DOWNLOADS · PROVENANCE · VALIDATION
Data
Everything on this site is reproducible. Our derived dataset is CC BY 4.0 — cite "Clanked (clanked.ai)". Upstream sources carry their own licenses, recorded per-file in the manifest.
Downloads
| File | Contents |
|---|---|
| rankings.json | All 393 metros: consensus, band, rank, grade |
| /data/cities/{slug}.json | Per-metro: index scores, top occupations, scenario grid, beta histogram (e.g. austin-tx) |
| manifest.json | Provenance: every source file's URL, vintage, license, SHA-256 |
| scenario_constants.json | The full scenario model: parameters, defaults, presets, docs |
Validation report
Generated by the build's gate suite (2026-06-11); the site cannot deploy if a gate fails.
PASS — Brookings San Jose anchor: 42.9% vs published 43% (tolerance ±2)
PASS — San Jose = most exposed large metro: rank 1
PASS — Las Vegas bottom decile of large metros: percentile 5 (LV level: 22.5% vs Brookings 31% — knife-edge at beta 0.5 threshold + May 2025 vs May 2023 vintage; documented)
PASS — Coverage >=50% everywhere: min 58%
PASS — <=5 C-grade metros: 2 C-grades
PASS — Programmer/writer family in top-10 consensus: ['15-1251', '15-1254', '27-3042', '27-3091']
PASS — Agricultural metros at bottom: ['Salinas', 'Visalia', 'Hanford-Corcoran', 'Yakima', 'Bakersfield-Delano']
PASS — 2026 modeled displacement <1.5%: max 0.16%
PASS — Provenance manifest complete: all key inputs manifested
Citation
@misc{clanked2026,
title = {Clanked: AI exposure by metropolitan area},
year = {2026},
url = {https://clanked.ai},
note = {Data vintage: BLS OEWS May 2025}
}