So I'm glad that cypherpunks haven't died out. I was one of the first cypherpunks, and I was, you know, a little worried for a moment . Um, she said, “I’m with Dean Trible from Agoric, and we’re building a product called YAX, but I’m actually going to talk about agents and what it takes to do that? Um, in case people weren't paying attention, um, financial activity is shifting dramatically from people involved in all the cycles to agents capable of performing actions on behalf of people. There are serious obstacles to this right now, and a lot of this conversation is about how to address them.
But we know that agents will be placing capital, and not just Giza, it's AI placing its own capital, but, um, Fireblocks, Coinbase, Circle, um, this morning MetaMask is rolling out APIs so that agents can talk to these systems. And it's not systems. And it's not , you know, and it's largely driven by the desire for intelligent automation that can adapt to the complexity of our evolving economy, um, and operate at the speed of this electronic economy. And this is irreversible. I mean, this is just in its early stages.
There are relatively small amounts of tens or perhaps a few hundred million." dedicated to this right now. But this is a market that people see as inevitable. That's why everyone is getting involved because of the benefits, and the financial world is already too complex to manage manually, and it's becoming more and more advanced. So agents, as you know, you and I who are not professionals in direct money management, are forced to keep up with the times and the market is agreeing, I mean they are redeploying agents not only into advisory roles but into executive roles.
This means executive roles. This means that agents, uh, will be transacting directly with money. But there are key obstacles to this. Over the past few months, you may have seen an increase in issues with agents going off the rails. Either it was deleting the production databases of the running services, or, uh, there was an agent that didn't have any infrastructure issues.
It's just that the agent was denied money, and since he had the authority to send his money to arbitrary recipients, the attacker could convince him to send them his money so they could help. And, uh, many of us have seen these attacks on us. Agents are simply the object of this. And so we know, and the Industry knows, and the market knows, that right now you really can't trust agents with your money. That this is the number one obstacle to a world where agents can act on our behalf so that we can cope and thrive in an increasingly wealthy economy.
So what can we do to make agents trustworthy? Because this is the number one problem that needs to be solved in the future to thrive in this economy. So there are a lot of things you can't do. There are many things that do not solve this problem. Right.
In most of the financial industry, you let people perform actions and then audit them after the fact. With the speed of computers, with the speed of light that our economy is moving towards, it's already too late, isn't it? The money has been sent. They are convertible. They were exchanged, and they're going to some country that we didn't want to build missiles with, right?
So monitoring is too late. It detects theft, fraud, and simple mistakes after they are no longer recoverable no longer recoverable . The other thing you see is...Uh, that's what people set up rules for actors, right? They tell the actor, “Okay, you know, you can only put your money in these vaults.
You can only put your money on these chains. You can only send your money back to me." And then, to help, someone convinces him that it's much cheaper, better, faster to return my money by giving it to them first. And then somehow I never see the money. Right?
So we try to get the actor we want to cast and the agent to do the right thing and just say, "Well, here's how we're going to do it." “This doesn’t work. That's what we're trying to control, right? Right? And that's like giving the junior clerk you just hired the right to decide when he's traded enough or when he has enough leverage.
And there are many banks that have failed and who can tell you that this is not necessarily a good idea. And the other thing good idea. And the other thing that you see, and we saw this in the exchange presentation where they discussed how you can access the API to trade on this exchange trade on this exchange , is that they have additional agents. They either have clear guidance for your current agent, which is the second point, or they have additional agents who will review what you are doing and decide if it is approved. And these agents are no more trustworthy than the agent who is making the transaction, right?
This is not the type of deterministic high integrity you want to have to protect access and execution of real funds. And so, you know, essentially, more of the same won't help. Therefore, this requires changes in the structural environment. And you see again this growing awareness of the issue of protective fencing as something that everyone is doing now.” who they are talking to when they talk about agents.
But a lot of them are talking about it, like I said, I have this smarter agent and he's confident that he'll work better. So the answer is not to have more of these. This is a direct fulfillment of the mandate, you know, of this financial industry. This idea of limited powers that this thing can perform no matter what it tries to do, right? It's about what she can do, not what she wants to do.
And that's not the case, and you're not going to achieve that by you're not going to achieve that by giving her the keys and hoping she does the right thing, right? It's like giving the person who buys your groceries for you the authority to sign your bank accounts, right? You wouldn't do that. This doesn't make any sense. You want them to have limited execution, whatever they do, they can only spend a certain amount of money.
And really, you'd like them to be able to spend it only on the products you asked them to buy, would n't you? It's also similar to how Robin Hood released an API for trading, where you put down some money and give her the keys to...it. Well, simply giving him less money with keys is like cutting a bill for your child. This is a great learning experience.
We'll learn a lot from this , but that's not what you're going to do with your retirement account, is it? This is exactly what you want to do even within a company for fund allocation. It's just a place to learn. Very powerful. I'm very glad they're doing this.
This confirms how important this is as an industry movement, but it is not a solution at all. The second way people solve this problem is through storage, right? That is, I put my money in Coinbase, Anchorage, Fireblocks, etc., and they will provide execution. I can create rules in that context and then my agent interacts with their API and it can only do what I configured in the rules and that's really nice and you know and banks are starting to add it, Mercury has added something like that and so on.
This is really useful and valuable. But storage providers are implementing non- are implementing non- custodial solutions custodial solutions because storage cannot be moved. at the speed, uh, at which the economy is developing. It can't move at the speed of computers, it can't move at the speed of light, right? And again, this is a valuable and important niche.
But the future of agent commerce is a rich, dynamic, adaptive e- economy powered by smart powered by smart contracts, cryptocurrency, and even, uh, traditional trading services or merchant services. Storage services. Storage is not a solution. This is not a complete solution. And again, you can't do that with terms of service, right?
You can't do it with, you know, I'm going to sign, and going to sign, and , and, and, and sign the risks and rights, and my agent will follow these rules. Again , this is just a duplication of what we are trying to prevent. So, how do you ensure that the mandate is fulfilled? Good. Well, we have this tool, smart tool, smart contracts, that solves this huge problem called the principal-agent problem.
If you're familiar with this in the economic literature, uh. So what do we want? The user, who is perhaps talking to an artificial intelligence, decides what the mandate is, decides what they want to do. I was talking to some people at Morpho yesterday, and a common question they get is, "I want my money to sit in this, you know, relatively low-risk, you know, 6% vault until some vault goes up, you know, to 8% up, you know, to 8% or more." And regularly in different chains, storage increases up to 15 or 20% because there was a big demand for liquidity, right?
The demand for USDC liquidity, USDT liquidity changes, which chain has the highest demand, as the industry develops. So the mandate I want is to put them in this specific vault with a reasonable fixed interest rate until it goes up to 8%, then move your money there for a few days, then move it back. This is what I want my strategy to look like. And part of the mandate is that here is the set of repositories that you are allowed to participate in, and here is the maximum that you can put into any one repository any one repository . And then the strategy of how to react quickly, and how to withdraw funds quickly, and whether the movement is sufficient and real.
And let me look at the Twitter signals to see if the reason for the sharp rise in interest rates is demand or a liquidity crisis. Um liquidity crisis. Um , and based on all of this, the agent will act. So, I'm taking this strategy that I came up with. I take the policy, sign it, and send it to a contract, yes, a smart contract on the blockchain, which will enforce the boundaries and scope within which I am willing to allow my agent to execute, and then I hand off the rest of the strategy to the agent.
He will offer high-level intentions. There are many systems that deal with how many rules and protections I can have. There are many systems where it will offer such binary bits of ETH transactions. Good luck figuring this out figuring this out . Good luck making sure there isn't some bad idea that slips through, or some vulnerability in some extra argument that was optional in some operation.
This is not a semantic thing that a person will understand from a political perspective. What do you want, yes, you can. Enter this repository of creeds. Yes, you can go into this RWA Gauntlet repository. Yes, you can put it on base, or Arbitum, or Avalanche, or Salana, or wherever you want.
So what you need is for the agent's interaction with the smart agent's interaction with the smart contract to be in these easy-to- understand high intentions. Again, this is not a transfer of money to account 0x426, because that, as you know, can be a problem. This is a transfer of money to my account at the base, which is part of my portfolio. The smart portfolio. The smart contract must know what it is.
You understand that yourself, don't you ? And then the smart ? And then the smart contract checks whether this intent is consistent with the policy. And if and only if this is so, then he fulfills this high intention. He then takes it to the base and places it in this particular morphophobe.
No problem. I know how to do it. I don't need detailed EVM transactions. I need the ability to invoke a trusted relay like CCTP and then at the base put it into a specific font, and that gives you limited authority and execution, you know, with the high level of integrity that we expect from programs running in- chain. Good.
But the way most systems work now, of course, only allows this to be done on a single chain, or it has to be done off- chain, which has its own storage issues that have yet to be resolved through the courts. This has its own execution risks as you are operating in one jurisdiction and so on. You really want this to work in a chain. This is where the orchestration approach provided by the Agoric platform comes in, uh, which Agoric implements on the Agoric L1 chain, where one contract can perform an asynchronous multi-step action, keeping track of, “Okay, I’m going to move data using CCTP, it’s reliable; I'm going to use Axelr to send messages but not transfer assets to Base to say " Base to say " make this deposit, " and that way one contract can manage all this cross-chain activity. OK.
And that means that this mandate determines how much chain volume I can cover, how much money can move, how fast, by what routes, how much I'm willing to put in any particular vault, or how much we're willing to expose to any particular curator particular curator ? All of these things are implemented in one place that can be checked. And then the secure coordination and orchestration to make that happen across the entire Web 3 network is part of the execution infrastructure in the blockchain. This is what orchestration is. And this extends this mandate execution to a level of trust that encompasses the entire reachable orchestration domain.
And orchestration is an important scheme in many ways. I think the way we did it is the most powerful and scalable way. But there are other ways to do it. It's just a fundamental strength that we need to be able to say in one place, "This is what I want." do, "and then coordinate web 3 activities.
Because our customers, the users out there, when we go out into the much broader world, they're not really concerned with ETH versus Base versus Arbitrum. They're concerned with web 3 and investing in this fund or that asset. Well, as an example of this orchestration, we have our YAX product, which does this initially to optimize the accretion of returns. So, one signature as a user intent of "This is how I want to allocate my capital," and it goes out and creates child accounts on all these different chains, moves the money there in a reliable way, so you're not sitting there with this bridge risk and you can be interrupted in the middle of this complex deployment. So, one signature authorizes it, the money goes out, it deploys to all these repositories or, uh, DeFi services.
And so this is the application that's available now. And this is the agentless version that gives you control with a very simple interface, right? It's a high-level abstract rich, abstract rich, uh, the orchestration environment, it's a Rich Crosschain Rich Crosschain environment. Um, and so the environment. Um, and so the agent place where that goes, now it's where you create your portfolio that gives you this mandate.
These are the funds that my thing can participate in. These are the constraints on how it's allowed to trade. And then I set that mandate. So I sign the same thing. I have the same interaction with MetaMask or Rabby or, uh with MetaMask or Rabby or, uh , vault or whatever, and I submit it to the transaction.
All of this is non-custodial. Everything is controlled by my wallet. And then what my agent is, and then I delegate to my agent. I delegate to an open call. I delegate to Hermes.
I delegate to an agent in an environment that has rich, well- curated data services inside the YAX ecosystem. And the agent can then immediately go to work, looking at Twitter sentiment, looking at historical data, looking at current APYs, looking at what you have, and proposing new distributions; responding to provisioning that's starting to look a little questionable; responding to storage overflows, over-demand, and so they have a liquidity crisis. All of these things are to protect your security and respond to issues that could jeopardize your financial position, and also to take advantage of dynamic opportunities that are emerging. And so, it proposes these distributions, and, you know, along with that, this is one of the elements of our deployment where we had to work hard to get the agent to propose something that was outside of its mandate; but it did, and it actually got rejected on-chain, um. And so, it proposes them, submits them to an on-chain smart contract that checks them against the mandate that the user signed, and then does a cross-chain, uh, distribution of the yield.
And one of the things that, you know, it then uses, what we discovered with the deployment, uh, first round of YAX, which allows you to do deployments with one signature, is that the cross-chain yield is really better than the single-chain yield, right? Quarterly over the last many, many years, which chain has the best yield, right? And it's not because, you know, sometimes it's an incentive program, sometimes, well, you know, Poly, the market was launched on Polygon, and suddenly you needed a bunch of USDC, and Avalanche just launched an initiative next quarter, and so suddenly you need a lot of demand, and interest rates are rising to follow the economic activity in Web 3. People want that kind of access to Web 3 with a level of automation level of automation that allows them to not worry about the details, and that's what this achieves, okay? And so, you know, it's in alpha right now.
That's what we're working on with partners []; we have some initial integrations, we're developing an agent experience, and if you're interested in participating, anticipate some of that. Our booth is over there on the left. But let's move on from that, because this is really an example of a template. A smart contract in one place where you send a high-level intent, and then it checks it against the user's mandate and executes it if it's successful. Right it's successful.
Right ? The basis of this is smart contracts. Now let's go back, you know, to that earlier thing about encryption. Punks about encryption. Punks .
You know, a long time ago, when we first started doing smart doing smart contracts, we were not just talking about stand-alone code, but code that enforces the terms of a contractual agreement between multiple parties. Okay? So eBay, PayPal, Venmo, that's all the smart all the smart contract business before blockchain. They allowed strangers to collaborate, you know, and then AMM and Pers, it's now a rich, much more dynamic, fast-paced world of smart contracts that no longer requires a trusted intermediary like eBay to allow strangers to collaborate. And this is a huge, you know, multi-trillion dollar solution that's working for our economy right now, accelerates it and expands it across the world.
But there's this thing that I mentioned earlier, and that economists know about, and that's ironically called the principal- principal- agent problem, which is: I have an agent working for me. How do I know that he's going to do what I want him to do going to do what I want him to do ? And smart contracts now have this new urgency of enforcing these principal- principal- agent relationships not with strangers, but rather strangers, but rather , yeah, you know yourself and your agents, whether they're humans or computers or artificial intelligence. you need to agree on and enforce their scope. So it's, you know, kind of expanding what we can do.
And that's what creates the level of trust that we need to safely run agents, and that's what will allow us to expose them so that we can get to that agent world that people imagine, where the economy doesn't just run at the speed of computers but is constantly being enriched in a way that our agents can safely participate in and safely support this ever-changing landscape. So don't give away the keys to your AI. And you'll see a lot of that when you see all these open claws and Hermes and stuff. Give them a mandate. Make sure that what the user wants to do is done in an environment where you can safely let your agents play.
Thanks for your time, everyone, and please check out the YAX app on Twitter or stop by our booth over there. Um, I think you're going to love what this tool can do for you in terms of... in terms of capital allocation and profitability, and, um, join us with an agent in building the future of agency with this.
Automatic transcript — names and jargon may be misspelled.