# AI-facing documentation needs an explicit authority boundary ## Claim AI-facing documentation needs a configured authority boundary that excludes speculative sources from the product's accepted behavior. <!-- atomicity: 1 idea ✓ --> ## Evidence > [!quote] > "AI tends to go to GitHub issues and collect stuff that are not merged yet and just fantasize about the tool that we all want." The proposed configuration marks which sources count as authoritative, keeping unmerged requests and imagined behavior from being presented as current product truth. ## Scope & conditions This applies when an assistant can retrieve from mixed-status material such as released docs, merged code, open issues, and proposals. The boundary has to encode source status; a general instruction to prefer trustworthy material is too vague to separate shipped behavior from aspiration. ## Links - [[3. Resources/Permanent Notes/AI Search Inherits the Trust and Fragility of the Sites It Cites]] — source authority is the configuration-level answer to retrieval systems inheriting weak or speculative evidence - [[3. Resources/Permanent Notes/Sampling a Fraction of a Source Produces False Authority Over the Whole]] — both prevent confident answers from outrunning what the retrieved evidence can support - [[3. Resources/Permanent Notes/Bounded Action Spaces Not Better Models Make Agents Safe to Run]] — authority configuration bounds what an agent may treat as fact just as action controls bound what it may execute - [[3. Resources/MOCs/AI Agent Safety MOC]] — home for retrieval authority and bounded agent behavior - [[3. Resources/MOCs/Documentation & Technical Writing MOC]] — applies-in: the docs discipline map, "docs that agents read" cluster ## Source - [[3. Resources/Readwise/Articles/Milana Cap About WordCamp Belgrade, Tech Events, and Open Source Contributions Kinsta Talks]] — describes a config file that distinguishes accepted sources from unmerged GitHub issues