Everything is super succinct. Opus 5 lands, it almost completely disregards the intent.
I suppose watermarking requires a certain text mass.
Just those two words. I use it A LOT recently.
Maybe just don’t generate garbage in the first place?
My theory is that Claude's learned approach to comments is to treat them as a sort of persistent in-band thinking trace, or a "memory" tied to an in-code location, which is a little at odds with the way humans use comments (human comments are intended to be read and understood by other humans, whereas Claude comments are their own dialect).
I bet this is a result of iteratively training Claude on output from other successful Claude sessions. Presumably it's good for making benchmark scores go up.
Claude will include actual comments ("// ...") into Excel sheets, and include the thinking that led to the output, instead of just focusing on the final result.
So if Claude questioned whether a vendor should be replaced, and you said "oh no, they are critical and we're already negotiating a great price") you'll now need to be careful to not send your vendor a document that contain text like ("Cost: X. // Management confirmed to not fire this vendor as they are critical to infrastructure and a better price will be negotiated later")
My CLAUDE.md has rules about not including any redundant comments in the code that are obvious from the code itself. I reiterate that occasionally while working. It's absolutely disregarded and any Claude-written code is full of comments. Some of them are simply redundant, like "Collect Foos and pass them to the requested sink" on a function that's void CollectFoos(IFooSink sink). But worse, many comments include in the moment reasoning like "added parameter bar because we can no longer use the frob to automatically derive bar". That's stuff for a commit message, or just a mental note, and absolutely not for comments.
I haven't found any way to stop Claude from doing these, so I have to tell Claude afterwards to clean the comments up. Which it does, making a note in memory to comment less, and it still does the exact same thing next time.
Exactly my experience! Since the release of Opus 5, no amount of instructions helps. In CLAUDE.md, in a separate file, in memory, as brief bullets, as long detailed guides, with reasoning from medium to max — nothing.
Even worse, recently, after getting another opus in a tiny bugfix session, I prompted directly, "drop the comments from the current code changes" — Claude instead just slightly trimmed them. I couldn't believe my eyes.
I have a relatively low bar for prose, could live with some junk. But Claude's comments are _poisonous_. They always require maintenance, instantly become out of sync with the actual code, and are a token black hole — for all agents, but especially for Claude itself.
Gave up and canceled Anthropic subscription yesterday. To my taste, it has become unusable for coding.
For me, Claude knows how I want the comments due to all the memories and CLAUDE.md, so funnily it's now enough with even a brief groan from me like "Come on, the comments" and then Claude goes through its recent additions and fixes comments quite well per my long-term instructions. But only ever during an extra pass that I initiate, never during the initial writing of the code.
I've noticed this a lot, and before your remark I couldn't put my finger on what was wrong. Now I know: Claude is writing its thought processes and maybe parts of the conversation it had with you as comments in the code!
I always end up manually trimming those comments, which is cumbersome.
// load_tree() loads the binary tree with data, but only the recently updated data, not all data (INTERNAL_NOTES.md section 4)
Given the code base has a minimal amount of such comments, it's also less likely to go "copy what the rest of the codebase does".
Of course I've now jinxed it and some update will cause it to ignore the instructions coz I didn't write them in the new model's style or something.
Edit:
I've had explicit instructions for communication style in CLAUDE.md, in Claude's project "memory", in global "memory", in "skills": it couldn't care less where it was. It would just ignore it.
When I would point this out it would just say "Yes, I violated communication guidelines, I won't do that again". Only to do that again in the next session.
This applies to everything: code guidelines, communication guidelines, preferences, decisions etc.
Absolutely infuriating if you’re using Claude in an environment where you can’t run hooks.
https://platform.claude.com/docs/en/build-with-claude/prompt...
Also using Codex or Pi makes you realise how slow and clunky the cc harness is. Even the desktop app is more responsive and has better UX.
Funny how quickly the tides change.
This is something that annoys me working in companies over the years. It’s that you can't just suggest "calm down, chasing the latest thing will not make you faster and is a huge distraction to actual work". Whether it's dot-com tech 20 years ago, latest JS framework 10 years ago, now it's the AI thing of the day. Being calm is interpreted as anti-whatever.
But I agree, the GPT models are so much simpler to work with, they have so much less personality and fewer quirks. They also are a little less aggressive about triple checking every little assumption immediately in a stack of 30 tool calls (but I haven't used 5.6 Sol yet so maybe that's not true anymore).
I doubt this is the reason. The fact that Chinese labs are all distilling Claude/GPT/etc isn't exactly a well kept secret, they don't even bother removing the name "Claude" from the training data, so the models randomly refer to themselves as "Claude" all the time.
I think it's far more likely to be a side effect of how much synthetic data is being fed back into the models to make them better at coding. The degradation of Claude's prose has been gradual but steady ever since they shifted towards focusing only on code with Opus 4.5.