Per-domain HDBSCAN drift leaderboard
Weekly unsupervised clustering of per-domain feature vectors (block rate, country footprint, dominant blocking method). Drift score = L2 distance between this week and last week in the same standardized space. A higher score means the domain's blocking footprint changed more.
Domains
—
Clusters this week
—
Clusters last week
—
New clusters this week
0
| # | Domain | Status | Last → This | Drift score | Block rate | n countries |
|---|---|---|---|---|---|---|
| 1 | freespeechdebate.com | new | C → noise | — | 12.5% | 2 |
| 2 | xxx.lanl.gov | new | C → noise | — | 0.0% | 0 |
| 3 | www.secfirst.org | new | C → noise | — | 0.0% | 0 |
| 4 | shareblue.com | new | C → noise | — | 75.0% | 1 |
| 5 | google.com | present_both | noise → noise | 10.641 | 100.0% | 22 |
| 6 | theguardian.com | present_both | noise → noise | 8.250 | 100.0% | 23 |
| 7 | nytimes.com | present_both | noise → noise | 7.896 | 100.0% | 23 |
| 8 | www.naral.org | present_both | C24 → noisemoved | 7.208 | 14.5% | 16 |
| 9 | www.qhtyzx.com | present_both | C3 → noisemoved | 7.155 | 15.9% | 18 |
| 10 | bricspress.live | present_both | C18 → noisemoved | 7.043 | 16.8% | 19 |
| 11 | www.righttodie.ca | present_both | C18 → noisemoved | 6.921 | 16.0% | 17 |
| 12 | bbc.com | present_both | noise → noise | 6.909 | 100.0% | 24 |
| 13 | psiphon.ca | present_both | noise → noise | 6.487 | 100.0% | 29 |
| 14 | getlantern.org | present_both | noise → noise | 6.445 | 100.0% | 26 |
| 15 | protonvpn.com | present_both | noise → noise | 6.269 | 100.0% | 30 |
| 16 | tumblr.com | present_both | noise → noise | 5.979 | 100.0% | 24 |
| 17 | instagram.com | present_both | noise → noise | 5.043 | 100.0% | 28 |
| 18 | reddit.com | present_both | C0 → noisemoved | 4.739 | 99.4% | 22 |
| 19 | tiktok.com | present_both | C0 → noisemoved | 4.329 | 99.5% | 26 |
| 20 | www.privaterra.org | present_both | noise → noise | 4.063 | 13.9% | 15 |
| 21 | facebook.com | present_both | noise → noise | 3.956 | 96.5% | 27 |
| 22 | substack.com | present_both | noise → noise | 3.534 | 23.4% | 23 |
| 23 | chat.mistral.ai | present_both | noise → noise | 3.407 | 50.0% | 1 |
| 24 | chat.deepseek.com | present_both | noise → noise | 3.406 | 0.0% | 0 |
| 25 | claude.ai | present_both | C0 → C0 | 3.354 | 11.1% | 27 |
| 26 | gemini.google.com | present_both | C0 → C0 | 3.336 | 12.6% | 28 |
| 27 | chatgpt.com | present_both | C0 → C0 | 3.335 | 12.1% | 28 |
| 28 | openai.com | present_both | C0 → C0 | 3.334 | 12.2% | 28 |
| 29 | chat.openai.com | present_both | C0 → C0 | 3.300 | 14.3% | 30 |
| 30 | youtube.com | present_both | C0 → noisemoved | 3.238 | 100.0% | 25 |
GET /v1/anomaly/domain-drift/leaderboard
Raw JSON of the same data
HDBSCAN drift finding
Methodology, weekly cron schedule, honest caveats about week-over-week stability
Why a separate domain-drift surface? The per-country DBSCAN at /atlas/anomaly catches country-day shape anomalies. HDBSCAN at the domain level catches global blocking-pattern changes — “is anyone blocking openai.com differently this week?” The two are orthogonal: one detects per-country novelty, the other detects per-domain novelty.
Generated: 2026-07-05T04:00:04.554630Z