The plain matrices hold on a tightly written prompt

RELEASE_NOTES_v3.md

@@ -54,12 +54,13 @@ That costs 0.0027 of pure-text cosine on the residual and 0.0013 on the matrix
 which is why the 8B figures above are slightly below what a text-only corpus
 would have given.
 
-**Prefer the `-mlp` files.** They measure closer to the 32B, 0.9449 against
-0.9289 on the 8B, and on the demo scene both plain matrices get the seated pose
-wrong: the prompt asks for one bare leg crossed over the other and they sit her
-cross-legged. But the gap is narrower than that sounds — the cat, the straw hat,
-the laundry and the bouncing knee all come through on the plain matrices too. On
-this scene they missed one attribute among many, and it happened to be the pose.
+**Prefer the `-mlp` files on the measurement, not on this scene.** They sit
+closer to the 32B, 0.9449 against 0.9289 on the 8B. But on this prompt the five
+renders are faithful, plain matrices included: the pose, the dress, the white
+pieces, the cat, the straw hat, the laundry, the bouncing knee all hold on all
+five. A tightly written prompt survives even the linear baseline, and the
+difference between the files shows up in the numbers well before it shows up on
+screen.
 
 ## What this does not fix, and will not