Cairn

Community knowledge is dying in private tabs. Every specialized community creates knowledge worth remembering, but most gets buried in ChatGPT, Discord, and private notes. After 365 surveys and 32 interviews across three AI communities I help moderate (260,000 people), the pattern was clear: individuals drown, communities cannot retrieve what exists, and wikis die because no human can maintain them. Cairn is a community knowledge map for the age of AI. Inspired by Andrej Karpathy's LLM wiki, Cairn organizes knowledge into collections, groups, and cards. Members contribute Field Notes (scars AI cannot fabricate) and Takes (opinions with stakes). When someone asks, Opus 4.7 reads the collection and returns a sourced answer with a reading map. When the map does not know yet, Moss wakes up. Moss is a custodian agent on Claude Managed Agents. She detects blindspots, drafts sourced Briefs, and places them in a pending queue. Her mind is an auditable filesystem (SOUL, CONSTITUTION, MEMORY) the community can inspect, correct, and roll back. Cairn grows from readers' curiosity, with humans in charge of shared memory. Over time, the best curators become visible domain experts. Cairn is the foundation for a consultation economy around the experts each community surfaces.