[Music] We are ready for our next speaker and for this one like to bring on Mike to tell you all about the future of AI having crypto wallets. Please give him a big round of applause.
Thanks Kik. Hi everyone. Uh my name is Mike. um co-founder at superfluid protocol is the streaming asset protocol where you earn every second. Today I'm here to talk about um why a wallet is a gamecher for AI and I promise you I'm not going to be very high level.
It's going to be a very down to the bottom technical pragmatic talk. So if you are a builder or if you are working in a blockchain web3 company, you'll probably take in some uh insights from from the call. Now, uh, just a show of hands before I start, can you raise your hand if you've been using AI to search the web this week or the last week? Awesome. Nice.
Almost everybody. And how about learning about a new project in crypto, right? And what about building an application? Okay, a lot of developers in here. Or maybe not.
We don't need them anymore. So, not joking. Come on. Um, so the reason I'm on stage today is because we are entering an era where the dominant users of software are not humans anymore. They are agents and they will be driven by our intents, driven by our management, but we will probably not interact with software as we do today.
So today agents are already pretty smart. They can do a lot of things. They can reason, plan, they can take action uh based on your, you know, desires. They can execute fairly complex workflows and they can make a lot of autonomous decisions, right? Um but it's not enough, right?
So what's missing for true agency? Like what is giving an agent that full ability to take initiative by itself? Well, definitely ownership. So an agent is not able to own itself or to own a lot of assets. There is a lot of issue with access to capital.
So an agent cannot raise funds easily today. It cannot have a lot of financial services available. You know the banks are not very happy to just you know provide a lot of capital to software today and at the same time they just miss a coordination layer where they can talk to one another and then you know take action in a swarm fashion. uh the swarms we have today are fairly rudimental and that's where crypto comes in. So if you give an agent a wallet, it can freely transact on chain.
It can also own assets. Think about NFTTS, think about credentials to use software etc. It can also hire hire people, hire other agents, hire service providers and it can make decisions on governance of protocols like for example governing a DAO via voting on chain. It can coordinate at scale with humans and other agents including with institutions because at the end of the day every company is just a group of humans working together. So with blockchain AIs or AI agents are an economic actor.
We are entering a world where agents are going to take decisions and shape our economy together with us. The good news is that we get to manage them. Let's see for how long. But for the time being, we can still coordinate their action ourself. We can monitor them.
We can instruct them. We can understand their intents. So a wallet truly is the difference between an assistant agent and entrepreneur agent. So imagine you know in in a while having companies run by AI agents or having conferences designed run and optimized by AI agents. Right?
We already seen a lot of DeFi plays where a lot of venture capital and investors have been using agents to optimize DeFi strategies. For example, they attribute them some capital. They let them try to increase the capital and they let them try to experiment, right? Try to optimize the return on investment. So, I want to make it very tangible for you.
Um, can you raise your hand if you ever voted in a DAO or if you own any tokens that would allow you to vote? Okay, so here are two delegates for you to choose. You can either choose Mark or Laura. Which one would you go for? Hands up.
Who's going to go with Mark? Who's going to go with Laura? Pretty obvious choice, right? Yes, because Laura is on the spot. She's always working 247, representing your interests, is informed on everything that relates to the DAO, is up to speed with the products itself, is up to speed with competitors, with the narratives on Twitter, right?
She knows everything basically and she's tailored to represent you uniquely with your own set of values and your interests. So, this is already emerging. It is not in the future anymore. It's here today. Right?
People are using AI and especially agents to summarize Dow proposals and then they are using them to understand how they can swing the vote in their interest. Right? You can imagine a lot of the VC investors in the space they have quite a lot of vested interest in the Dows taking specific decisions that benefit their portfolio. Right? So this is already in place today.
Now we accidentally built the perfect operating system for autonomous agents. It almost feels like AI came from the future to give us Bitcoin so that we could build an infrastructure for the agents themselves to interact and transact with one another. So if you look at you know for example Ethereum let's say it's clearly machine readable is software is open source is transparent it enforces the rules. So you know is you cannot game a smart contract once it's deployed on chain h is deterministic is permissionless and it it has basically scriptable interactions where you know if a then b right um there is a lot of open interfaces and APIs that are available so that the rules are composable I can say if I a then b and then if c then d and then if b and d happen then do x right so we have full composibility and now we can build an economy with that. Um the other thing missing is reputation.
If you have agents operating in autonomy then how can you measure their quality right and because well all transactions on blockchain technologies are auditable and transparent then you have a built-in reputation system I can tell if an agent has been operating in the way I wanted to or or not. So I can actually make decisions with that. So um I asked GPT what's the volume we're doing today in the traditional markets and if you the audio is not going to play but basically ch has said to me and I verified this 60 to 80% of the volume on traditional financial market is algo trading. Humans are in for the 20 to 40% today in traditional financial markets. So it's it's already happening in traditional finance.
It's obviously going to happen in crypto and it's going to happen real fast. In fact, there's a lot of algo trading which is not AIdriven today in the crypto market. So I'm here to tell you that the next wave of protocol users are AI agents and not people. This means we have been optimizing for the wrong users because we have been first optimizing for developers and then we've been optimizing for human users and crypto if if you know if you know solidity is not really intended for humans is more intended for machines to understand what's going on and it really feels tailored to uh some software users. So let's look at how protocols are built today.
Um, I built Superfluid together with my co-founders and I can tell you that we've been tailoring it to human needs at the moment. So, we thought about what do people need and also we needed a UI that you guys can use to stream tokens on chain. So, we built our own UI and then we have made some incentives that are appealing to you to go and use this protocol to move funds so that you can earn every second. We also created an experience around the fact you guys asleep at night and you only have so much energy and you know it takes a while to click around the UI and we had to cover for a bunch of errors that we do you know fat finger typing all of that we don't want you to send 10,000 times what you intend to send so there's a lot of experience around human error as well and at the same time we have a lot of uh emotions at play So you know pumpf fun on Salana is exploiting these emotions on the rush the FOMO and so on. So we have a lot of emotions baked into protocols today.
However agents are quite different. They are always on 24/7. They can scale instantly. Why? Because they can hire resources.
DPN offers them as much GPU, CPU or storage they can afford to buy or rent. They are very rational differently from us, right? They're they're code. They tend to be very very logical. Also, they respond to incentives very differently than we do.
And we we have seen that they seek predictability, reliability and low latency. So the question we have now if we really want to succeed in the web 3 of tomorrow is how does an agent choose your protocol or my protocol is a very different choice than a human. So welcome to agent experience is a concept that emerged a few times but nobody really took the time to define it clearly as a as an operator in the crypto industry. So I went on and tried to dissect it. Now how do we define agent experience is basically all of the interactions, perceptions and outcomes that an autonomous software agent has while discovering, evaluating and executing actions on a digital system or protocol, including the ability to parse reason and optimize for its own goals.
This means that when you design with agent experience in mind, you optimize for different things than humans. First of all, you have legibility. I, as an agent, will need to understand your code. I want to break it down into small problems because I'm not AGI yet. So, I want a modular approach.
I want your docs to be clear, structured, and easy to parse in in a machine readable format. I want interfaces to stay the same or I want to be told if they change so I don't have surprises, right? I I don't want to spend my computing resources to, you know, just try and fail many times using your protocol. So, I want a low latency. I don't want to wait for ages to get a response on my query and so speed is is very important.
Um I also want to be able to understand intents of other agents or or UN users. Um I definitely care about composibility. Can I take the output of a protocol or or another agent and feed it as input to what I want to do and vice versa? I want it to be predictable and also economically interesting for me. So I want deterministic rewards that are not going to change and I can predict for and plan for in a financial plan that I have for my own and I don't want some sort of opinionated interpretation because I can't predict for that.
So it's really changing the way that we design systems and protocols. And also I want some sort of deep auditability because the chain allows me to tell how things went and what the outcome was. So I basically want to make sure that I use that for logs of events and also for building my own and evaluating others reputation. So it's a very very different game that we're used to and it's really going to change the way we shape also our companies the entire funnel from user acquisition all the way down to retention of the users we're going to have to think about agents. Now to give you a very tangible example at superfluid we manage real-time value transfer on chain and when a user thinks of a one-off transaction or a one-off value transfer normally is is like I paid you or not simple in the case of agents is going to be very simp similar so it's an instant transaction either I did it or I didn't one or zero but when it comes to recurring value transfer or repeated value then we like to think in months subscription on Netflix once a month right your salary once a month sometimes every two weeks agents they don't have the concept of time what is the right subscription period for an agent they love to think in terms of ones and zeros so from what is a repeated value transfer we have now a ongoing stream I am paying you, I have access, I am not paying you, I don't have access is really a real-time relationship.
So in this case, imagine like a live feed. Do I as agent have access to that? Am I subscribed to that? If I am paying I do. I don't want to keep a history of transactions for a arbitrary period of time humans imposed on me the month let's say.
Okay. So agents really think in terms of ones and zeros and that's why repeated value transfers for agents is a stream. Now this is exactly what it says. So you go from have I paid for the current period to am I paying or not. So many many protocols will thrive with agent users by being tailored to machines.
Unfortunately today I cannot go much deeper and unpack this with you in in detail. So, I actually put together an article on Twitter X that you can check out scanning that QR code. And because we don't have time for a Q&A, please drop any comments you have on my talk there and we'll take the conversation online. So, thank you very much for for staying with me along this journey today and reach out to me on the same Twitter link anytime. Thank you so much.
Thanks, Mike. We actually do have a few minutes. So, if anybody has questions, please just raise your hand. We'll kind of do them. If not, I have a couple questions from just the slides.
So, does anybody have any questions about wallets and agents?
I'll I'll get my mic to you so you can say it out loud.
Hi. Um, I have a question about the agent to agent protocols. Um because uh in web two there are already uh not protocols between each other but just connecting. Uh so do you have insights about agent to agent? Yeah.
I mean if you have you heard about things called agent swarm? Yes. Right. So basically sometimes you have a master agent that has access to other agents to do specific tasks. Now I don't remember the name of the of the VC fund but someone has created an agent that then leverages other agents to do actions on DeFi.
And one of the smart things that they did is they created an agent to work with a specific protocol. let's say a for DeFi and that agent has the ability to work on on a he needs to transact incentives tokens to other agents that are using other lending protocol to maximize the value of the ROI. So now they are paying each other with a currency that is just for the two agents to basically determine which one is the most efficient. Now luckily this is kind of closed loop for now. So it's not out there in the economy, but we can really envision a world where, you know, I'm going to literally pay for access to that agent over time.
So luckily, we are the early days and, you know, we have a lot to to build. But yeah.
All right. So my question is, who do you think has come the closest to doing something like this thing right now in in the web 3 world? Are you seeing really cool experiments or are they just kind of prototypes but they're not actually trying to make it a real thing?
I think it's very early days. So what we are seeing is there's a lot of startups right now or or companies that are building the communication layer between agents so that they can talk to each other. They understand the intents and the outcomes. Uh there are other startups that are building u directly useful execution layer agents. So maximize my portfolio value let's say um most of these experiments and thought pieces are coming from the VC space.
So you know if you if you look at some of the leading V web 3 venture capital funds blogs you will find a lot of golden nuggets in there. Um but yeah there's a lot basically the answer is no. Nobody has built the whole stack. Um I have seen a lot this week that is at preed stage.
That means it's there's a lot of opportunity and uh there's a infinite garden of room to uh to grow. So well thank you so much. Give a big round of applause.
Thank you. [Music]
Automatic transcript — names and jargon may be misspelled.