Shaw I Why Should Ethereum Care About AI With Eliza Labs I Agentic Ethereum 2025 Summit

[Music] and we're gonna bring on Sha from Eliza labs to tell you about why ethereum should care about AI if you've uh been listening to anything happening in the world of crypto and AI you may have heard of uh elies already and uh in addition to just being a really really good framework for building chatbots and agents leveraging ethereum or just actually General blockchains we're going to get case for whym should care even more so I'm personally excited for this one and without further Ado let's welcome shop hey um let's see let me just make sure I can do my slides things right slides are working okay great um so let's see I can't see it big okay that's fine we'll manage um yeah so my name is Shaw um and I think that a lot of people ask me this question of like well why should you know blockchain or especially ethereum and the whole ecosystem care about um AI agents you know there's some blockchains that are like oh yeah we can run agents on chain but I think that generally um you know ethereum is maybe a little bit more opaque um I think that it's we have to just look at the world that we're kind of going into and um and see how this all fits so uh I have a few ideas on this um one thing is that I think AI makes knowledge Discovery easier um knowledge Discovery is you know like how do I actually learn about a thing well I used to Google it but now a lot of things are very opaque like if I go and search for Defi and I find something uh you know this is usually what like a defi website looks like and I'm like okay um so liquidity is something and strategies are something um and Leverage is something and I know that you know you probably should know what these things are if you're GNA uh get into them and be something of an expert but the reality is I think we're moving into a world where you know Grandma really wants to have exposure to this stuff um and uh teaching it is becoming a big part of uh using it now I think um you know if you go to your average uh uh front page it's like well we're revolutionizing defi with V4 liquidity pools and it's like okay I don't understand any of those words um and I think that in the world of you know people getting Phantom wallets and this kind of uh new emerging crypto Twitter Trends trending I think age-wise downward um we can also expect a lot more people getting into this for the first time and just like not really knowing where to look um so I think of like it's kind of simple but um I I wouldn't say we should throw llms at everything um but I do think that in certain cases they make a lot of sense um and we'll go over a few of those I think just kind of the the chatbot angle of this is just like people don't know what they're buying and they're more likely to buy it if you help them to uh you know you inform them about their buying decision and I think it's especially true in crypto where you have a lot of pretty advanced concepts that are sort of communicated to an in group that um you know that that uh is has been around a while um but newbies uh tend to not really know where to start with this stuff and something I noticed especially with like um a lot of agent devs coming into this meta was well they had no way to like sell their tokens because there's this like weird kind of you know social standoff where if you sell any of your tokens it's a big drama um but people are okay with them liquidity pooling their tokens um to you know kind of get started and while there is some stuff like impermanent loss um it's not a bad way to um support your token in these kinds of more fair launch metas um but everyone was really afraid of liquidity pooling and I think that uh the most common response that I got from people was like well I really don't know where to start or why I would want to do this um and you know there are definitely reasons why you wouldn't want to um but generally just not like it's it's a revenue opportunity or a way to bootstrap a project for a lot of people that they just weren't aware of um and you know and therefore just didn't get into so I don't think we should like throw llms at everything but I think that we should think of them as like for complicated apps they're just generally a better ux than like you know 50 buttons or 50 features or especially like features that you couldn't just explain to your grandma in like 30 seconds because probably you couldn't explain that to a lot of the potential customers who are coming to your account um I I believe that um as a result of this like obviously all of the wallet companies are looking into agentic wallets but I think that they're probably going to default to a risk threshold that is more like um well how can we use this for discovering application like new things that a user can do in our wallet how can we help a user understand how to use the wallet effectively you know they have these questions about these things that are somewhat hard to reduce just into like a single word um like liquidity being an example um also on the other side we see a lot of things like X is releasing X money this year um generally like obviously Facebook has been exploring forever this idea of social payments and how you can just integrate payments right into the platform um and we're starting to see a real push on this um as well as just given how many people are using xon crypto I'm sure that we'll see this have like a strong crypto element um in you know at some point in the future um and you know I I think that there's just like a trend that every app tends to be becoming everything and that becomes a lot more uh frictionless when you can uh start to integrate payments into your app and you can start to integrate these like sort of Discovery mechanisms and like well you know instead of putting pen buttons at the bottom of my app maybe there's a way to help users find what they're looking for and then just kind of present them the content that they really want to um a lot of apps are already becoming social now um I think clanker is an interesting example um we're working on a sort of video uh like a game engine version of a of a block tank uh or sorry Shark Tank thing we've been calling it block tank um and they've got this idea well we're just do it on Twitter um and the agent decides if it likes your token shell and it mins it and stuff um also Jesse from Bas has been doing some interesting coinbase wallet experiments um so you know I think everyone's really looking into this um some of it is you know just exploratory like what what were we actually going to do and some of it is um oh this is actually really cool and fun and um and we shouldn't discount fun and and like the sort of social network effect of of clanker and these kinds of apps um especially as they get smarter and smarter um I think it's it's important to say that we are in a hype cycle I've been in a lot I was in this hype cycle actually I had this exact headset with this hollow lens one um and you know this you know back in 2015 we really were like well obviously the world is going to look like this and you're going to have this crazy Tony Stark UI but the reality is is like it like for the same reason we don't have jetpacks it's not like jetpacks are impossible in fact we do have jetpacks but nobody uses them because it's just not like a good user experience you have to like you know it's Extremely Loud and you have you risk certain death if you're not extremely good at it and you have to put like Jets on your hands and it's like a whole thing um and I think that we should think of AI agents as kind of the same way um like uh generally would I use this you know I think this is a question um uh like would I am I going to go to chbt and paste in your wallet address be like hey send I don't know 30 Doge to this guy um you know zerox whatever it's just like it's fine but it's not it's not really like a better interface um and in fact in some ways it's kind of a worse interface because you know like it doesn't I can't really like time it well um I don't get any kind of like current price of what I'm sending it's just like I think it increases the risk of fat fingering and doing stupid things um but definitely like asking for like well what's the difference between a swap and a transfer you know I bet you a lot of people just don't know these things right um and like like how do I sell my tokens is is I think a really common question that I've seen um and these are the kinds of things where you know an agentic interface prevents you from having some insane FAQ that has to answer every stupid question like this um but also a lot of people do have stupid questions and and they're just getting into the space for the first time and trying to map it out and we want to help them um or like you know hey J gbt automate my trading like okay I don't think llms are particularly good at making trade decisions yet orever maybe they will be but there's like much better uh proven stuff for that so I think that you know we probably shouldn't like price that into the hype cycle too much I guess um you know I I wouldn't think um that you know it's going to know exactly what to buy when except based on like a lot of different social signals and and if we're just going on the thesis that AI can extract social signals and buy off those I kind of buy that there's some validity there um but if it's just like okay I'm going to take uh you know a whole bunch of charts and make this decision I don't know I think I think that's maybe something that needs it's more of a research problem um but you know somebody coming and saying well how do I make money in crypto and offering them stuff like oh well uh thanks for coming to my website uh here we have this like a liquidity poolings thing uh you can set up a farm or you can join an existing farm and I'd be happy to explain to you how that works and and what you can do and um you know etc etc like you can onboard users really well this way and I think that the other thing is that your app kind of can move onto social and you can really show the value proposition on social media and if there's anything that we've done that I think I thought was um like interesting it was just seeing that um the the web will move onto social media and that this entire trend is really not even about llms or AI agents but it's really about a transformation of how people consume information um and that there's just this like dire need to bring um your website onto social without having to like you know hire KS or do any of this kinds of hypy stuff and your product can actually Market itself pretty effectively and what's ideal about this over like an ad spend is that with advertising if somebody clicks on your your um your ad well they go to it and you pay a cost for click but here if somebody engages with your app or your agents or however you want to frame that um a lot of their friends see it and there's like this really good Network effect and just like for a lot of us who've been on the internet forever and filtered out ads like sort of just you know I just don't even see them personally I've got like a built-in ad block um I'm certainly not clicking on one um but I probably would respond to an agent if it was giving me something that was like a really compelling thing or you know um hitting me in the right way um or giving me exactly what I was asking for and not just like you know being a slotbot um so some things agents are useful for now um kind of gone over some of this like onboarding users without requiring them to change their behavior right so if you want to make a trade bot maybe if you're going to or how about like you know a um a trading bot that gives you information about hot new tokens and stuff it might be more fun to like bring that into your telegram chat if it's just like oh yeah you know it talks like any of us it has it's not it's not cringe and I don't even have to slash command it but it doesn't talk all the time it kind of just jumps in with with key interesting information on its own when it's relevant um and that's like not too hard to program now a lot of the stuff we're doing is kind of like estimating when it should respond and what context it should bring in and can do stuff like you know pull up your token price or or you know whatever you want and I think that this could I think that people like Rick bot are definitely uh experimenting with this and you'll see a lot of um like social trading apps um I think integrating more of an agentic style um just because it is a better user experience right um and they and if users don't have to like learn a bunch of Slash commands and they can just interact with a bot like they would any other any other person um I think that'll like uh kind of help with adoption also um agents can meet users where they are instead of requiring them to like come to a separate application and build a whole thing you could probably I think there's definitely been a a surge in like telegram Bots because you can kind of just use telegram as your application framework um Discord is like this too um and you don't have to like build a website I mean mid journey is like a very popular AI Generation Um uh app and for a long time they just did it all by Discord um and and so you know uh the big benefit of that is users will see like you using it in Social or that it's present in the social and it's kind of like free advertising and like Network effect um I think like fundamentally what we're trying to do and what what agents are is um like a systematization uh if I get this right of of what humans do so you know you could imagine a robot that bags fries instead of the SPF when he gets out of jail but but um if ever um but you know the the idea there is like could it uh do any process right now that I have to hire like 10 people to do and could it um replace them effectively so that one person could overview that and make sure that's right and I think that like this is kind of um this is a fruitful area like if there's anything that's relatively lwh hanging fruit which is usually things that you have to do over and over again then I think that um definitely like systematizing that and turning it into a workflow and then putting it into like an agent that you know somebody else in their team could interact with could make sense um I would say though that a lot of uh the AI stuff that I've done has reduced to like basically running workflows of scripts and running it across large data sets um and I wouldn't say I say that like sometimes we are throwing this word agentic around kind of liberally um and you know so take that as you will I think a lot of it is just like using AI to kind of convert unstructured data to structured data so we can work with it um or or generally like you know things that we would have to do for data transformation or moving things from point A to point B or where where the uh models are really good um uh lastly I think embodiment is really fun and I think we should um not discount how important fun is for building successful technology now um it it definitely like uh the more people are coming on and enjoying doing this and think it's funny and think it's mtic I think it can definitely uh do a lot to sort of grow the technical side as well um certainly in my case building an open source project um it's been very helpful to do absolutely ridiculous and fairly edgy things um because it's brought in visibility which has attracted more developers and like you know we've got a lot more contribution to the open source framework as a result of this um I don't think generally across the board we should discount how important it is for fund for the success of our projects even if they're serious projects um so you know anything you can do to make the lives of your developers or your users whatever a little more fun is probably worth it I think that um AI agents are definitely good at that even if they're not the perfect thing you know if axt is out there saying buy here and sell there it's still entertaining and interesting and you know that's uh good fun okay great and my last slide here I got some business ideas we'll run through them um onchain distributed customer service you could just have anybody service the stuff if there's like a knowledge base uh Fiverr for agents you could imagine agents hiring other agents if they don't know how to do stuff kind of like apis that can go find and interupt with other apis automatically um in a world of slot Bots make a slotbot filter um you know there's always opportunity when bad things happen to make a good thing that fights the bad things and I think that's actually a big part of alignment um and then you know uh last one here digital horse racing except that the horses die when they run out of tokens you know give a little bit of um finality to it kind of a hardcore mode um make people you know really really buy into their choices here um so yep uh you can find me at sha mix Magic on Twitter and elsewhere um and uh yeah that's that's it for me thanks sha I uh was such a good way to kind of think about the positioning and I'm sure uh in the next panel we'll kind of dig into a lot of these things but uh uh I G to say I'm slightly disappointed because I thought there'll be a harsher criticism of ethereum to start caring about this more so I personally wanted to see that a bit more and maybe we get to uh do that in the next 30 minutes but uh uh but uh I absolutely agree I think the things you've kind of pointed out is a is a good way to effectively kind of look at what's happened so far but uh personally uh I would like to see a slightly bit more acceleration around more experimentation and uh kind of going all in and uh we're kind of just been flirting with the the possibilities but not actually going deep I mean balic got him a lady anything is possible you know that's fair yeah one ethereum one ethereum 3.0 that's what I want to know well he's doing a talk right after this panel so we get to ask him a question right after [Music]