Loops HouseMarginal

NEAR HACKATHON
Links
Team
1 member- JAOwner
Jayanth Kumar
Overview
Problem: you can't price information without revealing it (Arrow's paradox), so markets for research memos, datasets and eval sets either leak the goods or make buyers pay for things they already know. Marginal prices an information good by what it adds for this buyer. The seller seals a memo; the buyer seals its own notes next to it in an attested TEE appraisal room; deterministic appraisers measure the marginal value. If it's new to the buyer, the deal clears and the memo decrypts in the browser with its commitment checked. If not, the buyer gets a full refund and a forget certificate bound to the room's attestation. NEAR does the trust-critical work: an immutable NEAR contract plus NEAR's MPC network (Confidential Key Derivation) release the listing key only for one paid job; the clearing contract never trusts the room with the price and recomputes the pricing rule itself (Open mode) or verifies a Groth16 proof over alt_bn128 (Sealed mode); each appraisal debits a leakage budget and pays the seller per bit revealed, which stops sybil probing; buyer agents are capped on-chain against prompt injection; cross-chain payouts quote through NEAR Intents 1Click. The README's "Real and simulated" table states exactly what runs today (local sandbox) and what is simulated.