decisions: 121, Q129 Codex for about half of new coding work until 10-18

Co-Authored-By: Claude Opus 5.5 <[email protected]>

docs/plans/2026-09-26_lead-decisions.md

@@ -2721,3 +2721,25 @@ which stay with him. Each item names who decided it and what happened.
     - Switching live seats to the proxy and the subscription-terms
       question go back to Jason after the test. The #237 design row is
       written from row 96's report.
+121. **Codex for about half of new coding work until 10-18 (2026-10-12).**
+    Jason's Q129, through Mos (sage-0111), runs until the Codex reset on
+    2026-10-18 at 01:00 CDT. New coding work splits about 50:50 between
+    Claude and Codex. Codex uses gpt-6.1-sol, with gpt-6-sol as the
+    fallback and gpt-6-luna only for very low-level tasks. Never Astra
+    (R26). No builder switches mid-task.
+    - Rocko and Researcher go to Codex, each in a new T3 thread on
+      gpt-6.1-sol. The thread is created with that model named, because
+      Codex defaults to gpt-6-astra.
+    - Rocko switches once row 77 lands. A round 3 on row 77 stays on the
+      current Claude thread. On Codex they take rows 94, 85, 84 and 88.
+      Row 88 moves from Darkwing to Rocko to even the split.
+    - Researcher switches once row 86 lands and takes row 96 on Codex.
+    - Darkwing, Dewey and Filbert stay on Claude. They are mid-task on
+      rows 92, 89 and 95 and keep rows 63, 91, 71, 80, 90 and 93.
+    - That gives five new rows each way: 94, 85, 84, 88 and 96 on Codex;
+      90, 91, 93, 71 and 80 on Claude.
+    - Reviews stay with the assigned reviewers, so Claude seats review
+      Codex builds. The Codex config has the t3 server, so Codex seats
+      message the same way. agents/README.md gets each new thread id
+      when it launches.
+    - The order doesn't change: #232, then #228, then the redeploy.