LLMs are terrible at tracking exact line numbers. This format uses context matching instead - it finds where to apply changes by matching surrounding text, not by trusting line numbers. The find_context_core() function searches for matching lines rather than jumping to line N.
