A catalog of the pains that surface when knowledge crosses multiple contributors - extending [[Knowledge problems]] with a collective layer. [[Tacit knowledge]] is the hardest of the catalog to see: it is never absent from a list because it was never on a list, and its loss becomes visible only when someone who held it leaves. ## The pains > "Can the right access boundaries be enforced - who can read, who can edit, which artefact can reach which agent?" ^securability When the answer is no, the artefact is either over-shared (anyone in the org sees it, including those who shouldn't) or under-shared (only the author sees it, defeating the point of capture). The third axis - which agents, human or machine, can be fed this artefact - is the AI-era extension: a wiki page with no permission boundary passes through an LLM context window to anyone who can prompt the agent. > "Is it unambiguous who creates, reviews, and retires it?" ^ownership-clarity When the answer is no, the artefact has no one on the hook for keeping it true. No one updates it because no one feels qualified; no one retires it because no one is sure if it is still in use. The artefact decays in plain sight while everyone defers to "whoever wrote it" - a person who may have left two years ago. > "Is it structured so that agents can consume it without reformatting?" ^agent-readiness When the answer is no, agents either fail (cannot parse arbitrary HTML or scanned PDFs) or fabricate (the LLM guesses where the parser could not). The cost is invisible until an agent returns a confidently wrong answer; then it cascades through every downstream consumer of that answer.