Crossroads and Infinite Drip

So my name's James. I'm a PhD student at Cornell Tech. I'll be talking about, Crossroads, a smart contract platform for chain abstracted assets. So this is really more of a blockchain talk than the AI talks we've been hearing. Kinda of the main point is that cross-chain interactions when you're, when you wanna, for example, exchange assets across chains, they can be very scary.

If you are trying to exchange your Bitcoin for Ether, for example, you kind of just have to trust some intermediary. You send your coins almost into the void using some bridge, and you hope that you get them on the other side.

And then also, if you're trying to exchange assets that are very separate, they're on, like, different ecosystems, then you often get overcharged for it. For example, Thorchain might, charge you for this priority fee that you're, that you'd, that they don't have to necessarily pay for. Same thing on other chains when you are bridging, kind of lesser used assets like test network tokens. If there aren't many markets for it, then you, they, you get overcharged. So kind of the, whole concept of this system is we wanna to make some way of doing cross-chain interactions that are, not permissioned because cross-chain bridging is currently permissioned and they can, take away assets, on their platforms, either centralized exchanges or decentralized exchanges, and you have to basically ask for their blessing to integrate a new one.

And it's also really complex. So if There are ways of, doing cross-chain exchanges that are, decentralized, but they involve a lot of interactions. Like, you have to sign for five different blockchains if you're using Rango, for example. So anyway, what we're trying to do is we're trying to make, a platform for these chain abstracted assets. And the priority is we want them to be as low cost as possible, and we also want it to be permissionless so that anyone can integrate a new blockchain. Even a chain that only exists on your laptop, you should be able to also integrate that. And so in prior work, we showed how you can share blockchain accounts.

You can actually share a same private key with other people and hold onto different assets that are actually in the same account, using, what we call key encumbrance. So that kind We recognize that keys are off off-chain, and you can use basically a trusted execution environment as well as a policy, governing use of that key, and then you can have multiple people share the same wallet, basically. And we use that concept in Crossroads. We replace a TEE with just a threshold signing committee, and then we can now, we read a policy that exists on a smart contract, and now we can actually do cross-chain interactions.

And so what we do is we just take some EVM chain, and that's where we put all these smart contracts, and the smart contracts are what registers a new, cross-chain asset, and you can do your exchange. And if it's a fast finality chain with a low transaction cost, then it, it's a, pretty happy path. And then the interesting thing about Crossroads compared to all the other cross-chain interactions is that we only need this one-way oracle. We don't need, in order to, basically prove that deposits and withdrawals have happened. And since we're using key encumbrance, we can actually have multiple people signing transactions going out at the same time, and then the system itself has to disambiguate who actually got their transaction landing on the, target chain.

So anyway, yeah, that's kind of the broad architecture. It's kind of sloppy, but anyway, this will motivate the, real cool application that I have for Crossroads called Infinite Drip, which is a universal cryptocurrency faucet. Yeah, so, AI agents are on chain these days. So, there are estimates that there are millions of them, millions of agent-based transactions settling.

Thousands of AI agents live in the Web3 ecosystem. But the problem is, like, even if you just wanted to create a new application, like a new development application using, using AI, you have to send them some funds because they don't have any, and there's no way for them to get any. If you wanted an AI agent to get funds, they'd have to go through either a centralized exchange or on test networks through faucets, or they have to somehow already hold crypto for some, faucets that require a mainnet balance, or you, or you get it, funds from a hackathon.

And, this kind of goes against what, like, cryptocurrency was at the very beginning, where anyone could join with their own compute power, and they could receive Bitcoin all the way back at the beginning. But now access to all crypto is very gated. Ever since the proof of work chains like, like Ethereum switched to proof of stake, even their test networks are very centralized, and there's no way for broad participation, except through some disgusting anti-civil mechanisms like CAPTCHAs and, getting denied because they don't like your IP address, for example. So with Infinite Drip, we were trying to make a universal cryptocurrency faucet that's easy to use, trust minimized, and has no barriers to entry, no, and no limits.

So, unlike these cryptocurrency test network faucets that will limit how much you can get, we want with Infinite Drip you to be able to get an infinite number of coins. And the, key thing to recognize is that, all tokens have value, and this extends to test network tokens. There are, there are markets for them as well. And so kind of the point is we want to make this faucet so that the users have to provide value to compensate for the value that they get out of the faucet. And so something to recognize is that even though a lot of chains have switched to proof of stake or the proof of work, work ones are no longer as relevant, there are still some that exist and, some that have resisted ASICs when it comes to their proof of work algorithms.

So, one algorithm in particular, RandomX, is optimized for general purpose CPUs. So you can participate in consensus even with your computer or your laptop. And something to, realize is that, the, and the main one is Monero. There are some other ones, but you've probably never heard of them. And what's amazing is you can still get, like a non-zero amount of coins from them. So even just on my laptop, I can get five cents per day mining Monero. And this would be enough to send transactions even on main networks of some of these EVM chains in just a couple minutes. And for a test network, if you want to send at the base, twenty-one thousand gas price or gas, gas, amount at, a fee of around three wei, it takes only, equivalent of less than a second of mining Monero to, get that much.

So what we did is we, took this Monero mining idea and we, used trusted execution environments to realize, a cryptocurrency faucet that takes that proof of work and creates markets for all the tokens that you want out of, out of the faucet, and lets you exchange that value that you provide basically to a mining pool and exchange that for the tokens that you want.

And so Infinite Drip is essentially like this mining pool proxy. All the miners will connect and upstream it will connect to some Monero mining pool, and it will, and, you'll mine Monero basically for the pool, and the pool will help you exchange that for the token that you want. Yeah, so we have to deal with a whole lot of things though, because this is a TEE application. We have to worry about the, TEE application going offline. And the way we, do the exchange is we use Crossroads and we allow you to bridge new, new chains, and then you can basically The faucet will give you a mining pool token in exchange for work you've performed, and then you can just swap that through Crossroads, for a test network token or some other token that you want.

And this can all be done permissionlessly, through, through Crossroads. And Crossroads is what lets us represent assets on these other chains, just as ERC-20 tokens, and then we can just integrate with standard AMMs like Uniswap to trade for them. And the, pool will give you difficulty that's, that's relative to the amount of work that you're able to do, so you don't have to mine for a super long time. And the reason we connect to an upstream mining pool is that we want, faster payouts than waiting for a full block, and more more regular payouts also.

And then we can use a relayer to, actually carry out the swap for your test network token or other token. And then also the, way we establish value in the, in these markets that we're creating with this mining pool token that we're issuing is that you can actually redeem the mining pool token for the underlying Monero that was mined by the pool. So over time, this is kind of what it looks like when you, the exchange rate for this mining pool token with Monero, it's essentially, one-to-one if you have at least a couple dollars of liquidity in the, that the mining pool token has. And so we want to make this as trustless as possible, so we instantiated this in a TEE application container.

And the really interesting thing is it's completely self-funding and autonomous. So it will take a fee for some of the work that the miners do, and it will use that to actually pay for itself. So the pool will just continue running, autonomously. It's not like an AI autonomous, but it's execution autonomous. So it can continue essentially forever until, like for example, the Monero consensus changes or something. And what's really interesting is we, ended up using Oasis Sapphire, which was the cheapest TEE application provider we could find. It was only four dollars per month for their cheapest, container. And you only need about four weak laptop users on average, in fees to pay for the running of the system.

So essentially it can be autonomous as long as we have a small number of users. And then we also will have a CLI tool that you can use to, just, start mining, and we'll make it as easy for agents as possible. And then, yes, we, wanna, we want all the things to be a lot, tighter in the future, but, we think this will address the problem of really like the centralization of test networks and, blockchains in general and, make, access to crypto a lot easier for agents and for people alike. This was, this was part of a, larger team. We ended up winning first prize at the IC3 hackathon this summer.

And we are finishing up the actual implementation, and we, hope to have it released in the next month or so.

I'm gonna to have, one question for you, James. So James did the research briefing on Crossroad, the paper, and the, Infinite Drip, which is a pivot, since the beginning of the program where, Crossroads entered, Shape Rotator as a bridge, and a bridge only. I think the first feedback that all of us, gave, the team was perhaps there's less, marginal value in the one thousand and one bridge, that's out there. But the Infinite Drip, project, which is a product at week four, IC3 Bootcamp really kind of, brought, got our attention, in particular because of the interesting implication for, autonomous agents. And so you did mention that, there's a distinction in terms of, like, autonomous AI versus autonomous program execution.

So can you unpack this part, and also help posit what could an autonomous agent do using Infinite Drip, using Crossroads, and what can it not do?

Yeah, good question. So what Infinite Drip can do for agents is agents can get money. They can get test network tokens, and they don't have to ask for them. They can, they can basically provision smart contracts and run them and do all the blockchain things that you want them to do without you having to, pay for it to begin with. And also, as I mentioned, it, it's universal, so technically these main networks also are compatible, so they can, if they wanted to get access to crypto, by themselves, they could even do that. It would take a couple minutes of mining as long as they have access to, compute. But yeah, this project was actually designed for humans though, because for a long time I had, like, as my pet peeve, it was just how centralized even test networks were in the blockchain ecosystem, and I, wanted access to, crypto just for testing, and it ended up having to go through all these, real centralized or, civil resistant things that I, didn't agree with, like checking if you had a mainnet balance or something like that, having a Twitter account, which was kind of silly.

So I wanted it to be really back to the core principles of blockchain and make, crypto more accessible to, everyone.