diff --git a/README.md b/README.md
index 1cb305e..5c89b01 100644
--- a/README.md
+++ b/README.md
@@ -144,8 +144,9 @@ behind each verdict is unit-tested in `test_audit_classifier.py`.
 ## Upstream
 
 The behaviour was reported as [microsoft/jericho#84][issue] and confirmed by a
-maintainer; an upstream fix targeting a major release is tracked in
-[microsoft/jericho#88][pr].
+maintainer; the fix is [microsoft/jericho#88][pr], merged 2026-08-14 for the
+v4.0 major release. Every release through 3.3.1 carries the silent default,
+including the 3.2.0 this artifact pins.
 
 [issue]: https://github.com/microsoft/jericho/issues/84
 [pr]: https://github.com/microsoft/jericho/pull/88
diff --git a/test_reading.py b/test_reading.py
index 3157895..6f364eb 100644
--- a/test_reading.py
+++ b/test_reading.py
@@ -77,7 +77,7 @@ def test_committed_run_credits_randomization(artifact):
 # --------------------------------------------------------------------------
 
 def test_post_fix_run_reports_the_confound_absent():
-    """The branch that matters once microsoft/jericho#88 lands.
+    """The branch that matters on an engine carrying microsoft/jericho#88.
 
     Numbers modelled on a real run against a patched build: the unseeded default
     draws a fresh seed per episode (recorded as -1) and spreads across states, so