> with a clear structure, good naming, and sensible comments. Additionally there is the additional problem, that LLM comments often represent what the code wou

> with a clear structure, good naming, and sensible comments. Additionally there is the additional problem, that LLM comments often represent what the code would be supposed to do, not what it actually does. People write comments to point out what was weird during implementation and what they found out during testing the implementation. LLM comments seems more to reflect the information present before writing the implementation, i.e. the use it as an internal check list what to generate. In my opinion deceiving comments are worse than no comments at all.