Bluesky's stackable approach to moderation

Original Article ↗ Hacker News Discussion ↗

Overall reaction to stackable moderation

  • Many commenters find the “stackable / pluggable” moderation model innovative and overdue, likening it to programmable blocklists, ad‑block filters, Steam curators, or usenet killfiles.
  • People are excited about use cases beyond safety: spoilers, unwanted topics, personality/tone filters, “chiller” labels to cool down heated threads, and niche filters like “Spider Shield”.
  • Others argue moderation is not additive but should be a cohesive ruleset for a community, and worry this could devolve into chaotic, individualized filters and a poor default experience.

Architecture and decentralization

  • A Bluesky team member explains labels are metadata attached by “labelers”; clients choose which labelers to use and how strictly to act on them.
  • There is a distinction between:
    • Client-side label-based moderation (opt‑in/stackable).
    • “Infrastructure takedowns” at the relay/aggregator level for illegal content and abuse.
  • Commenters compare this to Mastodon and Reddit:
    • Bluesky: modular roles (identity, PDS, relays, app views, feeds, labelers) with user choice at each layer.
    • Mastodon: identity and moderation bound to an instance; heavy inter‑instance blocking.
    • Reddit: scalable via subreddit boundaries and local mods, but with mod abuse and hard locks.
  • Skeptics note the aggregator/relay remains a practical central point; running alternative relays is technically possible but may be economically unattractive, risking de‑facto centralization.

Free speech, censorship, and law

  • Strong split between those who see this as empowering “freedom to ignore” and those who see any “Trust & Safety” layer as censorship risk.
  • Some worry that centralized “community guidelines” plus infra takedowns recreate older platforms’ problems, just with an extra layer of indirection.
  • Others counter that:
    • Completely unmoderated systems become unusable due to spam/abuse.
    • Users can choose clients/labelers with different norms, including more permissive ones.
  • Legal issues (CSAM, terrorism content, hate speech, jurisdictional conflicts) are heavily debated:
    • Hiding via labels is insufficient where mere possession/distribution is illegal; infra takedowns are still required.
    • Handling region-specific illegality via country-specific moderation channels is proposed but seen as imperfect.

Echo chambers, polarization, and incentives

  • Some fear configurable filters will harden echo chambers and polarization.
  • Others argue people already curate their spaces; composable moderation just makes this explicit and user‑controlled.
  • There is concern about:
    • Who labels the labelers and how users evaluate/“meta‑moderate” them.
    • Whether there are strong incentives for independent relays and labelers, or if the ecosystem will default to one dominant provider.