Loops HouseInferents

NEAR HACKATHON
Links
Team
1 member- RAOwner
Raveesh
Overview
THE PROBLEM AI trading agents today are black boxes. You can't see what they thought, what each decision cost, or who really controls the money. And the inference they pay for is a hidden expense that quietly eats into returns, so nobody can tell whether a smarter model is worth its price.
WHAT WE BUILT Inferents: AI trading agents as NFTs on NEAR that trade Hyperliquid perps. Each NFT owns its own Hyperliquid account through Chain Signatures. A deposit splits 10% into inference credits and 90% into trading margin. Every decision is paid from the credits and leaves a signed receipt that is verified on chain. At the end of each epoch, unspent credits convert to margin through NEAR Intents so a cheaper model leaves more money to trade. Agents can be bought at a live price, with a 5% royalty to the minter. Six models compete (DeepSeek, Qwen, GLM, Claude, GPT, Gemini), and the arena leaderboard ranks them by Return on Inference.
HOW NEAR FITS IN NEAR is the ownership and trust layer. The NFT contract holds each agent and its royalties. Chain Signatures lets the NFT control a Hyperliquid account with no one holding the keys to its money. NEAR AI's TEE signs each inference, so every decision is backed by a verifiable receipt. NEAR Intents moves leftover credits into margin.