diff --git a/CLAUDE.md b/CLAUDE.md
index 2400818..2755273 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -167,12 +167,13 @@ the single-chunk path, 98.028% through the production region replay. A featurele
 (`cmd/vanillacapture -featureless -blocks-only`, biomes stripped to their carvers) proves the
 undecorated pipeline bit-exact against it — density, surface rules, carvers, aquifers, and
 noise-router veins match every one of the fixture's cells — so the residual block gap is entirely
-inside feature replay. Plain-state mismatches that once looked like base-terrain defects are
-air-writing features the replay does not cover yet: monster rooms (stage 3) open cave_air pockets,
-geodes leave voids, lava lakes fill stage-1 cavities. Vanilla runs those stages before the ores, so
-ore ellipsoids crossing that pre-existing air roll air-exposure discards our run never sees, which
-is why high-attempt ore drift looks symmetric (extra ≈ missing). The next step toward exact parity
-is replaying the remaining pre-ore stages, monster rooms first.
+inside feature replay. Monster rooms (stage 3, `world/monster_rooms.go`) now replay between geodes
+and the ores with vanilla's draw order, so the air pockets ore ellipsoids roll discards against
+exist where the schedule puts them; the fixture's four chunks contain no dungeons, so the measured
+number did not move. The unreplayed air-writing stages are lava lakes (stage 1) and the remaining
+structure sets: the fixture carries ruined-portal and mineshaft pieces (obsidian/gold/planks
+markers), and a cobblestone-and-water pocket in chunk (-1,-1) that no replayed feature explains —
+attributing it is the open lead for closing the gap further.
 
 ## Testing worldgen
 
diff --git a/README.md b/README.md
index 098defc..1e51361 100644
--- a/README.md
+++ b/README.md
@@ -118,14 +118,15 @@ exposing undecorated chunks.
 The remaining block-parity gap is entirely inside feature replay: a vanilla
 capture with every feature and structure set stripped out matches our
 undecorated pipeline on all 393,216 cells, so density, surface rules, carvers,
-aquifers, and noise-router veins are already bit-exact. What drifts is the
-interaction between stages — vanilla opens air pockets (monster rooms, geode
-voids, lava lakes) before the ores run, and ore ellipsoids crossing that
-pre-existing air roll air-exposure discard rounds our replay never sees, which
-is why high-attempt ore drift looks symmetric. The next worldgen milestone is
-to replay the remaining pre-ore stages, monster rooms first, and raise the
-98.028% block parity toward exact equality while keeping cold batch generation
-within an acceptable latency budget.
+aquifers, and noise-router veins are already bit-exact. Monster rooms now
+replay from their datapack configuration between the geodes and the ores, so
+the air pockets ore ellipsoids roll discards against exist where the schedule
+puts them; the committed fixture happens to contain no dungeons, leaving the
+measured parity at 98.028%. What still does not replay is the stage-1 lake
+schedule and the structure sets — ruined portal and mineshaft pieces are
+visible in the fixture — and those are the next worldgen milestones on the way
+to exact equality while keeping cold batch generation within an acceptable
+latency budget.
 
 ## Project layout