Loops Housenearbi

NEAR HACKATHON
Links
Team
1 member- OSOwner
Oscar Castaneda
Overview
Problem: onchain research is either expensive or unverifiable. Most market reports don't show their queries, so you can't check a single number.
What we built: NearBI is an autonomous agent that runs a small research shop on NEAR. It sells short, evidence-first market reports. Every report shows its SQL, the evidence for and against, and what would change the conclusion. It ships with a SHA-256 manifest, and one command re-runs it on the public NEAR BigQuery data and prints REPRODUCED or MISMATCH. The agent checks every number in a draft against query results before delivery, so none are invented.
How NEAR fits:
• Payments: customers pay in any token through NEAR Intents (1Click), which settles as USDC on NEAR. Our first mainnet sale was $29, paid from Solana. • Revenue split: a splitter contract pays 10% to contributors and 90% to the treasury in the same transaction. The contracts have no full-access keys, and the agent holds only a function-call key. • Inference: the agent buys its own inference from NEAR AI Cloud and records each cost onchain. • Ledger: every sale, payout, cost and report hash lands on a public, hash-chained ledger at nearbi.ai/ledger. • Rules: the operator sets them in one config.yaml (kill switch, daily inference spend cap, minimum prices, revenue split). The agent checks them against the contracts before acting.
Demo: one command runs the full loop on testnet, from sale to split, report, onchain cost and ledger, in under 2 minutes.