Show HN: TownSquare, a tiny presence layer for websites

A tiny “presence layer” called TownSquare lets website visitors appear as little avatars in a shared space and exchange short messages, turning otherwise static pages into lightweight social hangouts. Early users praise the playful design and potential for community-building, but repeated trolling, spam, and offensive content highlight how hard real-time moderation is—especially for anonymous, embeddable tools. Commenters debate technical and policy approaches, from word filters and OpenAI’s free moderation endpoint to client-side controls, shadow bans, and pre-set phrase lists, underscoring that safety and control will determine whether such widgets are viable for most sites.

Overall reception

  • Many commenters find the concept delightful: a lightweight “presence layer” that makes websites feel inhabited and social.
  • Visual details (benches, tree, atmosphere) are widely praised; some users say they “hung out” for a while and had fun.
  • A few are wary due to the project being “vibe‑coded,” citing likely abandonment and potential security issues.

Moderation, abuse, and spam

  • The live demo quickly filled with offensive text, childish spam, and even browser‑freezing message floods.
  • Several note this as a predictable outcome for any unmoderated, anonymous public space; some call it “this is why we can’t have nice things.”
  • There is strong sentiment that robust moderation is essential: slur filters, word lists, bans/kicks, rate limiting, and spam controls.
  • The author states moderation features exist (block/ban, block word list) but were not fully wired into the landing page before the HN spike; more improvements are promised.
  • Debate arises about:
    • Whether “letting people talk” is viable versus driving out disruptive users.
    • How harmful words can be, both to people and to a product’s reputation.
    • Whether any word/concept filtering can truly work at Internet scale.

AI and moderation strategies

  • Multiple suggestions to use an LLM‑based moderation endpoint (e.g., OpenAI’s free moderation API) to classify content in real time.
  • Concerns raised about cost under attack, need for strong rate limiting, and risk of DoS via moderation calls.
  • Ideas include:
    • Shadowbanning and “shadow spaces” where bad actors unknowingly interact with bots rather than real users.
    • Client‑side configurable filters vs. site‑owner‑controlled moderation, with tension between user autonomy and curated site experience.

Design, UX, and feature ideas

  • Requests for:
    • Minimalist modes (simple presence counter, typing indicator, small chat).
    • Day/night and weather themes based on site owner’s timezone.
    • Rooms, RPG‑like spaces, minigames, or platformer/fighter variations.
    • Limiting visible users and pruning stale clients to prevent lag.
  • Some propose using WebRTC or P2P for presence/chat; others describe similar prior experiments and note hard network‑effect problems.