty strong guarantees about what AI model is running underneath whether it's being run correctly a little bit of privacy uh you know as a someone who runs a company that wants to use AI also very cognizant of like I don't want to up necessarily upload my proprietary data to Sam wman servers and you actually kind of opened my eyes a little bit today that there's another aspect to the infrastructure which is maybe more permissions management so maybe do you want to touch upon that yeah absolutely I think you teed that up really nicely like functionally for an agent that does things on behalf of somebody there's three components there's the llm which is the brain which could be either on open AI or run in a verifiable way whether it's with ZK ml or tees or one of these other privacy enhancing Technologies there's the agent context and memory and so some of the work that like recall has done around encrypting agent memory with lit making sure that the user is not bound to a specific system but can move from model to model to model and then the third component and this is where we're really focused at lit protocol is the tools and so an an agent is just this this vectorized space it's just software how does that llm then reach out to systems in the world from the perspective of an llm doing e-commerce where you buy flowers for your mother on Mother's Day and doing a million doll transaction on bitcoin are functionally the same thing and so the view that we've taken at lip protocol is and and and this you know something that we've been working on for many years is that in all cases in software systems identity and Authority is represented through a secret of course in crypto that's private keys in a lot of web 2 systems that's the API key in e-commerce it's a credit card on social it's a password and so what we built with the lit agent wallet is essentially like ooth but for transactions so you're probably all Familiar of what this looks like for data when you sign in with Google and the app says this app wants access to your context and calendar and then you click approve and that app can then use your data for that case the way the lit agent wallet works is the agent runtime or the app or the service these are all kind of synonyms here says this agent wants to be able to use Unis swap with a daily spending limit of $1,500 or this agent wants to be able to use your Twitter account or your credit card or your binance account or your Charles Schwab account to purchase stocks and so we've really kind of taken this perspective that a that a transaction is a transaction independent of the state machine whether it's decentralized or centralized and we actually think that this is imperative not just for the people in this room at eth Denver but attracting the web 2 agent develop velers and seeing decentralized web Technologies as a core piece of infrastructure to orchestrate transactions where the end result of those transactions may be on blockchain and may not be but the blockchain and decentralized web Technologies as the coordination and permission layer that kind of becomes like the new iPhone right like as an analogy you have the iPhone everybody or or an Android phone and that's the home assistant it has your passwords it has your pass keys and then the service that you use from the iPhone are all of the apps that you install and so the agent world I think the model ends up looking really similar except uh it's not a piece of Hardware um it's this core home assistant that then you can send to other agents and other services out there in the internet what people are largely calling the Swarm these days and yeah we're really excited about that perspective especially in so far as we can bring web 2 developers into the web 3 space without them having to understand the different specifications of the different smart accounts on different blockchains and kind of Ratchet up the abstraction stack a pretty classic move that has happened in crypto for many years and now in the context of Agents okay awesome yeah um let's see I guess what is a what does adoption look like do you think yeah I mean it's still super early I'd definitely be curious around your perspective here but you know we're in web 2 we're seeing these B2B agents that I mentioned earlier there's very cool experiments that agents that trade um there's a lot of like experiments where you say hey chat GPT what are the best tokens to buy I've been really impressed with um what the team at third web has done where they've trained a model over all of the evm trains that model is called nebula like specific speci ific models for blockchains are enablers of more sophisticated type of transactions but you mentioned kind of the the social media agents or what has been characterized as the internet Cinematic Universe I think it's worth kind of making the distinction like at least to my mind the most interesting things that are getting adoption are like the most boring things the things that are just taking a human out of the loop from having to pay attention to something or click a button and just having the llm do that in its place how about how about you what do you think is going to be the first type of breakout agents well if I'm being like super honest I think that um kind of those types of Agents I mentioned before that act on your behalf they're not really that good today yet we're obviously on this Innovation curve and they're getting better but like I tried operator the other day to automate some of my tax reporting tasks like it had to go to a bunch of you know crypto wallet viewers and collect balances and assets into a spreadsheet kind of thing and I found that operator like made a lot of mistakes it started to open other like spreadsheets in my Google Docs and stuff and I had to turn it off but um but there actually are like some very very useful things that these things can do today they can like Source data you know on our research team we we ran an experiment where we um sent openai deep research to kind of look at token allocations for different projects and the data that came back was essentially as good or better than what we had done manually because it went like directly to the source you know to the published allocations that the that these issuers have published now as a company again like we're very cognizant of this proprietary uh data and so like one way that like web 3 agent adoption could even look like is giving you those properties of privacy or maybe even being able to locally run um you know an llm or an agent or domain expert on something that sort of protects you from you know leaking this data yeah super interesting I mean like classically there's this idea that like privacy is not a sellable feature um certainly there are cases where it becomes very important you know open AI comp requirement in certain cases there's a compliance require and then like then the interesting question is do you do compliance up to the state of the world today which is web 2 which is to say okay cool we've signed this Enterprise contract with open AI they've made these guarantees on paper that they're not going to read the data or resell it or you know whatever keeps them sock to compliant or whatever might be or like what is the reason to move to like the next level of ver IFI ability certainly when you're dealing with money that's one of them um but I think like there's an opportunity for us as a space collectively to kind of move out of the the Doom and Gloom pitch which is largely like don't trust web 2 companies because they're scary and bad and I totally feel that and understand that um but the reality is is most people trust most corporations most of the time and there's really an opportunity to talk about what are the higher order benefits to building the machine intelligent Society on decentralized web infrastructure and to me it really just comes down to things like collective intelligence Global cyber security threat detection these things only work in an interoperable world where the benefit of using decentralized web Technologies isn't this idealistic one but is is is more material such that we can have better types of ations and to my mind like using systems like decentralized storage with lit encryption and and other encryption methodologies and tees and FHA and privacy enhancing Technologies is the only way to create intelligent protocols that are functionally owned by no one a similar property that ethereum has but upgradable by everyone awesome all right so in the last two minutes why don't we do bold predictions about the a AI agent decentralized AI space okay you go first oh okay all right well I'm going to make a bold prediction that we will be uh producing AI models in web 3 that will be competitive with you know the open AIS uh and anthropics of the world I have some pretty good confidence on that um and I also will predict that that will result and a lot more llms out there a lot more agents and they will be in this kind of web 3 format nice um mine is more of a Wishful prediction but uh you know the I guess the the prediction and the hope is that um we start to like really rival visa and the web Tu methodologies of of trafi of transactions um and like the flippening potentially is intimate but not the ethereum Bitcoin flippening but um the count of of daily transactions being on crypto rails versus on Fiat rails and agents being a driver for that over the next 24 to 36 months awesome David thank you so much for this conversation super Jake thank you so much cheers [Applause]