Show HN: Struct – A Feed-Centric Chat Platform

Original Article ↗ Hacker News Discussion ↗

A new feed‑centric chat platform called Struct aims to replace traditional channel-based tools like Slack and Discord by making every conversation a thread, then surfacing them in customizable feeds with AI‑generated titles, summaries, and search. Commenters are intrigued by its potential to reduce information loss and overload, especially for long-running projects and community support, but raise concerns about reliance on OpenAI, privacy and compliance transparency, lack of mature integrations and APIs, and the ever-present risk of creating yet another silo. The exchange also highlights broader fatigue with “AI‑powered” products, differing preferences for real‑time chat versus forum‑style communication, and the practical hurdles of persuading organizations to migrate from entrenched tools.

Overall reception & concept

  • Many find the “feed + threads” approach compelling, especially for surfacing important conversations and reducing Slack/Discord chaos.
  • Others feel a unified “All Threads” feed is overwhelming and prefer channel-based separation.
  • Several note it feels closer to forums/Discourse/Zulip or “message boards with real-time chat” than to classic IRC-style chat.

Threads, feeds, and UX design

  • Struct treats channels mainly as permission groups; conversations live as threads in feeds.
  • Custom feeds (by tags, people, channels) are seen as powerful; critics worry about distraction and loss of clear “rooms.”
  • Some users perceive the interface as impersonal and lacking a sense of presence; suggestions include surfacing avatars/online status more.
  • Comparison with Zulip: others argue Zulip already has “all messages”/inbox views, but the Struct team stresses a different, feed-first real‑time design.

AI features and concerns

  • Positive: summarization and automatic thread titles are widely seen as a strong, practical AI use case; AI search over years of chat history is attractive.
  • Negative: several are fatigued by “AI” marketing, worry about unreliability and hallucinations, and dislike dependence on OpenAI.
  • The AI bot for Q&A over past threads is optional; core differentiator is positioned as feeds/threads, not AI itself.

Privacy, security, and compliance

  • An early copy bug around “Privacy” messaging created suspicion; this was acknowledged as a CMS mistake and corrected.
  • Concerns raised about data going to OpenAI, lack of self-hosting, unclear legal entity details, and GDPR/SOC2 expectations.
  • The team says OpenAI business terms prevent training on customer data and that SOC2 and better transparency are planned.

Integrations, migration, and ecosystem

  • Strong interest in Slack integration as both a migration path and a “Superhuman for Slack” client; it already syncs threads, with partial history import and plans for export tools.
  • Discord integration exists but currently indexes only threads, not channel chats; Slack channel chats are auto-threaded.
  • Teams, Matrix, and a broader plugin/API ecosystem are requested; Teams integration is considered but seen as a hard market.
  • Many emphasize that deep app/bot integrations are critical for leaving Slack.

Pricing and business model

  • Pricing is per-org with a base fee plus usage-based AI token billing; some praise the transparency and see it as cheaper than Slack.
  • Requests include clearer tables, concrete usage examples, and hard monthly caps to avoid surprise bills.
  • Debate arises over SSO being a “premium” feature in general SaaS pricing; some argue SSO should be default, others see it as a fair high-tier feature.

Technical implementation & reliability

  • Stack details shared: Go, Postgres, OpenAI plus Microsoft embeddings, Typesense, React/Next.js, Hetzner hosting, Tauri desktop app.
  • OpenAI timeouts and Windows installer/redirect issues are reported; improvements to error handling and signing are promised.

Broader reflections on communication

  • Some question whether better tools are the right optimization versus simply having less communication/meetings.
  • Others argue that real-time chat and O(N²) team communication are here to stay, so structuring and retrieval matter more than raw reduction.