fix(notebook): state the level-0 exception in the efficiency prior

The finished suite showed four holdout games clearing level 0 and still scoring 0.00: ft09 1633 actions vs baseline 43, cd82 2502 vs 55, vc33 289 vs 7, m0r0 1179 vs 30 - all about 40x, and (1/40)^2 is 0.06% of that level's points. So 'exploring level 0 is cheap' is true only relative to the later levels. If level 0 is the only level you clear, its efficiency IS your score. The addendum now says that, and says to stop exploring once the objective is visible - which is exactly the failure mode those four games display. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
arcprize · samrishtt · May 31, 2026 · May 31, 2026 · May 31, 2026 · May 31, 2026
commit 02377cd8866d5e230195128180080c98194dead6
diff --git a/1.33 scored in arc agi 3 competiotn in kaggle/arc3-duck-v13-priors.ipynb b/1.33 scored in arc agi 3 competiotn in kaggle/arc3-duck-v13-priors.ipynb
@@ -632,7 +632,11 @@
     "  clearing. Spend actions to establish the mechanics: which button moves what, what\n",
     "  blocks movement, what the objective is, what an action does when nothing appears\n",
     "  to change. The first level also carries the smallest weight of any level in the\n",
-    "  game.\n",
+    "  game. But \"cheap\" here means cheap *relative to the later levels*: if the first\n",
+    "  level turns out to be the only one you clear, then its efficiency is your entire\n",
+    "  score for this game. So explore it freely while you still do not understand it,\n",
+    "  and the moment you can see the objective, go and complete it rather than\n",
+    "  continuing to explore for its own sake.\n",
     "- Being crisp pays on later levels. Once you know the mechanics, execute. Later\n",
     "  levels carry more weight and charge every wasted action quadratically, and you\n",
     "  should re-derive nothing you already established earlier in the same game - the\n",

diff --git a/tools/make_submission_notebook.py b/tools/make_submission_notebook.py
@@ -230,7 +230,11 @@ def finish_game(self, generated_tokens: int = 0, uncached_input_tokens: int = 0)
   clearing. Spend actions to establish the mechanics: which button moves what, what
   blocks movement, what the objective is, what an action does when nothing appears
   to change. The first level also carries the smallest weight of any level in the
-  game.
+  game. But "cheap" here means cheap *relative to the later levels*: if the first
+  level turns out to be the only one you clear, then its efficiency is your entire
+  score for this game. So explore it freely while you still do not understand it,
+  and the moment you can see the objective, go and complete it rather than
+  continuing to explore for its own sake.
 - Being crisp pays on later levels. Once you know the mechanics, execute. Later
   levels carry more weight and charge every wasted action quadratically, and you
   should re-derive nothing you already established earlier in the same game - the