← back

Layer N x Pond—Building AI Trading Agents as XVMs

May 29th, 2024 by Layer N

Testnet Live

We’re partnering with Pond (opens in a new tab) to explore AI-powered trading agents built with our XVMs. This represents another big step toward implementing a vast range of AI use cases across Layer N.

Pond: Crypto-Native AI

Pond is building a Graph Neural Network (GNN) – a predictive engine powered by onchain data.

It starts with representing Web3 as a graph. Wallets, smart contracts, DIDs, and other onchain entities form the nodes, connected by edges representing their relationships. The result is a massive spider web of users, transactions, and smart contract interactions. From this, Web3 itself becomes a massive repository of training data for artificial intelligence.

Pond uses this graph to train its neural network. Just as Generative Pre-trained Transformers (GPT) are trained on giant language data sets to predict the next word, GNNs are trained on graphs to predict the next outcome. The result is a crypto-native, AI-powered prediction engine built from and for Web3.

Pond on Layer N

Together, we’re implementing AI XVMs. That’s almost 20% of the alphabet. So, what does it mean?

Layer N’s unique architecture is built around XVMs – dedicated virtual machines for individual applications. By building a separate VM for each dApp, we can create custom logic for each one to improve performance, while still connecting to the rest of the ecosystem via our shared communication layer.

We’ve already built AI functions into XVMs via our partnership with Modulus (opens in a new tab). Any rollup in our ecosystem can deploy an AI model. By running the rollup and AI function separately, we get verifiable outputs without compromising on speed. Pond’s GNN will be deployed on Layer N as a dedicated XVM using this AI tooling.

Our first project together is a trading agent, which predicts price action based on onchain data. The XVM runs the execution of each trade, making calls to Pond’s models using our AI function tooling. Since it’s deployed as an XVM connected to the rest of Layer N, the agent will be able to act on its predictions, making calculated trades on Layer N exchanges like Susa and Nos.

Follow Pond (opens in a new tab), and we’ll keep you updated on our progress here, and via X (opens in a new tab), Telegram (opens in a new tab), and Discord (opens in a new tab).