Yea, this seems to be a consistent issue. The models don’t know when to think strategically and often make local edits that keep growing individual functions with more and more conditional logic.