Loops HouseIntents Thesis

NEAR HACKATHON
Links
Team
1 member- AROwner
Arpit
Overview
Problem: Claims that NEAR is becoming a cross-chain settlement layer via NEAR Intents usually rest on volume headlines. A headline does not show whether a quote is there when a user asks, or how far the returned amount sits from a price anyone can see.
What we built: A pre-registered, falsifiable test published as a static report. On 2026-11-14 we sent 1,240 dry: true quotes to NEAR Intents 1Click over 62 cycles (ten routes across NEAR, Ethereum, Solana, Bitcoin, Tron, Base and Arbitrum, two sizes each) and compared each token amount with a Kraken cross-mid from the same cycle. Pass and falsify bars were written in PREREGISTRATION.md before the sample.
Results: C1 availability passed: 99.4% of quotes returned, all 10 routes at 70% or above. C2 price passed: stable routes median -2.1 bps versus Kraken, volatile -28.1 bps, with the 25 bps unauthenticated fee left in. C3, descriptive: DefiLlama's NEAR Intents volume was contracting, last 30 days -21.5% versus the prior 30. Reading: the quote layer answered and priced close to market, but public volume was not expanding.
How NEAR fits: NEAR Intents 1Click is what is measured. The method and dataset manifests were SHA-256 hashed and anchored on the testnet contract intents-thesis-d163.testnet (txs 8LMEu3g9MpxDhBoZnJeH2Crkh7GEda7G5xtyJrnYmNPX and 2fKHAWgThhs2QaUit8SD5QbuqZubHC9zM3buomuthQ8L), so the data cannot be quietly edited. scripts/verify_anchor.py checks the anchors on chain.
Limits: Three hours on one day is a snapshot, not a trend. Dry quotes are not settled swaps. The comparison is mid against mid. The repository holds the built report, its data files and the verify script; the collector, analysis scripts and raw rows are not included, so the dataset hash cannot be recomputed from the repo.