Show HN: I made an AI agent write a spoiler-free wiki for the novel Anathem (avoutarchive.com)

A fan-built site called the Avout Archive uses an AI agent to generate a spoiler-free, chapter-aware wiki for Neal Stephenson’s novel Anathem. Readers select the last chapter they’ve read and the site restricts browsing and search results to only the material revealed up to that point; pages that were updated in the chosen chapter are highlighted. Content is explicitly labeled “AI-Generated” and the developer warns of possible inaccuracies and creative interpretations. The project is unaffiliated with the author or publisher and invites issue reports for errors. Technically the site demonstrates a practical pattern for context-sensitive LLM applications: a chapter selector enforces information gating, search/browsing is scoped to the user’s state, and entries are surfaced by “frecency” (a blend of frequency and recency). That design makes the tool useful as a reading companion while minimizing accidental spoilers, but it also illustrates common trade-offs—LLM hallucinations, provenance and IP concerns, and the need for user feedback loops. For the AI/ML community this is a neat, low-friction example of deploying agents for controlled knowledge generation and UX-driven safety, showing how simple UI constraints and ranking heuristics can shape responsible LLM outputs.