I had the same feeling. It's like every sentence becomes a struggle to parse: referencing things that haven't been introduced, using too many pronouns, the lack of verbs and the passive tense... These past few days it's felt more like Opus 5.1 than 5.5.

I just had Claude Code with "Opus 5.5" ignore one of my standing instructions for the first time and then proceed to generate an Opus 5 style verbose and barely readable commit message. I questioned it about Fable and Mythos and it has no idea what those were, yet they should have been in the Opus 5.5 knowledge cutoff.

On the web with both high and xhigh, it had knowledge of both.

Yes, I've observed the same. It's a shame.

Yep, its prose is getting really bad

In my experience, no amount of output style customization gets it anywhere near the output I'd like.

I've had more success running a prose linter[1] with a ruleset[2] that gets rid of most of the egregious stuff. You can also pretty easily write custom rules for it.

And sometimes, when I'm fed up enough, I just throw all the prose in the codebase at Antigravity/Gemini to rephrase. The result is so much more readable.

[1]: https://vale.sh/

[2]: https://github.com/tbhb/vale-ai-tells

You can change it ... and Claude will maybe pay attention to what you changed 25% of the time.

"You can tell an AI model to write in a different style" and "the AI model will actually write in the style you tell it to" are two different things that often don't intersect

Agree. It’s still awful. Gemini by far sounds the most natural IMO