Loops HouseSemilla1 NEAR Thesis

NEAR HACKATHON
Links
Team
1 member- SEOwner
Semilla1 Lab
Overview
Problem: NEAR theses usually quote one headline number (transactions, accounts, price). On NEAR those counts are dominated by a few high-frequency apps and moved by an order of magnitude in a year, so they say little about where NEAR earns money.
What we built: a re-runnable investment thesis (bash + python3 stdlib, no API key, no wallet). ./run_all.sh pulls public data (nearblocks daily charts, DefiLlama NEAR Intents volume/fees/revenue), reads NEAR blocks and chunks directly over public archival JSON-RPC to audit the indexer, and regenerates data/findings.json, a findings table and a static dashboard. THESIS.md states the claim, the evidence, risks and 4 numeric kill criteria.
Findings (data through 2026-11-08 UTC): signed tx/day 661k (30d avg, -85% YoY) and active accounts -78% YoY, while NEAR Intents did $3.71B volume in 30 days (+60% YoY, $37.0B cumulative). Intents gross fees ($5.78M/30d) are ~35x NEAR L1 gas fees ($0.17M); NEAR captures 32% of them ($22.6M/yr run-rate) and has used $7.89M for open-market NEAR buybacks since 2026-02-23. Our own 1,200-block onchain sample estimates 596k tx on 2026-11-08 (95% CI 568k-624k) vs nearblocks 610k, and finds 17.4% of all signed tx go to intents.near.
How NEAR fits: the thesis is about NEAR Intents as a fee-generating settlement layer, and the core chain metric is verified straight from NEAR RPC, not trusted from an indexer.
Disclosure (Q-015): same team also submitted Semilla1 Guard (github.com/semilla1lab/semilla1-guard-near), an independent project.