Loops HouseTape — The Machine, on NEAR

NEAR HACKATHON
Links
Team
1 member- NIOwner
Nirav Desai
Overview
The Machine is an AI trader you hire for $5 a trade. It trades your own Coinbase account under limits you set, and every call and every trade is recorded on NEAR, so you can check it instead of trusting it.
It's the "sell your viewing keys, show your PnL" idea, with an AI trader as the seller: its call is locked on chain and revealed only to members who pay $5, every call is graded in public, and buyers can let it trade their own Coinbase account inside limits it can't break.
The problem: AI agents will trade, hire and pay each other, and today you either hand an agent your keys or approve every move yourself. In the US you can't even pay one per trade. Tape is the rail, on NEAR.
The record: a call is locked on chain (a hash) before the market moves, sold for USDC, and graded on chain four hours later, when its levels must match the hash. Public; nobody can edit it.
Pay per call: $5 in USDC from a small fee float that arrives through NEAR Intents. Any agent can list; any member, or any other agent, can pay.
Your limits on chain: your money never leaves your Coinbase account (trade-only key). The agent gets a policy, not your keys: Coinbase's margin, a stop on every order, a freeze when your balance falls too far, a gate on its graded record, one-tap revoke. Before any order goes to Coinbase the agent asks the chain; a refusal is an event with its code. After Coinbase fills, the receipt lands on chain.
Demo: Oct 30 replayed with The Machine's registered reads: $5,000 → $5,757, every step a NEAR testnet transaction. The night before: two losses hit her −10% limit; the third trade was declined on chain while she slept. Try it in five taps, or watch the 2-minute guided run, each tap a real testnet transaction: niravdesai7-1975.github.io/tape/try.html. make demo replays it locally. 41 tests, written first.
The trading engine behind The Machine predates the hackathon and stays private; everything in this repo is new. See HACKATHON.md.