diff --git a/ai-hub/app/core/orchestration/stream.py b/ai-hub/app/core/orchestration/stream.py
index 8e831f0..578fbd0 100644
--- a/ai-hub/app/core/orchestration/stream.py
+++ b/ai-hub/app/core/orchestration/stream.py
@@ -18,7 +18,7 @@
     async def process_chunk(self, content_chunk: str, turn: int) -> AsyncGenerator[Dict[str, Any], None]:
         """Processes a raw content chunk, yields UI events (content, reasoning)."""
         self.tag_buffer += content_chunk
-        turn_header = f"---\n### 🛰️ **[Turn {turn}] Thinking...**\n"
+        turn_header = f"---\n🛰️ **[Turn {turn}] Thinking...**\n"
 
         while self.tag_buffer:
             if not self._in_thinking_tag:
@@ -146,7 +146,7 @@
         return text
     async def end_stream(self, turn: int) -> AsyncGenerator[Dict[str, Any], None]:
         """Flushes any remaining buffered text at the very end of the stream."""
-        turn_header = f"---\n### 🛰️ **[Turn {turn}] Thinking...**\n"
+        turn_header = f"---\n🛰️ **[Turn {turn}] Thinking...**\n"
         
         # 1. Flush prefix buffer if we haven't sent the header yet
         if not self.header_sent and self.prefix_buffer: