Labeler: Anti "Anti-AI" Labeler (did:plc:s45hbf5dqdkjpwuuq4djo6l2)

Methods and caveats

• Behavioral observation only. No content analysis, user profiling, or brute-force enumeration beyond declared/observed/protocol surfaces.
• Rules detect geometric patterns (rate anomalies, target distribution, churn). Thresholds are configurable.
• Every alert includes a receipted hash over rule_id, config, inputs, and evidence for reproducibility.
• Sparklines show hourly event counts over 7 days. Badges summarize which rules fired in the period.
• Observation surface: label events from com.atproto.label.queryLabels and labeler declarations from com.atproto.sync.listReposByCollection. Does not observe content, profiles, or social graph.
• Classification is based on structured evidence from multiple surfaces (registry declaration, DID document, endpoint probing, observed label activity). Each labeler page includes a "Why classified this way" expander with full evidence.