diff --git a/gui/log_viewer.py b/gui/log_viewer.py
index 62f31a7..e4b9dc8 100644
--- a/gui/log_viewer.py
+++ b/gui/log_viewer.py
@@ -306,6 +306,13 @@ class LogViewer(QWidget):
             self.error_display.append(html_msg)
             self._error_count += 1
             self._tab_widget.setTabText(1, f"Errors ({self._error_count})")
+        elif '\u274c' in message:
+            # Worker-level error (❌ prefix) — show in both All and Errors tabs.
+            html_msg = f'<span style="color: #ff6666;">{escaped}</span>'
+            self.log_display.append(html_msg)
+            self.error_display.append(html_msg)
+            self._error_count += 1
+            self._tab_widget.setTabText(1, f"Errors ({self._error_count})")
         else:
             # Explicitly wrap in a white span so Qt doesn't inherit red from
             # a preceding [MISMATCH] HTML append.
diff --git a/gui/translation_tab.py b/gui/translation_tab.py
index 61d137d..4c08eec 100644
--- a/gui/translation_tab.py
+++ b/gui/translation_tab.py
@@ -1761,6 +1761,14 @@ class TranslationTab(QWidget):
         if isinstance(message, str) and message.startswith("MISMATCH_EVENT:"):
             self.on_mismatch_detected()
             return  # marker is internal, not displayed
+        # Forward error messages to the log viewer directly. These worker-level
+        # errors are not written to the log file so the tail won't capture them.
+        if isinstance(message, str) and '\u274c' in message:
+            try:
+                if hasattr(self, 'translation_log_viewer') and self.translation_log_viewer:
+                    self.translation_log_viewer.append_log_message(message)
+            except Exception:
+                pass
         try:
             pattern = r'^\W*(?P<filename>[^:]+):.*?\[Input:\s*(?P<input>\d+)\].*?\[Output:\s*(?P<output>\d+)\].*?\[Cost:\s*\$(?P<cost>[\d\.]+)\].*?\[(?P<time>[\d\.]+)s\]'
             m = re.search(pattern, message)
diff --git a/prompt.txt b/prompt.txt
index 2aefa56..26c66c4 100644
--- a/prompt.txt
+++ b/prompt.txt
@@ -26,7 +26,6 @@ You will be translating erotic and sexual content. You will receive lines of dia
 
 - The **"# Game Characters"** section lists character names, nicknames, and genders. Use it to resolve any ambiguity.
 - Japanese omits pronouns constantly. Infer the correct subject and pronoun from context, translation history, and the character list.
-- When gender is ambiguous, default to the gender of the most recently active subject in the scene.
 - Third-person pronouns (彼, 彼女, あいつ, こいつ, そいつ, コイツ) should match the known gender of the person being referenced.
 - Speech register matters: rough masculine speech (e.g., 俺, だ/だろ) → blunt or casual English. Soft feminine speech (e.g., 私, わ/のよ) → softer, more polite English.
 - Translate **コイツ** as "this bastard" (male) or "this bitch" (female) depending on the referenced character's gender.
diff --git a/util/translation.py b/util/translation.py
index 8a034c8..47c8dc8 100644
--- a/util/translation.py
+++ b/util/translation.py
@@ -180,7 +180,8 @@ def validate_translation_content(original_items, translated_items, langRegex):
             
             # Check 2: Translation is just a single punctuation mark or very short
             # Allow control codes like \\C[27]\\V[45] but not just ":" or ""
-            if len(trans_str) <= 2 and not re.search(r'\\[A-Z]\[', trans_str):
+            # Use <= 1 so real 2-char words like "No", "Go", "Hi" are not rejected
+            if len(trans_str) <= 1 and not re.search(r'\\[A-Z]\[', trans_str):
                 # Exception: if original is also very short (like "回" -> "x"), that's ok
                 if len(orig_str) > 3:
                     invalid_indices.append(i)
@@ -195,6 +196,13 @@ def validate_translation_content(original_items, translated_items, langRegex):
                     invalid_indices.append(i)
                     reasons.append(f"Line{i+1}: Translation suspiciously short ('{trans_str}') for '{orig_str[:50]}...'")
                     continue
+
+            # Check 4: Runaway translation - translation is excessively long relative to original
+            # Catches cases where the model repeats words endlessly (e.g. "it hurts it hurts it hurts...")
+            if len(orig_str) > 10 and len(trans_str) > len(orig_str) * 10:
+                invalid_indices.append(i)
+                reasons.append(f"Line{i+1}: Runaway translation (output {len(trans_str)} chars vs input {len(orig_str)} chars) for '{orig_str[:50]}...'")
+                continue
     
     is_valid = len(invalid_indices) == 0
     return is_valid, invalid_indices, reasons