Opus 5.5 went back to the old writing style a few days ago.

I really didn't like Opus 5 and was vocal about it. Opus 5.5 is noticeably better, but literally just yesterday I was fighting parsing some summary it had given me and it was a bit like the old days. IOW, it is better but still has some room to improve.

You can easily change its writing style: https://code.claude.com/docs/en/output-styles

You can also customize further by putting notes in your global MD.

opus 5.5 is still unbearable but better than 5.

I enjoy a lot Argon, maybe more than chatgpt, but damn, like all Gemini models, lots of hallucination and sycophant to the max, so a bit useless for agentic coding.

sol is ok. it still such a superthinky model that it gets tangled up trying to present way too much? i struggle to connect with it often, have ti ask it to try again.

terra has been prettt good for me for me, when i need something looked at an explained to me.

i dont ask for much non-technical output though, which i expect yeah is its own cans of worms.

How much of this is bad writing style, versus a homogenization of global grammar conventions.

Almost every response it’s telling me “I could not verify X” where X is something that doesn’t need verifying. I was doing some game work yesterday filtering through some historical data, and it tried to put caveats about data accuracy in the game’s UI! If Fable made Age of Empires you wouldn’t be able to play it because the entire screen would be full of footnotes and caveats.

Opus 5.5 got better for me but it still has some problematic behavior.

Interestingly, you can have informative conversations about the LLM linguistic tics with the LLMs themselves. One word that chatgpt used to describe what I was observing was linguistic anaphora (not rhetorical). It’s basically shortening/compression achieved via pronoun, and Claude does a lot of it. Worth reading about. A lot of it seems to do with how the new models are trained for brevity.

Another pattern I’m starting to become very sensitive to is Claude making up on the fly jargon for something under consideration, especially in code review. This morning for example it referred to “a claim” out of nowhere, and when I questioned it, it admitted that it was using the noun as shorthand for a record getting dequeued (“claimed”) inside of an sql transaction. It makes sense if you already know what it is referring to but in a code review it’s totally incomprehensible on first read, and ultimately counterproductive.

I still find Opus 5.5 way too dense. MUCH better than Opus 5 which was unreadable.

Even the Plain English hack only is a temporary solution.

I don't think its this. Some models across both labs have had issues with writing and communication and it typically seems to pop up on successor models of really strong ones.

Opus 4.7/4.8 heavily suffered from these communication quirks. while the Sol models at the time worked well. Right now its flipped though and Opus 5.5 has none of these issues while Sol 6.1 does.

Small example I've had today: "Me: Hey Sol 6.1 xhigh, does X have property Y? Sol 6.1: Yes, if we change X to have property Y. Right now it has property [...], [...], [...], and it does not have property Y." The model shouldn't be answering questions like that. It makes understanding extremely hard and slow. For whatever reason it's either being encouraged or at least not stomped out during RL.

The best programmers I've met or read throughout my life are phenomenal communicators. I'd even go so far as to say it's likely one of the reasons they're so good at programming. After all, the majority of our job is converting human requirements in lossy language to fleshed out behaviors in formal, lossless language.

Heck, some of the most common debates are on the semantic clarity of abstractions or syntax.

[dead]