diff --git a/docs/assistant-intelligence-roadmap.md b/docs/assistant-intelligence-roadmap.md
index f16ab7a..91b7d16 100644
--- a/docs/assistant-intelligence-roadmap.md
+++ b/docs/assistant-intelligence-roadmap.md
@@ -84,9 +84,9 @@ significant query words before falling back to the full KDB.
 2. Add latency and recall-budget gates to the assistant harness:
    https://github.com/tsotchke/gpt2-basic/issues/41
 3. Add an inspectable persistent conversation database:
-   https://github.com/tsotchke/gpt2-basic/issues/38
-4. Add a pack intent router for assistant queries:
    https://github.com/tsotchke/gpt2-basic/issues/39
+4. Add a pack intent router for assistant queries:
+   https://github.com/tsotchke/gpt2-basic/issues/38
 5. Add larger programming and offline-reference domain packs:
    https://github.com/tsotchke/gpt2-basic/issues/40
 6. Define Windows 95 and OS/2 assistant shell parity:
diff --git a/docs/public-launch.md b/docs/public-launch.md
index 115392e..1023e65 100644
--- a/docs/public-launch.md
+++ b/docs/public-launch.md
@@ -143,8 +143,8 @@ Non-hardware follow-up is tracked separately so software work can continue
 without waiting for physical captures:
 
 - Dense recall storage: https://github.com/tsotchke/gpt2-basic/issues/37
-- Persistent conversation database: https://github.com/tsotchke/gpt2-basic/issues/38
-- Pack intent router: https://github.com/tsotchke/gpt2-basic/issues/39
+- Pack intent router: https://github.com/tsotchke/gpt2-basic/issues/38
+- Persistent conversation database: https://github.com/tsotchke/gpt2-basic/issues/39
 - Larger domain packs: https://github.com/tsotchke/gpt2-basic/issues/40
 - Latency and recall budgets: https://github.com/tsotchke/gpt2-basic/issues/41
 - Windows 95 and OS/2 shell parity: https://github.com/tsotchke/gpt2-basic/issues/42