diff --git a/DECISIONS/2026-11-02-tracker-long-term-home.md b/DECISIONS/2026-11-02-tracker-long-term-home.md
new file mode 100644
index 0000000..ec1a1f9
--- /dev/null
+++ b/DECISIONS/2026-11-02-tracker-long-term-home.md
@@ -0,0 +1,48 @@
+# Decision: Tracker long-term home
+
+Status:  Decided
+Date:    2026-11-02
+Review:  Annually, or on trigger (see below)
+
+## Choice
+
+**The Outstanding Items Tracker Doc remains the master cross-cutting view.** GitHub issues are the master for individual work items. Each layer references the other.
+
+## How the layers work
+
+- **Tracker Doc** holds the cross-cutting view: all open decisions, deadlines, areas of work, presales, deferred items. One place to scan "what is still open."
+- **GitHub issues** hold the per-item state: acceptance criteria, evidence, PRs, labels. Where the work actually happens.
+- **`DECISIONS/` in the repo** holds resolved decisions as versioned records.
+- **`claude/project-status.md`** holds the periodic snapshot and decision log.
+
+## When each is used
+
+- D-items get a GitHub issue when they have a done-when that needs evidence beyond "decided", have code work attached, or need to be visible to others
+- Operational items (deadlines, bugs, reviews) always get issues; the tracker row is the index
+- Deferred-by-choice items stay as tracker rows only, no issue, because the point is not to re-litigate
+
+## Reasoning
+
+- Operated this way for 5+ weeks; the pattern works at current scale
+- 5 decisions closed on 2026-11-02 all followed this end-to-end
+- Alternatives (repo file, GitHub-only, PROJECT_MEMORY.md, mirror) have real costs and no clear benefit at current scale
+
+## Trade-offs accepted
+
+- Tracker Doc depends on Anthropic hosting Artifacts (tail risk)
+- Cross-session edits need an open Doc tab (manageable)
+- Mitigation: export the tracker to a repo snapshot quarterly, so a historical record survives if the Doc becomes unavailable
+
+## Reopen trigger
+
+Reconsider if any of:
+
+- Artifact hosting changes materially
+- Project grows beyond solo and new people need write access that is harder in a Doc
+- Tracker becomes too large to scan in one view
+
+## Related
+
+- 27 Oct 2026: `PROJECT_MEMORY.md` retired; tracker + `claude/project-status.md` became project memory
+- D1 (Path A, 18 Oct 2026): the "once D1 settles" condition in the standing recommendation is met
+- #156 (closed with this decision)
\ No newline at end of file