in my experience most code begins its life as prose, whether it is pseudo-code in my head or the pure natural language of CEO/CTO-speak, project notes, task descriptions, PRDs, statements of work, feature requests, bug reports, brainstorming sessions, notebook scribbles, user stories, sprint objectives, and other human-sourced content.
along the way, ideas are translated from the human realm to the human-computer shared-language realm of computer programming. a well-placed comment of intent by a well-meaning but slightly-ignorant ;) co-worker can help reveal that a paragraph of code is mis-implemented.
in my opinion, the "coupling" of comments and code can be a good thing, as long as both are both concise and local. but the purpose of the comments should not be to document the intricacies of your particular choice of implementation, but a general paraphasing of the task at hand.