diff --git a/scrape.py b/scrape.py
index aae4172..b5aa799 100644
--- a/scrape.py
+++ b/scrape.py
@@ -33,7 +33,7 @@ SQUARE_BRACKETS_PATTERN = re.compile(r"^\s*\[([^\]]+)\]\s*$", re.MULTILINE)
 BALL_PATTERN = re.compile(r"●\s*(.*?)\n", re.MULTILINE)
 REFERENCE_MARK_PATTERN = re.compile(r"^\s*※\s*(\S.*?)\s*$", re.MULTILINE)
 ESCAPED_STAR_PATTERN = re.compile(r"\\\*(.*)", re.MULTILINE)
-NON_BREAKING_SPACE_PATTERN = re.compile(r"[\xa0\u00a0\u2002\u2003\u2009]")  # Various nbsp characters
+NON_BREAKING_SPACE_PATTERN = re.compile(r"[\xa0\u2002\u2003\u2009]")  # Various nbsp characters
 EMPTY_CODE_BLOCK_PATTERN = re.compile(r"```[ \t]*\n[ \t]*\n```")
 
 # Circled number patterns - precompile for better performance
@@ -459,7 +459,7 @@ def generate_atom_feed(articles: list[dict[Any, Any]], file_name: str) -> str:
         content = REFERENCE_MARK_PATTERN.sub(r"\n\n*\1*\n\n", content)
         
         # Replace circled Unicode numbers with plain numbered text (using precompiled patterns)
-        for symbol, (number, pattern) in CIRCLED_NUMBERS.items():
+        for number, pattern in CIRCLED_NUMBERS.values():
             content = pattern.sub(rf"\n\n{number}. \1\n\n", content)
         
         content = ESCAPED_STAR_PATTERN.sub(r"* \1", content)