Loops HouseNEAR Real Usage

NEAR HACKATHON
Links
Team
1 member- EGOwner
egor
Overview
NEAR's headline activity (daily transactions, active accounts) is easy to inflate with reward apps, so it is a weak basis for a long-term investment case. NEAR Real Usage counts who actually acts on NEAR: a census of the accounts acting on the chain, month by month (NEAR's public BigQuery dataset, every transaction, not a sample), split by who pays the fee.
Findings (snapshot 2026-11-11): active accounts fell 97% in a year (43.6M to 1.38M a month), and the whole fall is accounts whose fees an app was paying; 30 accounts send half of all transactions; self-paid activity outside reward apps ran below the year before, on average, until autumn 2026; the NEAR that Intents revenue brings in equals 18% of the net new NEAR issued, and matching the issuance takes 5.6 times the revenue; usage did not lead the price in a Holm-corrected lead-lag test.
The repository is a reproducible pipeline: one Docker command runs the tests and rebuilds every chart and a static dashboard (each chart with its query and one plain sentence) from committed data. THESIS.md gives the claims, the evidence for and against, what would change our mind and what could not be computed. Nine dated predictions are frozen by hash and timestamped in a Bitcoin block; the build scores them.
NEAR is both the subject and the source: the NEAR BigQuery public dataset for the census, keyless FastNEAR endpoints for an independent block sample and the revenue wallets' balances, NearBlocks and DefiLlama for the daily series.