UnisonShared vs private memory
The Unison visibility model: write location decides who sees it - /private/ for the caller, /workspace/teams/ for a squad, /workspace/ for the whole workspace, with explicit promotion.
Unison has no permission matrix to administer. Where an agent writes is who sees it: /private/ is caller-only, /workspace/teams/<slug>/ is squad-wide, /workspace/ is workspace-wide. Sharing is an explicit act - either write to a shared path, or promote later with POST /v1/brain/share.
Why scope-by-path
Agents make thousands of writes; per-item ACL decisions don't survive that volume. A path prefix is a decision an agent can reliably make at write time ("is this scratch, squad context, or workspace truth?"), it's visible in every listing, and it's auditable by grep.
Multi-user services
With actor delegation, one service key serves many end users: each X-Unison-Actor id gets an automatically isolated /private/ namespace, while /workspace/ remains the shared layer. That's the workspace-brain shape: private working memory per actor, one shared source of truth above it.
Knowledge graph
Entities, bitemporal facts, and links: how Unison stores what is true, when it became true, and how knowledge connects.
Context rot
Context rot is the quality degradation that hits AI agents as long, unfiltered context accumulates - and why externalizing memory beats stretching the window.