We already have that in the form of separate reasoning/thinking and speaking streams. Even with that it's awfully hard to get LLMs to keep it consistently concise. As soon as that context window starts growing it falls right back into verbosity without constant nudges back.
Right, I often bring up the film noir analogy for "reasoning" models, it's satisfying, like the revelation when a magic trick is explained, and many oddly disconnected questions about "why the scarf" or "where does the assistant go" all become sensible at once.
On a practical level, I believe more developers and adopters need these magic tricks spoiled, because otherwise they'll build a lot of important stuff on top of the idea that magic-is-real, leading to various forms of suffering in the long run.
That said, I'm no LLM / math academic, so if I'm totally wrong on the the trick, I'd like to know what needs revising.