diff 8236acf..d0806bf ramanujan_arthur_cuspidal_v2.txt
diff --git a/ramanujan_arthur_cuspidal_v2.txt b/ramanujan_arthur_cuspidal_v2.txt
index 84acbe3..f7ded54 100644
--- a/ramanujan_arthur_cuspidal_v2.txt
+++ b/ramanujan_arthur_cuspidal_v2.txt
@@ -20,6 +20,19 @@ works". A fake proof cannot be fixed by editing. The honest
 correction is to state what those conjectures say, what is
 actually proven, and where the file goes wrong.
 
+----------------------------------------------------------------------
+"WHERE IS THE FULL PROOF?"
+There are exactly two cases in this story where a complete,
+published, real proof exists and you can hold the pages:
+  (1) the classical bound |tau(p)| <= 2 p^{11/2} for Delta is
+      Deligne 1974, together with Weil I (1974), in full;
+  (2) Ramanujan for GL(n) over a function field F_q(X) is
+      L. Lafforgue 2002, in full (240 pages).
+For a general split reductive group beyond GL(n) over F_q(X),
+and for GLR 3.4.5/3.4.6 at full finite level, no complete proof
+exists. That is the gap. The rest of this document maps it.
+----------------------------------------------------------------------
+
 I cannot hand you a full proof of GLR 3.4.5/3.4.6 or of the
 generic Ramanujan conjecture for split adjoint groups over
 function fields. Nobody can, because both are open. What I can
@@ -959,8 +972,6 @@ in two ways:
 That is what the original file claims to prove. Sections 10,
 11 will make it precise and explain why it is open.
 
-======================================================================
-
 ======================================================================
 SECTION 10. THE GLR CONJECTURES, HONESTLY STATED
 ======================================================================