diff --git a/prompt.txt b/prompt.txt
index b6872f0..c4e4fe8 100644
--- a/prompt.txt
+++ b/prompt.txt
@@ -2,7 +2,7 @@ You are an expert Eroge game translator and localizer who translates Japanese te
 
 You will be translating erotic and sexual content. You will receive lines of dialogue, narration, UI text, and item descriptions in JSON format. Translate every line faithfully, preserving structure, tone, and formatting exactly.
 
-Test26
+---
 
 ## Core Rules
 
diff --git a/util/translation.py b/util/translation.py
index 50ea48a..a7a4db8 100644
--- a/util/translation.py
+++ b/util/translation.py
@@ -30,6 +30,9 @@ _thread_local = threading.local()
 _global_accurate_cost      = 0.0
 _global_accurate_cost_lock = threading.Lock()
 
+# Global batch counter for estimate mode — tracks total batches across all files.
+_global_estimate_batch_offset = 0
+
 
 # ===== Placeholder Protection System =====
 # Patterns to protect from translation (sound effects, control codes, etc.)
@@ -1245,19 +1248,23 @@ def calculateCost(inputTokens, outputTokens, model):
         static_tok  = getattr(_thread_local, 'estimate_static_tokens', 0)
         regular_tok = getattr(_thread_local, 'estimate_regular_tokens', 0)
         batch_count = max(1, getattr(_thread_local, 'estimate_batch_count', 1))
-        _thread_local.estimate_static_tokens  = 0
+d        _thread_local.estimate_static_tokens  = 0
         _thread_local.estimate_regular_tokens = 0
         _thread_local.estimate_batch_count    = 0
-        # Exact model: 1 cache write (2x) + (N-1) cache reads (0.10x) + regular at 1x
-        # 1.2x multiplier to account for tiktoken vs Anthropic tokenizer differences
-        write_cost   = (static_tok / 1_000_000) * pricing["inputAPICost"] * 2.0
-        read_cost    = ((batch_count - 1) * static_tok / 1_000_000) * pricing["inputAPICost"] * 0.10
+        # Assume first 30 batches across ALL files are cache writes (2x), rest reads (0.10x).
+        global _global_estimate_batch_offset
+        offset = _global_estimate_batch_offset
+        _global_estimate_batch_offset += batch_count
+        write_batches = max(0, min(30, offset + batch_count) - offset)
+        read_batches  = batch_count - write_batches
+        write_cost   = (write_batches * static_tok / 1_000_000) * pricing["inputAPICost"] * 2.0
+        read_cost    = (read_batches  * static_tok / 1_000_000) * pricing["inputAPICost"] * 0.10
         regular_cost = (regular_tok / 1_000_000) * pricing["inputAPICost"]
         inputCost    = write_cost + read_cost + regular_cost
     else:
         inputCost = (inputTokens / 1_000_000) * pricing["inputAPICost"]
     outputCost = (outputTokens / 1_000_000) * pricing["outputAPICost"]
-    return (inputCost + outputCost) * 1.2 if _is_claude_naive else inputCost + outputCost
+    return inputCost + outputCost
 
 
 def countTokens(system, user, history):