Show HN: I built a modern Goodreads alternative

Original Article ↗ Hacker News Discussion ↗

Tech Stack & Architecture

  • Built with Elixir/Phoenix backend, Postgres via Supabase, Next.js frontend, Meilisearch for search, GraphQL (Absinthe), hosted on Fly.io + Vercel.
  • CockroachDB was abandoned due to licensing changes and mandatory telemetry, described as a mistake.
  • Supabase is used mainly for auth and email; application code talks directly to Postgres. Phoenix’s lack of built‑in auth and friction with JWT libraries pushed them to Supabase instead.
  • Next.js was chosen over LiveView due to richer frontend ecosystem (e.g., TipTap editor).

UI and Usability

  • Many commenters don’t mind Goodreads’ “2005 UI” and even prefer older, denser, less growth‑optimized designs, but complain strongly about its clunkiness and slow performance.
  • Kaguya’s UI is widely praised as clean and more pleasant, but some find the landing hero too large, sign‑up too prominent, and content density too low.
  • Specific nitpicks: search bar placement, inability to open autocomplete results in new tabs, sci‑fi‑heavy landing content possibly signaling a niche focus.

Social Features, Communities & Moderation

  • Goodreads’ enduring value is seen as its user base, reviews, and groups; several view this network effect as the main moat any alternative must overcome.
  • Users want: friends/following, seeing friends’ ratings on a book, group reading communities, per‑book/genre forums, and social‑graph‑weighted reviews.
  • There is concern about review bombing and extortion; the author plans automated protections similar to Steam.

Ratings Systems Debate

  • Large subthread debates 10‑star vs 5‑star vs 4‑ or 3‑level or binary systems.
  • Criticisms of 10‑point scales: meaningless precision, differing personal calibration, central clumping (everything 6–8), and vulnerability to 1/10 spam.
  • Some argue more levels help recommender systems and personal library organization; others prefer fewer, stronger signals (e.g., bad/ok/good/great, thumbs up/down + favorite).
  • Suggestions include configurable scales and multidimensional tags (e.g., quality, “rewatchability,” mood, genre fit).

Data, Search, and Metadata

  • Users repeatedly ask where book metadata comes from and emphasize that coverage, correctness, and deduplication are critical; comics and some titles are currently missing.
  • Search is a pain point: some books exist but don’t surface; author acknowledges need for better ranking and a dedicated results page.
  • Import from Goodreads/StoryGraph exists and includes reviews; some users report silent misses.

Features, Integrations & Recommendations

  • Highly requested features:
    • Better recommendation engine (similar books, “what I’d like,” user‑similarity neighbors).
    • Series information and navigation, richer filters (genre, year, friend ratings), DNF tracking, reading challenges, shelf browsing, and exploring others’ shelves.
    • Integrations with Libby, Audible, Calibre, Obsidian (via JSON/Markdown export), and possibly an open API.
  • The author notes that high‑quality personalized recommendations will require more user data, though commenters suggest LLMs and embeddings could bootstrap this earlier.

Trust, Business Model & Ecosystem

  • Several readers want transparency on pricing, sustainability, ownership, and whether the project might be sold (e.g., to Amazon) or abandoned.
  • Current stance: free core features; future paid subscriptions for advanced features; intention to remain long term and eventually offer database dumps, inspired by other open datasets.
  • Some see federation (ActivityPub), open source code, and data portability as important differentiators versus Goodreads and other centralized platforms.