With C++ formatting is optional. A better test case for LLMs is Python where indention specifies code blocks. Even ChatGPT 3.5 got the formatting for Python and YAML correct - now the actual code back then was often hilariously wrong.
Well there’s your issue!
Clearly it isn’t just “broken” for everyone, “Claude Code modernizes a legacy COBOL codebase”, from Anthropic:
In this case, a five stage pipeline, built on demo environments and code that were already in the training data, was successful. I see more red flags there, than green.