Loops HouseNEAR, Filtered

NEAR HACKATHON
Links
Team
1 member- SAOwner
Sambit Sargam Ekalabya
Overview
The problem: most claims about NEAR's growth are made on raw on-chain numbers that bots and airdrop farmers inflate, and are rarely tested against a result that could prove them wrong.
What we built: NEAR, Filtered, a pre-registered, reproducible investment thesis. We wrote five falsifiable hypotheses (real usage, concentration, retention, cross-chain shift, value accrual), fixed the exact metrics, thresholds and rejection criteria, and committed them to git before pulling any data. Four of five went against us, and we published that with the evidence against ahead of the evidence for. Bot-filtered economic accounts fell 69.6% over 12 months, robust across all 27 pre-registered filter settings. The one survivor: cross-chain settlement share through NEAR Intents rose from 1.73% to 10.73%, but mostly because native activity collapsed (Intents +11.6%, native transfers -83.7%). Fee burn is 1.42% of gross issuance, and no documented mechanism routes Intents fees to holders.
How NEAR fits in: all data comes from the NEAR public BigQuery dataset and the NEAR Intents Explorer API, and the thesis tests whether NEAR's shift to cross-chain settlement creates value for NEAR holders. Every number is generated from results/metrics.json, and a number-drift test fails the build otherwise. Anyone can verify it in two minutes with make verify: no cloud account, no keys, no network. Live dashboard: near-filtered.vercel.app. Technical walkthrough: youtu.be/yO0XIfxgYvY