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 | copilot.github.com | new | C → C0 | — | 0.0% | 0 |
| 2 | skype.com | new | C → C0 | — | 0.0% | 0 |
| 3 | messenger.com | new | C → C0 | — | 100.0% | 12 |
| 4 | washingtonpost.com | new | C → C0 | — | 100.0% | 13 |
| 5 | aljazeera.com | new | C → C0 | — | 100.0% | 13 |
| 6 | voanews.com | new | C → C0 | — | 100.0% | 11 |
| 7 | rferl.org | new | C → C0 | — | 100.0% | 11 |
| 8 | duckduckgo.com | new | C → C0 | — | 100.0% | 7 |
| 9 | expressvpn.com | new | C → C0 | — | 100.0% | 5 |
| 10 | vpngate.net | new | C → noise | — | 7.6% | 1 |
| 11 | wechat.com | new | C → C0 | — | 100.0% | 2 |
| 12 | tiktok.com | present_both | noise → C1moved | 2.780 | 100.0% | 34 |
| 13 | whatsapp.com | present_both | C0 → C1moved | 2.717 | 100.0% | 33 |
| 14 | chatgpt.com | present_both | noise → noise | 0.336 | 11.9% | 26 |
| 15 | medium.com | present_both | C1 → C1 | 0.287 | 100.0% | 21 |
| 16 | substack.com | present_both | noise → C1moved | 0.282 | 100.0% | 21 |
| 17 | psiphon.ca | present_both | noise → C1moved | 0.272 | 100.0% | 25 |
| 18 | claude.ai | present_both | noise → noise | 0.269 | 11.2% | 27 |
| 19 | chat.deepseek.com | present_both | noise → C0moved | 0.260 | 3.9% | 1 |
| 20 | google.com | present_both | noise → C1moved | 0.252 | 100.0% | 20 |
| 21 | tumblr.com | present_both | noise → C1moved | 0.251 | 100.0% | 20 |
| 22 | theguardian.com | present_both | C1 → C1 | 0.234 | 100.0% | 21 |
| 23 | telegram.org | present_both | C0 → C1moved | 0.226 | 100.0% | 25 |
| 24 | facebook.com | present_both | C0 → C1moved | 0.224 | 100.0% | 25 |
| 25 | reddit.com | present_both | C1 → C1 | 0.222 | 100.0% | 21 |
| 26 | nytimes.com | present_both | C1 → C1 | 0.219 | 100.0% | 21 |
| 27 | signal.org | present_both | C0 → C1moved | 0.218 | 100.0% | 26 |
| 28 | twitter.com | present_both | C0 → C1moved | 0.217 | 100.0% | 28 |
| 29 | getlantern.org | present_both | C1 → C1 | 0.214 | 100.0% | 23 |
| 30 | youtube.com | present_both | C0 → C1moved | 0.210 | 100.0% | 24 |
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: 2026-06-07T04:00:04.518444Z