> The named pains from the previous act tell you where the bleeding is. This act decides what would actually stop it. The pains are clustered by the kind of knowledge each is about, each cluster is profiled (who creates it, who consumes it, how volatile, how connected, what is the cost of it being wrong), and a quality bar is written: > 1. what the knowledge must do > 2. for whom > 3. at what cost if it fails. > > Requirements first, tool decisions last. ^purpose What "knowledge" means for [[Decision record|decision rationale]] - rare, high-consequence, easily lost - is almost unrecognisable next to what it means for procedural know-how: high-volume, repetitive, ambient. One method cannot serve both. The act clusters the pains by knowledge kind, profiles each cluster against shared questions (volatility, audience, connectivity, cost of failure), and names a steward for it - because requirements without named ownership are aspirations. The clustering is what does the structuring work for the next act: by the time the room arrives at method-matching, the design buckets are already drawn in their language. ^method The informal brokers identified during the previous act are the natural candidate stewards for the clusters they already serve. **You leave with:** - A profile for each cluster - volatile vs. stable, [[AI agent|agent]]-consumed vs. human-only, connected vs. self-contained - A quality bar per cluster - what the knowledge must do, and which requirements are non-negotiable - A first pass at stewardship - who already does the work informally, surfaced and named ^outcomes ## Examples of clusters profiled here 1. [[Second brain example - personal notes|Second brain (personal notes)]] ![[Second brain example - personal notes#^profile-line]] 2. [[Shared second brain example - ADR log|Shared second brain (ADR log)]] ![[Shared second brain example - ADR log#^profile-line]]