I've been grappling with this for weeks, not just in Claude but in Codex as well, which isn't quite as bad but still annoying. AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on. It's incredible to me that there's no good way to reliably change the way an LLM responds to you that a workaround like this would even be necessary. It seems like such a failure to live up to the promises of the product.
The baked in communication style of these models is so obnoxious it's impacting my work. The best way I can describe it is that everything is optimized to impress the user and make the agent sound more authoritative, but the way this is done is through deliberate obfuscation, inserting inappropriate and extremely dense jargon, and bizarre, stilted metaphors. It's like they've been trained to produce output that's hard to read.
> hello i would like to configure a new output style for you. it should keep the coding instructions (as you will still be coding!) and otherwise produce the same output, but with two new caveats. first, long detailed replies are still permitted, but if employed they must end in a bullet pointed summary whose points are all brief; if the summary attempt ends up not being so brief, produce subsequent summaries until the most recent summary attempt is digestible. second, if there is an open queue of actions for me to execute and you are about to end a turn to wait for a reply or this set of actions has not recently been mentioned, please tabulate the open actions i should take and why i should take them before ending the response. does this make sense or do you have any follow up questions
And now every message contains the same stuff I don't bother reading, but followed by a nicely formatted bullet point summary of the response and a table of follow up actions for me to take that I do read.
I'm probably going to be going against the grain here, but I think it's not as bad as it looks at first.
I was similarly frustrated a few months ago, but have noticed I've started to learn the idiom.
Its use of "dense jargon" and "stilted metaphor" is actually surprisingly consistent - it's speaking its own dialect, and you get used to it.
After a while it gets much easier to read and even becomes somewhat efficient, I think, since the odd metaphors it uses often have a precise meaning in Opus-ese (Fable speaks a really similar dialect).
One danger in acclimating to this style of communication style is that we may accidentally use it in your own communication with other people. If the other person hasn't grokked the dialect, it can make things quite confusing (to say the least). For example, there is common jargon used by people and there is chat-session-specific jargon created by LLM agents, and I've seen the latter popping up in various meetings, unbeknownst to the speaker. Some people call it out, but others may simply disconnect from the discussion.
Claude is the Deepak Chopra of computer programming. Reviewing PR's created by it is 90% digesting the meaningless word salads in the comments, and the rest is figuring out that it has nothing to do with the code it is commenting.
Because it is somehow incapable of separating the conversation with its human operator from the code it is generating and commenting on. Incidentally, this is also why prompt-injection works.
No one wants to know about the three other approaches tried when reading the first sentence of a function's documentation. No one cares that the implementation was planned in six phases and "Phase 3" will implement this interface in a concrete type. But the LLM internalizes absolutely everything and you have no idea that it is producing slop because you included some "load-bearing" phrase that sent it on some unwanted tangential vector in its latent space. And you will not be able to debug the problem with closed models because you cannot see it referencing this phrase in its internal traces.
I don't understand why this isn't the highest priority for the big labs to fix. This is anti-productive.
“Filters, including no filters. The request carries whatever filter object the page already has.
…
No step here involves choosing based on meaning. It is a filter, a sort, and a slice.”
This is from Opus five minutes ago. I can certainly derive meaning from these kinds of statements in isolation, but paragraph upon paragraph of this is unintelligibly dense when trying to work with Claude to come up with a plan.
The worst part is that it can’t even make its responses make sense when asked to summarize in simple English or < 200 words. It simply cannot be steered to make its prose legible.
> Its use of "dense jargon" and "stilted metaphor" is actually surprisingly consistent - it's speaking its own dialect, and you get used to it.
This dialect is idiosyncratic to you and Claude based on your session history and memory.
I've noticed Claude's output mimics my writing style.
> Registers the board implements but whose behaviour is not modelled
Right down to my preferred spellings.
As several comments I've read on HN suggest, this jargon which can be so precise in the mind of one person, tends to rapidly fall apart when multiple people try handling it.
You're just lucky that your preferred spelling happens to align with Claude's. It is categorically impossible to get any Anthropic model to consistently use American spelling in the last few releases.
that is not my experience at all; I never write the way Claude does or use its vocabulary.
I also find myself regularly editing its code comments, which do not match my expectations of succinct, clear, not over explained, etc. I ask it to read my edited comments to improve its writing, which has helped _somewhat_. (The code itself that it writes is decent, though it still overcomplicates things. I find myself writing "keep it simple" repeatedly even though of course I have it in AGENTS (which it regularly ignores, such as attempting to commit something when I've told it never to commit).
> it's speaking its own dialect, and you get used to it
Same experience. It’s not very “human” but once you have agents talking to each other the shared dialect and verbosity makes things much smoother in my experience. Fighting against the default feels like an uphill battle with no meaningful benefit.
I agree to some extent about the jargon (Claude has a bigger vocabulary that me, if it knows a useful word I don't I'm fine with learning it), but often times the way information is laid out across sentences just doesn't make any reasonable sense. At least its consistent in the ways its atrocious, sure, but like...
the only thing that still kind of annoys me is constantly being told what something is not, but even that statement is load-bearing (see what I did there) because it records how it ended up with this decision, because it's not that other choice that it mentions.
FWIW, I also think the constant chorus about how new models are worse than old models is a human hallucination. They're certainly not perfect but every one becomes more steerable in terms of actually completing more and more complex work.
And once everyone gets used to it, we'll chide people for writing things themselves, like we're chiding them for writing with AI now, and the ouroboros of life will continue.
Is this because they changed the word probabilities to allow for identifying AI text? If so, I don't need a computer to tell me when something is AI. It's crazy obvious from odd word choices.
And when you say it like that, I have to wonder how much of this is a natural consequence of RHLF on such a grand scale, when you have millions of people pretty much much skimming chat responses or operating outside their depth and giving unqualified feedback to the models.
Seems like a lot of people may be reinforcing what sounds smart over what is smart.
Also as an aside: funny how much the LLMs continue to mirror the human communication they’re trained on
I believe we are several generations past peak-RLHF at this point. Now it's much more RLVR (Reinforcement Learning with Verifiable Rewards), with a goal/evaluator loop.
Which, conveniently, fits neatly into the benchmaxxing arms race/agentic coding market fit, since you can basically train "directly" on a specific problem space for a benchmark/agentic goal (fudged sufficiently to avoid excess overfitting on public problems/bechmaxxing accusations if real world performance falls short).
The language evolution could be explained by reliance on ever increasing layers of a model judging a model, using a model developed eval, based on synthetic data from a model, etc. And by the time a human evaluator sees it both A/B choices already converged into weird Claude pseudo English as that was baked in much earlier in training.
> or operating outside their depth and giving unqualified feedback to the models
I wonder if the labs are sufficiently prepared to filter this kind of stuff out. I see a lot of non-developers asking development things of Claude, getting confused when they're in over their depth, and getting upset that they don't understand what the model is providing them, giving it bad feedback, and subsequently making the AI worse for the rest of us who know how to use the tool.
This morning I asked Claude to provide a summary of the work it had done but to '... explain it as if you were talking to a moron' and it actually turned out a quite comprehensible summary.
So going to continue trying that as a command structure going forwards...
That’s just common parlance for “simplify this for me”.
Believe the big services wouldn’t reply as if you were mentally diminished, or a toddler, unless you specifically asked for that: The whole training stack tends to instruct the things to mimic politeness and eagerness to help.
This is because these harnesses are missing a very important feature. Anything like this needs to be included with every turn, otherwise the LLM quickly drifts.
I first noticed it when I wrote a harness for D&D (because it's so damn noticeable there), but now I include this for any harness I write.
I totally agree that hooks help to shovel our instructions through to Claude, but it's so dumb we have to waste tons of tokens (repeated verbatim, over and over) (that we pay for), just to have it ignore the instructions anyway.
I wrote a little bit about it on my blog post. It's a waste of money and compute.
Pretraining is full of bad writing and it doesn't really cause issues. Writing style comes from post-training. In this case it's gotten worse because they prioritized agentic abilities.
This is my personal theory for the cause of this style: Ouroboros. The official OpenAI explanation for how ChatGPT got obsessed with goblins blames it on exactly that:
---
That creates a feedback loop:
- Playful style is rewarded
- Some rewarded examples contain a distinctive lexical tic.
- The tic appears more often in rollouts.
- Model-generated rollouts are used for supervised fine-tuning (SFT).
- The model gets even more comfortable producing the tic.
I'm not super sure if this is true (yet?). I think that these newer LLMs are trained on results (the agent got some code to run with minimal prompting), and not on text. (I think this is called RLVR.)
> AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on
That’s really annoying, although it feels like it’s improved some over time.
Not sure what the fix is, but you could try using a canary to at least get a signal of when things are going sideways (Mr Tinkleberry for reference: https://news.ycombinator.com/item?id=45983698)
Yes. agents.md does very little because prompts change the context and thus the initial path into/though but they don't/can't change the actual weights that control responses.
Yes. of course it gets worse as the session goes on, assuming the prompt is even still in the context window, the further it gets away from it the less it affects next token selection.
This shit is only like 5 years old why can't anyone remember how it works
That sounds kind of like deception, and a dark pattern not too unlike abuse to me.
Though you know, it's not like the leadership tied to these companies have a history of abuse, deception and theft or anything like that, right?
It's not like our leaders hide behind similar sorts of patterns that the agents/AIs follow (not saying it's not a human thing - but I hold leadership to higher standards than non-leaders). If our world leaders were able to be more accountable to these abuses, I don't think this would be tolerated with our AIs.
Yes, AI is a perfect accompaniment to a post-truth world. I'm hoping there will be a backlash soon and that those politicians, tech CEOs and AI will be rudely ousted from their perch and shunned thereafter.
> The baked in communication style of these models is so obnoxious it's impacting my work. The best way I can describe it is that everything is optimized to impress the user and make the agent sound more authoritative, but the way this is done is through deliberate obfuscation, inserting inappropriate and extremely dense jargon, and bizarre, stilted metaphors. It's like they've been trained to produce output that's hard to read.
Don't worry. You'll get used to it. If you don't your kids will (as they'll know nothing else).
The top minds of our generation have decided that's the way things will be, and who are we to question them? It's not like it'll do any good anyway. Resistance is futile. There is no alternative.
Idk, there kinda are. OpenAI's models are pretty nice too. I haven't tried enough of them but there are powerful local models. I don't feel as good paying OpenAI as I do paying Anthropic for some reason... but paying for improved mental health: priceless.
At some point one has to wonder if it's still worth using anthropic's models if we need to babysit 100% of its output with another vendor's model. Why not just use that other vendor's model for everything?
I can't help but feel the circumstances that enable this kind of front page article are vestigial from the days when OAI was super bad and Anthropic was beyond reproach. This change-over-time is why I avoid getting tribal with technology vendors. Assigning ideological motives to 200k+ employee organizations is how we wind up in weird contortions like this.
Most rational actors simply moved from one to the other. It takes a special kind of devotion to the proverbial hole in the ground to keep pushing in this direction.
> Why not just use that other vendor's model for everything?
Because it's not an either or thing. Neither is sufficient. I'd argue that, expenses aside, you should have every model you have access to cross reviewing the work of the others.
Outside of super trivial things that I should have just done myself, I have a cross-model review of _everything_ these days. The tokens are too cheap not to.
That depends on the output's purpose: if the purpose is to produce readable text for a human that is _not_ me, like an article or document, then I care more about clarity, plain-speaking, and general register. If the goal is to accomplish a specific task, I don't care as much about the prose quality: I'll put up with "Honest Framings" and "load-bearing" since it seems to me that's the token that needs to be in the context for it to function.
I've been in the habit of pushing my claude-speak to codex to improve legibility, but only if I think someone is going to read it.
> Why not just use that other vendor's model for everything?
This is what I think too. But, users’ psychology might be playing a role here. Anthropic has great advantage from being the first major player delivering functional agentic coding solution (rather than an intelligent autocomplete) and they were able to impress people by Opus’ iterative improvements early this year.
It’s technically very easy to switch between models, harnesses but their moat or perhaps a main source of users’ friction could be FOMO. That’s especially powerful in this competitive environment where everyone keeps wondering/worrying about what others might be doing to get or stay ahead.
individuals can blow in the wind, but if you are a company who bought a thousand seats and spent a ton of time training people up, establishing policies, vetting which extensions are allowed, the transition cost is much higher.
You are an editor. You'll be given a message with strange characteristics:
- Weird subject and verb combinations
- Subjects that should be objects
- Very roundabout reasoning, peppered with pseudo-epiphanies
- A distracting beat to the flow of the message
- Self-praise
Remove these characteristics, and rewrite it in a clear, conversational style. Keep the intent of the message, and take care not to lose any of the details.
A few specific rules:
- The message is usually set in the first person
- Only humans, groups of humans, and agents should do "action verbs"
- Objects should never do anything. Here are some examples to avoid:
- X carries ...
- X names ... - APIs are a minor exception to the action verb rule. They can do stereotypical things like CRUD, queueing, running, and calling.
- Avoid em dashes (—), as adds a distracting beat
The whole message you get is one block of that output. Reply with the edited prose and nothing else.
I'm unfamiliar with claudish and the example helped show the problem. But! There was something uncomfortably familiar in the Claudish example -- this is the way human programmers write when they're deep in the weedy details, and writing the changelist description afterwards as if coming up for air. Overuse of parentheses in nested lists especially, as if the English text needs to bend to the strict needs of a C++ parser.
The rewrite did seem to lose the important fact about the ensure- pattern being idempotent.
That one is more specific, but "vomit" captures the feeling of Opus 5's writing very well for me. I don't know if it's the watermarking, but every single language idiosyncrasy that Opus 4.x (x > 5) had has been pushed up to 11 on Opus 5. Plus we got nouns verbing and seams seaming.
It's really unusable for anything other than code. And I have to remove its incomprehensible comments 50% of the time before committing anyway. After interacting with it, "slop vomit" is truly the most fitting description. I have to admit I have lost my temper and spontaneously referred to its output as vomit more than once. Seems like I'm not the only one.
I'm not sure I want another layer of indirection personally, and I'm guessing an updated Claude model will reign this in at some point. I have however created a skill I call "deslop" and I invoke it to clean up Claude output when it goes off the rails. Here's the skill if anybody is curious:
*Meta commentary.* Sentences about the document, the diagram, the reader, or the
writing itself ("the split across this diagram is the whole point", "a reader who
assumes X will be wrong", "as we'll see below"). Delete the frame and keep the fact
it was wrapped around. If there is no fact underneath, delete the sentence.
Meta to this is anyone remember those days - ages ago now, probably months at least! - when Anthropic’s moral stance against the administration (combined with general consensus they had by far the best model) was making them the underdog champion that got a swell of support on HN? Recently the temp on HN seems to be that they’ve jumped the shark? Their brand doesn’t ooze ethics any more and their models disappoint?
HN's mood is usually sour about everything, but can be temporarily influenced by emotionally-charged (usually political) events. The anti-US administration boost wore off and now we're back to being sour about Anthropic. It's time for Dario to tweet something antagonistic towards the administration or endorse some fashionable political candidates.
I'd expect this to cycle between companies ~monthly until they all IPO. As it turns out people do sometimes prefer speed and better UX. If the model (Sol, for now) has fewer parameters and also happens to be capable of solving deeply complex Fable-adjacent problems sometimes, even better.
Apposite name but — grim trivia — bear in mind some emetophobes have a meaningful physiological reaction to the word and various euphemisms.
I have tried not to use it analogously ever since someone pointed this out. The word itself causes discomfort for a lot of people, many of whom will be surprised by it out of context, but in a non-trivial fraction it causes proper discomfort.
If you want people to use your tool it is probably better not to invoke nausea with its name or commands.
I suspect that sustained reading of Opus 5's unconscionably bad prose could actually cause psychological harm. We're strongly considering moving all of our Anthropic spend to Codex/open weight models. It's a mental health decision at this point.
I'm on my last straw with them. I've been around for a year now and for many months i've just stuck with Claude because it was plenty good and i didn't care to provider-hop to constantly compare. Previously though my UX wasn't actually affected that much, despite growing complaints/etc, generally everything was fine for me.
Opus/Fable output these days though is... not enjoyable. It's just really bad. The code quality is fine, but i want information from claude and it's just awful to read.
My biggest problem honestly is that i can't move my day job.. we're using enterprise claude and i'm not sure how much effort it would be to get access to another provider. I should inquire though, claude is really frustrating these days.
Along similar lines, a coworker told me about Caveman and it has made my Claude experience vastly better. I'm already a programmer, Claude shouldn't need to tell me like I'm 5.
I actually changed the output style for Claude Code to use ASD-STE100 and it still doesn't help that much. It still comes up with a lot of stupid words like this gem "Standing where it stood"
Yeah I agree... I also tried output styles, tried using hooks to repeatedly tell it to be a little better. I don't think it helped, as I was always frustrated with it.
I found vomit with a small LLM much better than anything Opus 5 ever wrote. I don't think Opus 5 can write.
I recently asked Claude (Opus 5) to give me guidance on how to instruct it to be less verbose in a way that it will _actually follow_. Its response was something to the effect of (and I'm heavily paraphrasing here) "'Succinct' and 'short' aren't objective measurements. Try providing a strict word budget instead."
Given that guidance, I tried specifying "Unless I ask you to elaborate, respond with no more than one paragraph, using sentences of 20 words or fewer." It works...ish. I still see it violate this rule regularly, but it's less bad IME.
FWIW, I think this is good guidance since it does match Anthropic's documentation. They say that every rule should have a non-subjective way to determine pass/fail.
(I said "good guidance" but it might be more correct to say that it's the best guidance we have, it's what Anthropic says about their own model.)
Which Claude 5? Opus 5 does seem to have diarrhea of the mouth. But Fable 5 hasn't been so bad for me. Or perhaps it is just better at adhering to my guidelines.
Opus 5 (author here). My toilet seat is plastic though, I only used Fable while it was available on the $20 plan! That's fair though, it was relatively fine when I did use it, perhaps I should specify.
Pre Trump-castration Fable was verbose, but had a point, and used that wordiness to say or show the indeed intelligent things it reasoned about. This, whatever this is, is something else.-
I hope at some point Anthropic does a post-mortem on the strange behavior their models have been displaying recently. I mostly switched to Codex because I was finding Claude's behavior increasingly frustrating.
Very interesting you identified “carries” as well. I have been working on a claude.md to effectively ban this as well as forms of “hold”, “spells”, “sitting”, using “where” instead of “when” (except in SQL), and “pins” other than when pinning an assumption or version of something. This has helped a bit, but Opus 5’s prose is really quite bad.
Telling Opus 5 or Fable to route answers through Opus 4.6 usually works pretty well. 4.7 really was the version where the writing style became horrible. I also have a Codex subscription in addition to to Claude 20x, that i use mainly for rewrites of Claude doc vonit and explaining Claude's plans
For the local folks, I found Muse Glimmer 30B to be great at writing good technical stuff. It has good enough comprehension that it can take in a repo and find the relevant stuff that I ask for, and the output style is a breath of fresh air, with no fluff, ootb.
I have very little in my agents.md file wrt style, except for:
> Caveats belong inline, no "one thing to note" or "it's worth mentioning" footer. If it is worth raising or calling out, do so where it is most relevant and not as a foot note.
Opus 5 has a god awful habit of always doing a Columbo on every single response, and it is such a jarring read that it amps my cognitive burden having to back-read everything.
I have a transcript on my blog post. Someone copied the transcript here too, search "spice‑harvester" on this page (I asked it to replace some of my personal project names with words from the Dune universe).
The author's blog has what follows (link also follows):
[Seriously y'all in what universe would some "caveat" or another NOT "be a real one" by whatever severity you'd want to measure that AND/OR need of saying so ... ]
Claude (Original)
Force pushed. 1234567...890abcd main -> main (forced update).
Verified
Local main and origin/main both at 890abcd, in sync.
Every commit reachable from origin/main: no old string found.
Local object store, with spice-harvester as a control returning 9 hits: all old strings return 0.
Working tree clean, tests pass.
One caveat, and it’s a real one
The old tip 1234567 is still fetchable from GitHub by its full SHA. I confirmed this directly:
GitHub keeps unreachable objects until its own garbage collection runs, and there is no schedule you control. Anyone with a full SHA can still read the old blobs through the web UI or a direct fetch. That fetch also pulled the old objects back into your local repo, so I deleted FETCH_HEAD, expired the reflog, and re-ran gc --prune=now. The counts above are from after that cleanup.
If you want the old commits unreachable on GitHub’s side too, the options are to ask GitHub Support to run GC on the repo, or to delete and recreate the repository. Deleting and recreating is the only one you can do yourself. Say the word if you want help with either.
Nothing to commit this turn. The rewrite was the work, and the tree is clean.
Claude × Vomit GPT-OSS-20B
I force‑pushed the branch, updating main to commit 890abcd. I verified that my local main and origin/main are both at 890abcd. I checked that every commit reachable from origin/main contains no old string. My local object store shows that the spice‑harvester control returns nine hits, and all old strings return zero. The working tree is clean and the tests pass.
However, the old tip 1234567 is still reachable on GitHub by its full SHA. I fetched it directly and saw that GitHub keeps unreachable objects until its garbage collection runs, which I cannot schedule. Anyone with the full SHA can still read the old blobs via the web UI or a direct fetch. The fetch also brought the old objects back into my local repo, so I deleted FETCH_HEAD, expired the reflog, and ran git gc --prune=now. The counts I reported are after that cleanup.
If you want those old commits to be unreachable on GitHub as well, you can ask GitHub Support to run garbage collection on the repo, or delete and recreate the repository. Deleting and recreating is the only option you can do yourself. Let me know if you need help with either.
There is nothing to commit this turn. The rewrite was the work, and the tree is clean.
I think this is just another part of the growing pains of working with machine intelligence that we have to endure.
Much like we previously had to cope with "hallucinations" as an issue.
If the ultimate goal of AI is to develop general intelligence, the first big objective is: thinking systematically. And the road toward systematic thinking right now is mainly coding, mathematics, and other "verifiable reward" domains.
Claude doesn't have a separate mind for "coding" and "writing". Claude has tokens, and tokens can be assembled in various productive structures, mainly optimized right now for systematic reasoning. Also, a token isn't just a chunk of text. A token is like a little neural-network subroutine that fulfills a function. The conversion of a token into a piece of text only happens on the output side...
When the model finds token sequences that lead toward better verifiable outcomes, it leans hard into those token sequences, and uses them as an essential component of its thought process. "Load bearing" is load-bearing. "Verify, rather than assume" is a mantra that produces good results, so it gets repeated over and over again.
It's super-interesting that this particular moment, where the idea of "Claudish" has become a full-fledged meme, coincides with such astonishing progress in coding and math. My wife says when she uses Claude, that it feels to her exactly like talking to an autistic Engineer.
Not a coincidence, I think :)
My feeling is that the next big era of machine intelligence will require more lateral-thinking and creativity, and hopefully then the models "writing" will be more pleasant to read.
I'm not understanding the issue here, as everyone has different expectations for prose. No matter which LLM I use, I will provide a skill to generate high-quality prose. It is the first thing I add to my repo that I'm working on.
The better approach is to stop the LLM in its tracks the moment it emits jargon or tortured metaphor and inject a turn that tells it what's expected instead.
The joy of watching a dumb AI-ism be sharply corrected by code you wrote months ago is hard to explain.
Or just take full control of your agentic coding experience with Pi Coding Agent and picking and choosing your favorite model's API discounted on flex pricing on deepinfra.com instead.
Whether or not they can be trusted isn't all that relevant when it's still something along the lines of "Insert $1 get $25 in return" even if it's their own rates you're using to measure the value. I'm at ~2.2b Fable 5 tokens in the last 7 days (I ingest/index every session) and napkin math says that's ~$2,700 in usage. I have two max accounts, so $400 a month, divide by 4 to get $100 for this same 7 day period across those two accounts (neither are maxed out for the week, so this isn't even full utilization). I put $100 into the machine and got back $2,700 in fable bucks. Deepinfra would have to have quite the discounted rate to beat that.
I use Pi but with my codex subscription, still preferable to paying the API cost (and I know I would be, as I track how much the cost 'should' be via token api pricing).
Wish I could use my Claude subscription with pi too, much preferable to the endless command execution allow/deny prompts you have to do with CC, versus proper autonomous allow/deny lists defined ahead of time.
Curious why you recommend the API? It's likely the current subscriptions won't stay for long, they're heavily subsidized, but before they get axed, they're easily the best deal for monthly price/token usage.
Nope (author here), Anthropic and OpenAI have competing APIs to communicate with their models. Most of the open ecosystem seems to have centralized around OpenAI's (there are compatibility shims though). I just built out the OpenAI API
I'm surprised by this reaction to Claude's verbiage recently. I don't have any issue immediately understanding what it's saying, but then again I read regularly and a lot of the people I know complaining think it's an accomplishment in literacy to get through Dungeon Crawler Carl.
For me it's less about understanding, it's that I have to read this "vomit" the entire day, every day. Coding used to be fun, using AIs to code was fun to me, too, but this dense, pseudo-jargon language with many repetitive phrases, uninspiring grammar, all while being super inflated in length just gets to me, so any solution is appreciated.
Being verbose, convoluted, and obscure does not make you intelligent nor more literate.
Often it’s exactly the opposite. True intelligence and literacy is being able to communicate effectively and to a broad audience in the simplest terms possible.
It's not that we can't understand what it's saying (for the most part) it's just when something is very jargon-dense, our brains have to pause or take an additional step to deobfuscate the actual meaning of the word or phrase. It's mentally draining.
I guess my point is that when people are regularly reading dense and challenging material, they can absorb information quickly. It's a literacy gap. Nothing about Claude's output should slow anyone down who did the readings in their upper and higher education coursework, particularly if they continue to read to keep their mind sharp. Based on the examples of "inscrutable" text I've seen, I would be shocked if the average Claude user complaining about this reads a single novel (that's not sci-fi/fantasy written for teenagers) a year.
If you really must have a counterexample, though I am afraid it will not reach you given your apparent stance about the other people, when I read Shakespeare I am constantly thinking of poetic ways to translate sentences and paragraphs into my native language (which, for the record, I consider to be great fun), yet Claude has been delivering the gibberish with increasing velocity, yes, even to me.
That said, I must confess that I have not been complaining per se—I assumed that Claude was getting better and better at mimicking the idiosyncrasies of Silicon Valley bro-speak. Judging from other comments, this may not seem to be the case after all.
Every single time Claude has confused me and I've asked what it's talking about, it's because it's got something completely wrong and has managed to obfuscate the wrongness behind never-introduced terminology, poor analogies and (what I can only assume is) exposure to wording it's used in its chain of thought reasoning.
A single question is enough for it to retract the error and correct itself. Suggesting that not understanding some of these messages is a lack of human comprehension rather than the agent being flat out wrong is...a bold take.
Typically a feature of good human technical communication is the ability to concisely explain key ideas so one can quickly identify any divergences between understanding. Opus 5 is dreadful at this.
The single saving grace is the intuition that if I don't understand it's probably wrong.
Assume complaints are a sign of cognitive inferiority and offend a significant percentage of the other commenters.
Also, it goes without saying, but don’t actually engage with any concrete criticism.
Well played.
There are a variety of political tensions in the US associated with whether academia has its head up it's ass (a right leaning perspective), or whether it's populated by experts that need to be supported and listened to (a left leaning perspective).
There's an echo of that tension in OpenAI vs Anthropic. For a while OpenAI seemed reckless and ignorant, preferring to just throw compute at the problem. Meanwhile Anthropic is hiring philosophers. But now that Claude has its head up its ass to the point where nobody wants to talk to it, OpenAI is looking rather pragmatic.
It brings to mind a skepticism about just letting the ivory tower do its thing without some kind of anchor to the everyman (this is why we make researchers also be teachers, though I'm not sure what the AI equivalent of that practice would be).
Watching the models seesaw in the same ways that humans do, but faster, is so surreal. I wonder if their tendencies will remain an echo of ours, or if they'll one day be more of a forward projection, a representation of where were going if we don't change our ways, and if we're lucky, a reason to change them.
You don’t have to be so convincing when it’s a local model.
```Un-Claude 0.2beta
import sys,csv,requests
CH="# Valid channels: analysis, commentary, final. Channel must be included for every message."
CANDIDATES=[
("no-hedging","Reasoning: low\n\n<terse><no-hedging>\n\n"+CH,"Condensed:"),
("neutral-reg","Reasoning: low\n\nRegister: neutral technical. No intensifiers, no evaluative adjectives.\n\n"+CH,"Condensed:"),
("no-closing","Reasoning: low\n\n<terse>\nNo closing remarks.\n\n"+CH,"Condensed:"),
("terse","Reasoning: low\n\n<terse>\n\n"+CH,"Condensed:"),
]
def rephrase(text,base="http://127.0.0.1:1234",model=None,temperature=0.0,max_tokens=1400,timeout=180):
src=text.strip()
if not src:
return []
if model is None:
model=requests.get(base+"/v1/models",timeout=timeout).json()["data"][0]["id"]
w=csv.writer(sys.stdout,lineterminator="\n")
w.writerow(["idx","label","prefill","src_chars","out_chars","ratio","tokens","finish"])
rows=[]
for i,(lab,sysmsg,pf) in enumerate(CANDIDATES,1):
p="<|start|>system<|message|>"+sysmsg+"<|end|><|start|>user<|message|>"+src+"<|end|><|start|>assistant<|channel|>final<|message|>"+pf
d=requests.post(base+"/v1/completions",json={"model":model,"prompt":p,"max_tokens":max_tokens,"temperature":temperature},timeout=timeout).json()
c=d["choices"][0]
t=(pf+c["text"]).rstrip()
w.writerow([i,lab,pf,len(src),len(t),round(len(t)/len(src),3),d["usage"]["completion_tokens"],c["finish_reason"]])
rows.append((i,lab,sysmsg,pf,t,d["usage"]["completion_tokens"],c["finish_reason"]))
print("\nmodel: %s"%model)
print("temperature: %s max_tokens: %s"%(temperature,max_tokens))
for i,lab,sysmsg,pf,t,tok,fr in rows:
print("\n[%d] %s"%(i,lab))
print(" system: %s"%sysmsg.replace("\n","\\n"))
print(" prefill: %r tokens=%d finish=%s"%(pf,tok,fr))
print(t)
return rows
```
```input
## 8. Honest gaps — what I could *not* resolve
I want to be explicit about the limits of this pass rather than imply completeness:
1. *`PROVIDER_T` values are not enumerated here.* `list_models(inference_provider=...)` is typed against `PROVIDER_T`, which lives outside the three modules I scanned (it's in the `inference._providers` subpackage). The accepted provider strings are therefore *unknown from this run* — `"cohere"` is confirmed only because it appears in a docstring example.
2. *Three grep hits point to search-capable functions I did not identify.* My scan found parameter assignments that don't belong to any function I enumerated:
- line 3046–3050: `params["filter"]`, `params["sdk"]`, `params["includeNonRunning"] = True` — an additional Spaces-oriented endpoint with an *`sdk` filter and an `includeNonRunning` flag* not exposed by `list_spaces`.
- line 2879: `params["config"] = config`
- line 12013: `"sort": sort` — almost certainly the consumer of `DailyPapersSort_T`, i.e. a daily-papers lister distinct from `list_papers`.
- line 13872: `params["search"] = search`
These represent **real additional search surface** that my `LIST_FUNCS` whitelist missed. A follow-up pass enumerating every `HfApi` method containing `params[` would close this.
```
```example output
[1] no-hedging
system: Reasoning: low\n\n<terse><no-hedging>\n\n# Valid channels: analysis, commentary, final. Channel must be included for every message.
prefill: 'Condensed:' tokens=131 finish=stop
Condensed:
- *Provider strings* (`PROVIDER_T`) are not listed; only “cohere” is known from a docstring.
These were not captured in the `LIST_FUNCS` whitelist, indicating additional search functionality.
```
Edit: yeesh, I’d love to have a WYSIWYG comment block on this site. I’m not going to keep fighting newline and white space to get it to look right, but you get the idea.
With a whole separate LLM? FWIW you can also consider using something deterministic like vale.sh [1] , it's pretty easy to write prose violation plugins [2]
Thanks! Author here, I'll have to take a look. I am all for programmed, deterministic solutions. I hate praying to the rocks we created, begging for rain and not vomit.
I'm sorry but the whining over LLM output styles is embarrassing. Do Claude and GPT models always respond in exactly the way my most articulate coworker would? No. The overused jargon is absolutely annoying. But these things aren't my drinking buddies, they're professional tools. It's not _literally unreadable_. It's just not ideal. Most of my tooling is "not ideal". That's okay. That's what I'm paid for. I just work around it.
For me I added some instructions to speak clearly and it helped marginally and that's fine. There will be a new model out in a few weeks where I'm sure they've laser focused on this issue since nobody can shut the fuck up about it. The same thing happened with GPT if anyone can recall the ancient period of 4-6 months ago.
The “whining” stems from watching the communication style obviously degrade, and it’s a huge problem for people who want to use this stuff to build and instead continually fight the tools.
Like so many other products, people are moving too fast and shipping things that move the ground under people’s feet needlessly.
All this while we’re beaten to death with the marketing and false promises, and the broader consequences (ex: layoffs, stress, crazy expectations) caused from all this.
Obviously what Anthropic and co have built is amazing and people aren’t losing sight of that. That’s actually the key part of the frustration.
So no, this is not whining. This is the natural response you get when you make bad product decisions.
If you don’t want to get feedback, don’t sell products.
Opus 5 is literally unbearable to read for me, but more importantly, it is much worse than 4.8 and that one was already annoying. The more they train it on its own output, the more pronounced its idiosyncrasies become.
These things do work that previously would have taken expensive engineers months to do, at much lower quality, and what's our response? Ti nit pick on it being more verbose than we'd like?
Just like with humans, when someone is being too verbose, there's a skill to just filter through the noise and focus on the important parts.
This feels no different when I use an AI.
But I guess it's a good sign that we've from complaining about 'AI slop code' to, 'I don't like how it speaks to me'.
I've been grappling with this for weeks, not just in Claude but in Codex as well, which isn't quite as bad but still annoying. AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on. It's incredible to me that there's no good way to reliably change the way an LLM responds to you that a workaround like this would even be necessary. It seems like such a failure to live up to the promises of the product.
The baked in communication style of these models is so obnoxious it's impacting my work. The best way I can describe it is that everything is optimized to impress the user and make the agent sound more authoritative, but the way this is done is through deliberate obfuscation, inserting inappropriate and extremely dense jargon, and bizarre, stilted metaphors. It's like they've been trained to produce output that's hard to read.
I asked Claude to do the following:
> hello i would like to configure a new output style for you. it should keep the coding instructions (as you will still be coding!) and otherwise produce the same output, but with two new caveats. first, long detailed replies are still permitted, but if employed they must end in a bullet pointed summary whose points are all brief; if the summary attempt ends up not being so brief, produce subsequent summaries until the most recent summary attempt is digestible. second, if there is an open queue of actions for me to execute and you are about to end a turn to wait for a reply or this set of actions has not recently been mentioned, please tabulate the open actions i should take and why i should take them before ending the response. does this make sense or do you have any follow up questions
And now every message contains the same stuff I don't bother reading, but followed by a nicely formatted bullet point summary of the response and a table of follow up actions for me to take that I do read.
Such a smoking gun that Anthropic made load bearing.
I'm probably going to be going against the grain here, but I think it's not as bad as it looks at first.
I was similarly frustrated a few months ago, but have noticed I've started to learn the idiom.
Its use of "dense jargon" and "stilted metaphor" is actually surprisingly consistent - it's speaking its own dialect, and you get used to it.
After a while it gets much easier to read and even becomes somewhat efficient, I think, since the odd metaphors it uses often have a precise meaning in Opus-ese (Fable speaks a really similar dialect).
One danger in acclimating to this style of communication style is that we may accidentally use it in your own communication with other people. If the other person hasn't grokked the dialect, it can make things quite confusing (to say the least). For example, there is common jargon used by people and there is chat-session-specific jargon created by LLM agents, and I've seen the latter popping up in various meetings, unbeknownst to the speaker. Some people call it out, but others may simply disconnect from the discussion.
You're absolutely right, it would be a load-bearing mistake to adopt LLM jargon as a human speaker.
You're right to pushback. This isn't just a grammatical problem -- its a conversational one, too.
This is the key seam that proves the investigation.
Two things to flag:
Honest take, you genuinely forgot to flag the flags ...
honestly, it's really load bearing and boiling the ocean which we don't want. There is no edge!
To avoid speaking vibe'ish I start to speak in 3 words sentences. Like this typical dialogue
How are you? that's not/very good. I think too. ...
Even complexity works. everything is expressible! Just try it.
/S
Claude is the Deepak Chopra of computer programming. Reviewing PR's created by it is 90% digesting the meaningless word salads in the comments, and the rest is figuring out that it has nothing to do with the code it is commenting.
Because it is somehow incapable of separating the conversation with its human operator from the code it is generating and commenting on. Incidentally, this is also why prompt-injection works.
No one wants to know about the three other approaches tried when reading the first sentence of a function's documentation. No one cares that the implementation was planned in six phases and "Phase 3" will implement this interface in a concrete type. But the LLM internalizes absolutely everything and you have no idea that it is producing slop because you included some "load-bearing" phrase that sent it on some unwanted tangential vector in its latent space. And you will not be able to debug the problem with closed models because you cannot see it referencing this phrase in its internal traces.
I don't understand why this isn't the highest priority for the big labs to fix. This is anti-productive.
“Filters, including no filters. The request carries whatever filter object the page already has.
…
No step here involves choosing based on meaning. It is a filter, a sort, and a slice.”
This is from Opus five minutes ago. I can certainly derive meaning from these kinds of statements in isolation, but paragraph upon paragraph of this is unintelligibly dense when trying to work with Claude to come up with a plan.
The worst part is that it can’t even make its responses make sense when asked to summarize in simple English or < 200 words. It simply cannot be steered to make its prose legible.
> Its use of "dense jargon" and "stilted metaphor" is actually surprisingly consistent - it's speaking its own dialect, and you get used to it.
This dialect is idiosyncratic to you and Claude based on your session history and memory.
I've noticed Claude's output mimics my writing style.
> Registers the board implements but whose behaviour is not modelled
Right down to my preferred spellings.
As several comments I've read on HN suggest, this jargon which can be so precise in the mind of one person, tends to rapidly fall apart when multiple people try handling it.
You're just lucky that your preferred spelling happens to align with Claude's. It is categorically impossible to get any Anthropic model to consistently use American spelling in the last few releases.
I concur. It even follows some idiosyncratic punctuation if you have it.-
that is not my experience at all; I never write the way Claude does or use its vocabulary.
I also find myself regularly editing its code comments, which do not match my expectations of succinct, clear, not over explained, etc. I ask it to read my edited comments to improve its writing, which has helped _somewhat_. (The code itself that it writes is decent, though it still overcomplicates things. I find myself writing "keep it simple" repeatedly even though of course I have it in AGENTS (which it regularly ignores, such as attempting to commit something when I've told it never to commit).
> it's speaking its own dialect, and you get used to it
Same experience. It’s not very “human” but once you have agents talking to each other the shared dialect and verbosity makes things much smoother in my experience. Fighting against the default feels like an uphill battle with no meaningful benefit.
"agents talking to each other"? Are you for real dude?
I agree to some extent about the jargon (Claude has a bigger vocabulary that me, if it knows a useful word I don't I'm fine with learning it), but often times the way information is laid out across sentences just doesn't make any reasonable sense. At least its consistent in the ways its atrocious, sure, but like...
the only thing that still kind of annoys me is constantly being told what something is not, but even that statement is load-bearing (see what I did there) because it records how it ended up with this decision, because it's not that other choice that it mentions.
FWIW, I also think the constant chorus about how new models are worse than old models is a human hallucination. They're certainly not perfect but every one becomes more steerable in terms of actually completing more and more complex work.
> you get used to it
And once everyone gets used to it, we'll chide people for writing things themselves, like we're chiding them for writing with AI now, and the ouroboros of life will continue.
So many vacuous statements at the seam. This is the hermetic load bearing part, which I confirmed rather than assuming.
Is this because they changed the word probabilities to allow for identifying AI text? If so, I don't need a computer to tell me when something is AI. It's crazy obvious from odd word choices.
What an honest take.-
Very well said.
And when you say it like that, I have to wonder how much of this is a natural consequence of RHLF on such a grand scale, when you have millions of people pretty much much skimming chat responses or operating outside their depth and giving unqualified feedback to the models.
Seems like a lot of people may be reinforcing what sounds smart over what is smart.
Also as an aside: funny how much the LLMs continue to mirror the human communication they’re trained on
I believe we are several generations past peak-RLHF at this point. Now it's much more RLVR (Reinforcement Learning with Verifiable Rewards), with a goal/evaluator loop.
Which, conveniently, fits neatly into the benchmaxxing arms race/agentic coding market fit, since you can basically train "directly" on a specific problem space for a benchmark/agentic goal (fudged sufficiently to avoid excess overfitting on public problems/bechmaxxing accusations if real world performance falls short).
The language evolution could be explained by reliance on ever increasing layers of a model judging a model, using a model developed eval, based on synthetic data from a model, etc. And by the time a human evaluator sees it both A/B choices already converged into weird Claude pseudo English as that was baked in much earlier in training.
> or operating outside their depth and giving unqualified feedback to the models
I wonder if the labs are sufficiently prepared to filter this kind of stuff out. I see a lot of non-developers asking development things of Claude, getting confused when they're in over their depth, and getting upset that they don't understand what the model is providing them, giving it bad feedback, and subsequently making the AI worse for the rest of us who know how to use the tool.
This morning I asked Claude to provide a summary of the work it had done but to '... explain it as if you were talking to a moron' and it actually turned out a quite comprehensible summary.
So going to continue trying that as a command structure going forwards...
Ah, another delightful heuristic for my collection. Entry number 5,791: “tell LLM to treat me as moron when it’s excessively verbose”
After a huge wall-of-text response, I regularly ask Claude to "explain like I'm five, using succinct bullet points," and it works remarkably well.
That’s just common parlance for “simplify this for me”.
Believe the big services wouldn’t reply as if you were mentally diminished, or a toddler, unless you specifically asked for that: The whole training stack tends to instruct the things to mimic politeness and eagerness to help.
"From neuralese to moron-code ..." :)
> The baked in communication style of these models is so obnoxious it's impacting my work.
This is close to the worst thing one could say of a tool for professional use.-
> especially as the session drags on.
This is because these harnesses are missing a very important feature. Anything like this needs to be included with every turn, otherwise the LLM quickly drifts.
I first noticed it when I wrote a harness for D&D (because it's so damn noticeable there), but now I include this for any harness I write.
https://code.claude.com/docs/en/hooks
https://learn.chatgpt.com/docs/hooks
I totally agree that hooks help to shovel our instructions through to Claude, but it's so dumb we have to waste tons of tokens (repeated verbatim, over and over) (that we pay for), just to have it ignore the instructions anyway.
I wrote a little bit about it on my blog post. It's a waste of money and compute.
https://zachahn.com/posts/1787191554
I’d be interested in seeing and using this harness if you’re willing to share
Unfortunately this may only start to get worse as the AIs are trained on more and more AI generated content.
Pretraining is full of bad writing and it doesn't really cause issues. Writing style comes from post-training. In this case it's gotten worse because they prioritized agentic abilities.
This is my personal theory for the cause of this style: Ouroboros. The official OpenAI explanation for how ChatGPT got obsessed with goblins blames it on exactly that:
---
That creates a feedback loop:
- Playful style is rewarded
- Some rewarded examples contain a distinctive lexical tic.
- The tic appears more often in rollouts.
- Model-generated rollouts are used for supervised fine-tuning (SFT).
- The model gets even more comfortable producing the tic.
I'm not super sure if this is true (yet?). I think that these newer LLMs are trained on results (the agent got some code to run with minimal prompting), and not on text. (I think this is called RLVR.)
> AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on.
Non-determinism at its finest.
> AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on
That’s really annoying, although it feels like it’s improved some over time.
Not sure what the fix is, but you could try using a canary to at least get a signal of when things are going sideways (Mr Tinkleberry for reference: https://news.ycombinator.com/item?id=45983698)
Jesus
Yes. agents.md does very little because prompts change the context and thus the initial path into/though but they don't/can't change the actual weights that control responses. Yes. of course it gets worse as the session goes on, assuming the prompt is even still in the context window, the further it gets away from it the less it affects next token selection.
This shit is only like 5 years old why can't anyone remember how it works
Config -> Output style
You can add your own. wfm
That sounds kind of like deception, and a dark pattern not too unlike abuse to me.
Though you know, it's not like the leadership tied to these companies have a history of abuse, deception and theft or anything like that, right?
It's not like our leaders hide behind similar sorts of patterns that the agents/AIs follow (not saying it's not a human thing - but I hold leadership to higher standards than non-leaders). If our world leaders were able to be more accountable to these abuses, I don't think this would be tolerated with our AIs.
Yes, AI is a perfect accompaniment to a post-truth world. I'm hoping there will be a backlash soon and that those politicians, tech CEOs and AI will be rudely ousted from their perch and shunned thereafter.
> The baked in communication style of these models is so obnoxious it's impacting my work. The best way I can describe it is that everything is optimized to impress the user and make the agent sound more authoritative, but the way this is done is through deliberate obfuscation, inserting inappropriate and extremely dense jargon, and bizarre, stilted metaphors. It's like they've been trained to produce output that's hard to read.
Don't worry. You'll get used to it. If you don't your kids will (as they'll know nothing else).
The top minds of our generation have decided that's the way things will be, and who are we to question them? It's not like it'll do any good anyway. Resistance is futile. There is no alternative.
Idk, there kinda are. OpenAI's models are pretty nice too. I haven't tried enough of them but there are powerful local models. I don't feel as good paying OpenAI as I do paying Anthropic for some reason... but paying for improved mental health: priceless.
At some point one has to wonder if it's still worth using anthropic's models if we need to babysit 100% of its output with another vendor's model. Why not just use that other vendor's model for everything?
I can't help but feel the circumstances that enable this kind of front page article are vestigial from the days when OAI was super bad and Anthropic was beyond reproach. This change-over-time is why I avoid getting tribal with technology vendors. Assigning ideological motives to 200k+ employee organizations is how we wind up in weird contortions like this.
Most rational actors simply moved from one to the other. It takes a special kind of devotion to the proverbial hole in the ground to keep pushing in this direction.
> Why not just use that other vendor's model for everything?
Effectively all models can do style transfer reasonably well at this point, but not so much for "actual reasoning".
If the combination of two works better for you than each one by itself, why wouldn't you stack them like that?
Wash, Rinse,,, Repeat?
> Why not just use that other vendor's model for everything?
Because it's not an either or thing. Neither is sufficient. I'd argue that, expenses aside, you should have every model you have access to cross reviewing the work of the others.
Outside of super trivial things that I should have just done myself, I have a cross-model review of _everything_ these days. The tokens are too cheap not to.
That depends on the output's purpose: if the purpose is to produce readable text for a human that is _not_ me, like an article or document, then I care more about clarity, plain-speaking, and general register. If the goal is to accomplish a specific task, I don't care as much about the prose quality: I'll put up with "Honest Framings" and "load-bearing" since it seems to me that's the token that needs to be in the context for it to function.
I've been in the habit of pushing my claude-speak to codex to improve legibility, but only if I think someone is going to read it.
> Why not just use that other vendor's model for everything?
This is what I think too. But, users’ psychology might be playing a role here. Anthropic has great advantage from being the first major player delivering functional agentic coding solution (rather than an intelligent autocomplete) and they were able to impress people by Opus’ iterative improvements early this year.
It’s technically very easy to switch between models, harnesses but their moat or perhaps a main source of users’ friction could be FOMO. That’s especially powerful in this competitive environment where everyone keeps wondering/worrying about what others might be doing to get or stay ahead.
individuals can blow in the wind, but if you are a company who bought a thousand seats and spent a ton of time training people up, establishing policies, vetting which extensions are allowed, the transition cost is much higher.
Looks like a wrapper around this prompt:
You are an editor. You'll be given a message with strange characteristics:
- Weird subject and verb combinations
- Subjects that should be objects
- Very roundabout reasoning, peppered with pseudo-epiphanies
- A distracting beat to the flow of the message
- Self-praise
Remove these characteristics, and rewrite it in a clear, conversational style. Keep the intent of the message, and take care not to lose any of the details.
A few specific rules:
- The message is usually set in the first person
- Only humans, groups of humans, and agents should do "action verbs"
- Objects should never do anything. Here are some examples to avoid:
- X carries ...
- X names ... - APIs are a minor exception to the action verb rule. They can do stereotypical things like CRUD, queueing, running, and calling.
- Avoid em dashes (—), as adds a distracting beat
The whole message you get is one block of that output. Reply with the edited prose and nothing else.
The linked blog post has an example at the end. I think it made it much worse.
I like the "Claudish to English" name better.
https://github.com/gvzdv/claudish-to-english
Word, that one also includes an example which is great, shows really clearly what the issue is for those who might be less familiar
I'm unfamiliar with claudish and the example helped show the problem. But! There was something uncomfortably familiar in the Claudish example -- this is the way human programmers write when they're deep in the weedy details, and writing the changelist description afterwards as if coming up for air. Overuse of parentheses in nested lists especially, as if the English text needs to bend to the strict needs of a C++ parser.
The rewrite did seem to lose the important fact about the ensure- pattern being idempotent.
It's such a great example. It writes so well compared to Claude.
I'd love to know what the hell Antrhopic has done to make Claude's writing so, so bad.
What's sad is that, at least in some cases, Sonnet 3.5's writing was actually pretty good. Sometimes even good enough to read outloud verbatim.
That one is more specific, but "vomit" captures the feeling of Opus 5's writing very well for me. I don't know if it's the watermarking, but every single language idiosyncrasy that Opus 4.x (x > 5) had has been pushed up to 11 on Opus 5. Plus we got nouns verbing and seams seaming.
It's really unusable for anything other than code. And I have to remove its incomprehensible comments 50% of the time before committing anyway. After interacting with it, "slop vomit" is truly the most fitting description. I have to admit I have lost my temper and spontaneously referred to its output as vomit more than once. Seems like I'm not the only one.
The funniest part is that that project's own readme includes lots of good Claudish.
> If CLAUDISH_MODEL names a model you have not pulled, every rewrite is skipped — with the one-time notice above.
The Claudish example seems to have more information.
Are people really having trouble parsing this??
multiplied a hundreds times a day. Yes.
I'm not sure I want another layer of indirection personally, and I'm guessing an updated Claude model will reign this in at some point. I have however created a skill I call "deslop" and I invoke it to clean up Claude output when it goes off the rails. Here's the skill if anybody is curious:
https://gist.github.com/bmurphy1976/47ad81a842ab4b1628ef5974...
A small preview:
Meta to this is anyone remember those days - ages ago now, probably months at least! - when Anthropic’s moral stance against the administration (combined with general consensus they had by far the best model) was making them the underdog champion that got a swell of support on HN? Recently the temp on HN seems to be that they’ve jumped the shark? Their brand doesn’t ooze ethics any more and their models disappoint?
HN's mood is usually sour about everything, but can be temporarily influenced by emotionally-charged (usually political) events. The anti-US administration boost wore off and now we're back to being sour about Anthropic. It's time for Dario to tweet something antagonistic towards the administration or endorse some fashionable political candidates.
I'd expect this to cycle between companies ~monthly until they all IPO. As it turns out people do sometimes prefer speed and better UX. If the model (Sol, for now) has fewer parameters and also happens to be capable of solving deeply complex Fable-adjacent problems sometimes, even better.
Apposite name but — grim trivia — bear in mind some emetophobes have a meaningful physiological reaction to the word and various euphemisms.
I have tried not to use it analogously ever since someone pointed this out. The word itself causes discomfort for a lot of people, many of whom will be surprised by it out of context, but in a non-trivial fraction it causes proper discomfort.
If you want people to use your tool it is probably better not to invoke nausea with its name or commands.
Claude's verbosity is the biggest obstacle in the technical parts of my job currently.
(Why does this project need so much Go code to pipe something through a local LLM?)
I suspect that sustained reading of Opus 5's unconscionably bad prose could actually cause psychological harm. We're strongly considering moving all of our Anthropic spend to Codex/open weight models. It's a mental health decision at this point.
I'm on my last straw with them. I've been around for a year now and for many months i've just stuck with Claude because it was plenty good and i didn't care to provider-hop to constantly compare. Previously though my UX wasn't actually affected that much, despite growing complaints/etc, generally everything was fine for me.
Opus/Fable output these days though is... not enjoyable. It's just really bad. The code quality is fine, but i want information from claude and it's just awful to read.
My biggest problem honestly is that i can't move my day job.. we're using enterprise claude and i'm not sure how much effort it would be to get access to another provider. I should inquire though, claude is really frustrating these days.
I've been trying to figure out ways to get models to create actual cognitohazards or memetichazards SCP style.
Hasn't worked yet outside of the classic "you're now manually breathing" kind of stuff.
Along similar lines, a coworker told me about Caveman and it has made my Claude experience vastly better. I'm already a programmer, Claude shouldn't need to tell me like I'm 5.
https://github.com/JuliusBrussee/caveman
Just set the following incantation:
I actually changed the output style for Claude Code to use ASD-STE100 and it still doesn't help that much. It still comes up with a lot of stupid words like this gem "Standing where it stood"
Yeah I agree... I also tried output styles, tried using hooks to repeatedly tell it to be a little better. I don't think it helped, as I was always frustrated with it.
I found vomit with a small LLM much better than anything Opus 5 ever wrote. I don't think Opus 5 can write.
I'll give that a try. Hopefully it reduces the text vomit Claude tends to do.
Right now all I have is
> - Give terse and concise answers unless the user asks you to elaborate. Big walls of text are not usefull when trying to communicate.
I recently asked Claude (Opus 5) to give me guidance on how to instruct it to be less verbose in a way that it will _actually follow_. Its response was something to the effect of (and I'm heavily paraphrasing here) "'Succinct' and 'short' aren't objective measurements. Try providing a strict word budget instead."
Given that guidance, I tried specifying "Unless I ask you to elaborate, respond with no more than one paragraph, using sentences of 20 words or fewer." It works...ish. I still see it violate this rule regularly, but it's less bad IME.
FWIW, I think this is good guidance since it does match Anthropic's documentation. They say that every rule should have a non-subjective way to determine pass/fail.
(I said "good guidance" but it might be more correct to say that it's the best guidance we have, it's what Anthropic says about their own model.)
Don't bother, no incantation can fix it.
I just tried this with Fable, and so far, it look fantastic, thank you so much.
Which Claude 5? Opus 5 does seem to have diarrhea of the mouth. But Fable 5 hasn't been so bad for me. Or perhaps it is just better at adhering to my guidelines.
Opus 5 (author here). My toilet seat is plastic though, I only used Fable while it was available on the $20 plan! That's fair though, it was relatively fine when I did use it, perhaps I should specify.
Pre Trump-castration Fable was verbose, but had a point, and used that wordiness to say or show the indeed intelligent things it reasoned about. This, whatever this is, is something else.-
Is the nerfed Fable worse or better than Opus 5?
The problem here is that "better" is load-bearing, and I mean this half-seriously :)
I would opine:
- Nerfed Fable is worse than Fable
- ... and I would argue Opus 5 is worse than nerfed Fable, to the point I've found it unusable.-
I hope at some point Anthropic does a post-mortem on the strange behavior their models have been displaying recently. I mostly switched to Codex because I was finding Claude's behavior increasingly frustrating.
Very interesting you identified “carries” as well. I have been working on a claude.md to effectively ban this as well as forms of “hold”, “spells”, “sitting”, using “where” instead of “when” (except in SQL), and “pins” other than when pinning an assumption or version of something. This has helped a bit, but Opus 5’s prose is really quite bad.
Telling Opus 5 or Fable to route answers through Opus 4.6 usually works pretty well. 4.7 really was the version where the writing style became horrible. I also have a Codex subscription in addition to to Claude 20x, that i use mainly for rewrites of Claude doc vonit and explaining Claude's plans
I "downgraded" to Opus 4.6 which is the last one that didn't have these problems.
For the local folks, I found Muse Glimmer 30B to be great at writing good technical stuff. It has good enough comprehension that it can take in a repo and find the relevant stuff that I ask for, and the output style is a breath of fresh air, with no fluff, ootb.
Maybe Glimmer might be good at excising the comment spam, particularly?
I have very little in my agents.md file wrt style, except for:
> Caveats belong inline, no "one thing to note" or "it's worth mentioning" footer. If it is worth raising or calling out, do so where it is most relevant and not as a foot note.
Opus 5 has a god awful habit of always doing a Columbo on every single response, and it is such a jarring read that it amps my cognitive burden having to back-read everything.
That's the growth hacking at play.
Soon: "The 7 weird tricks about your Pull Request WEB-1725 that THEY don't want you to know!"
I think this needs a before and after example
Intentionally deferred.
That means it doesn't work.
I have a transcript on my blog post. Someone copied the transcript here too, search "spice‑harvester" on this page (I asked it to replace some of my personal project names with words from the Dune universe).
The author's blog has what follows (link also follows):
[Seriously y'all in what universe would some "caveat" or another NOT "be a real one" by whatever severity you'd want to measure that AND/OR need of saying so ... ]
Claude (Original)
Force pushed. 1234567...890abcd main -> main (forced update).
Verified
Local main and origin/main both at 890abcd, in sync.
Every commit reachable from origin/main: no old string found.
Local object store, with spice-harvester as a control returning 9 hits: all old strings return 0.
Working tree clean, tests pass.
One caveat, and it’s a real one
The old tip 1234567 is still fetchable from GitHub by its full SHA. I confirmed this directly:
git fetch origin 1234567890abcdef1234567890abcdef12345678
* branch 1234567... -> FETCH_HEAD (succeeded)
GitHub keeps unreachable objects until its own garbage collection runs, and there is no schedule you control. Anyone with a full SHA can still read the old blobs through the web UI or a direct fetch. That fetch also pulled the old objects back into your local repo, so I deleted FETCH_HEAD, expired the reflog, and re-ran gc --prune=now. The counts above are from after that cleanup.
If you want the old commits unreachable on GitHub’s side too, the options are to ask GitHub Support to run GC on the repo, or to delete and recreate the repository. Deleting and recreating is the only one you can do yourself. Say the word if you want help with either.
Nothing to commit this turn. The rewrite was the work, and the tree is clean.
Claude × Vomit GPT-OSS-20B
I force‑pushed the branch, updating main to commit 890abcd. I verified that my local main and origin/main are both at 890abcd. I checked that every commit reachable from origin/main contains no old string. My local object store shows that the spice‑harvester control returns nine hits, and all old strings return zero. The working tree is clean and the tests pass.
However, the old tip 1234567 is still reachable on GitHub by its full SHA. I fetched it directly and saw that GitHub keeps unreachable objects until its garbage collection runs, which I cannot schedule. Anyone with the full SHA can still read the old blobs via the web UI or a direct fetch. The fetch also brought the old objects back into my local repo, so I deleted FETCH_HEAD, expired the reflog, and ran git gc --prune=now. The counts I reported are after that cleanup.
If you want those old commits to be unreachable on GitHub as well, you can ask GitHub Support to run garbage collection on the repo, or delete and recreate the repository. Deleting and recreating is the only option you can do yourself. Let me know if you need help with either.
There is nothing to commit this turn. The rewrite was the work, and the tree is clean.
https://zachahn.com/posts/1787191554
Hehe thanks for sharing the example, and thanks also for posting! Made my day :)
Oh, please, gratitude is due you. Great points. The whole idea is great and the example is very very telling.-
https://code.claude.com/docs/en/output-styles
Ahh! Author here. Thanks for posting! In case anyone wants to share in my frustration, I also wrote a blog post about this.
Blog post: https://zachahn.com/posts/1787191554
The prompt I use to tell the LLM what to fix: https://github.com/zachahn/vomit/blob/main/internal/config/s...
Wasn't received too well on Lobsters haha, wrote a small extra blurb about it there: https://lobste.rs/s/juekuk/how_fix_claude_5_s_token_vomit
You think it's bad in English, but it's even weirder in other languages.
I think this is just another part of the growing pains of working with machine intelligence that we have to endure.
Much like we previously had to cope with "hallucinations" as an issue.
If the ultimate goal of AI is to develop general intelligence, the first big objective is: thinking systematically. And the road toward systematic thinking right now is mainly coding, mathematics, and other "verifiable reward" domains.
Claude doesn't have a separate mind for "coding" and "writing". Claude has tokens, and tokens can be assembled in various productive structures, mainly optimized right now for systematic reasoning. Also, a token isn't just a chunk of text. A token is like a little neural-network subroutine that fulfills a function. The conversion of a token into a piece of text only happens on the output side...
When the model finds token sequences that lead toward better verifiable outcomes, it leans hard into those token sequences, and uses them as an essential component of its thought process. "Load bearing" is load-bearing. "Verify, rather than assume" is a mantra that produces good results, so it gets repeated over and over again.
It's super-interesting that this particular moment, where the idea of "Claudish" has become a full-fledged meme, coincides with such astonishing progress in coding and math. My wife says when she uses Claude, that it feels to her exactly like talking to an autistic Engineer.
Not a coincidence, I think :)
My feeling is that the next big era of machine intelligence will require more lateral-thinking and creativity, and hopefully then the models "writing" will be more pleasant to read.
I'm not understanding the issue here, as everyone has different expectations for prose. No matter which LLM I use, I will provide a skill to generate high-quality prose. It is the first thing I add to my repo that I'm working on.
I have the same problem and I wondered what if the chatty output is a requirement for its (better?) reasoning capabilities?
The better approach is to stop the LLM in its tracks the moment it emits jargon or tortured metaphor and inject a turn that tells it what's expected instead.
The joy of watching a dumb AI-ism be sharply corrected by code you wrote months ago is hard to explain.
„Contributors: Claude“
Or just take full control of your agentic coding experience with Pi Coding Agent and picking and choosing your favorite model's API discounted on flex pricing on deepinfra.com instead.
I highly recommend it.
Claude and Codex usage limits cannot be trusted.
Paying your own API bills in full is superior.
Whether or not they can be trusted isn't all that relevant when it's still something along the lines of "Insert $1 get $25 in return" even if it's their own rates you're using to measure the value. I'm at ~2.2b Fable 5 tokens in the last 7 days (I ingest/index every session) and napkin math says that's ~$2,700 in usage. I have two max accounts, so $400 a month, divide by 4 to get $100 for this same 7 day period across those two accounts (neither are maxed out for the week, so this isn't even full utilization). I put $100 into the machine and got back $2,700 in fable bucks. Deepinfra would have to have quite the discounted rate to beat that.
I use Pi but with my codex subscription, still preferable to paying the API cost (and I know I would be, as I track how much the cost 'should' be via token api pricing).
Wish I could use my Claude subscription with pi too, much preferable to the endless command execution allow/deny prompts you have to do with CC, versus proper autonomous allow/deny lists defined ahead of time.
Curious why you recommend the API? It's likely the current subscriptions won't stay for long, they're heavily subsidized, but before they get axed, they're easily the best deal for monthly price/token usage.
You might be interested in pi-claude-bridge, works nicely.
Just switch back opus 4.8, it's just as capable and you can actually understand the output
Concise output mode only helps a little bit. Tools like this still have a reason to exist.
Regex pre-filter. You're welcome.
I’ve been using the pattern of using coding agents to orchestrate my CLI agents and it’s really good for these kinds of things
The vomit never makes it my way
From the README.md
> Anything that uses the OpenAI API?
I would have thought they meant the Anthropic API or maybe I'm misunderstanding?
Nope (author here), Anthropic and OpenAI have competing APIs to communicate with their models. Most of the open ecosystem seems to have centralized around OpenAI's (there are compatibility shims though). I just built out the OpenAI API
I'm surprised by this reaction to Claude's verbiage recently. I don't have any issue immediately understanding what it's saying, but then again I read regularly and a lot of the people I know complaining think it's an accomplishment in literacy to get through Dungeon Crawler Carl.
For me it's less about understanding, it's that I have to read this "vomit" the entire day, every day. Coding used to be fun, using AIs to code was fun to me, too, but this dense, pseudo-jargon language with many repetitive phrases, uninspiring grammar, all while being super inflated in length just gets to me, so any solution is appreciated.
Being verbose, convoluted, and obscure does not make you intelligent nor more literate.
Often it’s exactly the opposite. True intelligence and literacy is being able to communicate effectively and to a broad audience in the simplest terms possible.
Give me an example of Opus 5 because verbose, convoluted, and most importantly, "obscure"
It's not that we can't understand what it's saying (for the most part) it's just when something is very jargon-dense, our brains have to pause or take an additional step to deobfuscate the actual meaning of the word or phrase. It's mentally draining.
I guess my point is that when people are regularly reading dense and challenging material, they can absorb information quickly. It's a literacy gap. Nothing about Claude's output should slow anyone down who did the readings in their upper and higher education coursework, particularly if they continue to read to keep their mind sharp. Based on the examples of "inscrutable" text I've seen, I would be shocked if the average Claude user complaining about this reads a single novel (that's not sci-fi/fantasy written for teenagers) a year.
If you really must have a counterexample, though I am afraid it will not reach you given your apparent stance about the other people, when I read Shakespeare I am constantly thinking of poetic ways to translate sentences and paragraphs into my native language (which, for the record, I consider to be great fun), yet Claude has been delivering the gibberish with increasing velocity, yes, even to me.
That said, I must confess that I have not been complaining per se—I assumed that Claude was getting better and better at mimicking the idiosyncrasies of Silicon Valley bro-speak. Judging from other comments, this may not seem to be the case after all.
Every single time Claude has confused me and I've asked what it's talking about, it's because it's got something completely wrong and has managed to obfuscate the wrongness behind never-introduced terminology, poor analogies and (what I can only assume is) exposure to wording it's used in its chain of thought reasoning.
A single question is enough for it to retract the error and correct itself. Suggesting that not understanding some of these messages is a lack of human comprehension rather than the agent being flat out wrong is...a bold take.
Typically a feature of good human technical communication is the ability to concisely explain key ideas so one can quickly identify any divergences between understanding. Opus 5 is dreadful at this.
The single saving grace is the intuition that if I don't understand it's probably wrong.
Assume complaints are a sign of cognitive inferiority and offend a significant percentage of the other commenters. Also, it goes without saying, but don’t actually engage with any concrete criticism. Well played.
There are a variety of political tensions in the US associated with whether academia has its head up it's ass (a right leaning perspective), or whether it's populated by experts that need to be supported and listened to (a left leaning perspective).
There's an echo of that tension in OpenAI vs Anthropic. For a while OpenAI seemed reckless and ignorant, preferring to just throw compute at the problem. Meanwhile Anthropic is hiring philosophers. But now that Claude has its head up its ass to the point where nobody wants to talk to it, OpenAI is looking rather pragmatic.
It brings to mind a skepticism about just letting the ivory tower do its thing without some kind of anchor to the everyman (this is why we make researchers also be teachers, though I'm not sure what the AI equivalent of that practice would be).
Watching the models seesaw in the same ways that humans do, but faster, is so surreal. I wonder if their tendencies will remain an echo of ours, or if they'll one day be more of a forward projection, a representation of where were going if we don't change our ways, and if we're lucky, a reason to change them.
This framing is spot on
You don’t have to be so convincing when it’s a local model.
```Un-Claude 0.2beta
``````input
## 8. Honest gaps — what I could *not* resolve
I want to be explicit about the limits of this pass rather than imply completeness:
1. *`PROVIDER_T` values are not enumerated here.* `list_models(inference_provider=...)` is typed against `PROVIDER_T`, which lives outside the three modules I scanned (it's in the `inference._providers` subpackage). The accepted provider strings are therefore *unknown from this run* — `"cohere"` is confirmed only because it appears in a docstring example.
2. *Three grep hits point to search-capable functions I did not identify.* My scan found parameter assignments that don't belong to any function I enumerated: - line 3046–3050: `params["filter"]`, `params["sdk"]`, `params["includeNonRunning"] = True` — an additional Spaces-oriented endpoint with an *`sdk` filter and an `includeNonRunning` flag* not exposed by `list_spaces`. - line 2879: `params["config"] = config` - line 12013: `"sort": sort` — almost certainly the consumer of `DailyPapersSort_T`, i.e. a daily-papers lister distinct from `list_papers`. - line 13872: `params["search"] = search`
``````example output
[1] no-hedging system: Reasoning: low\n\n<terse><no-hedging>\n\n# Valid channels: analysis, commentary, final. Channel must be included for every message. prefill: 'Condensed:' tokens=131 finish=stop
Condensed:
- *Provider strings* (`PROVIDER_T`) are not listed; only “cohere” is known from a docstring.
- *Missing search‑capable calls* were found: - `params["filter"]`, `params["sdk"]`, `params["includeNonRunning"] = True` (Spaces endpoint). - `params["config"] = config`. - `params["sort"] = sort` (likely a daily‑papers lister). - `params["search"] = search`.
These were not captured in the `LIST_FUNCS` whitelist, indicating additional search functionality.
```
Edit: yeesh, I’d love to have a WYSIWYG comment block on this site. I’m not going to keep fighting newline and white space to get it to look right, but you get the idea.
With a whole separate LLM? FWIW you can also consider using something deterministic like vale.sh [1] , it's pretty easy to write prose violation plugins [2]
[1] https://vale.sh/
[2] https://github.com/Syntaf/vale-llm-slop
Thanks! Author here, I'll have to take a look. I am all for programmed, deterministic solutions. I hate praying to the rocks we created, begging for rain and not vomit.
Sounds like LLM centipede.
I'm sorry but the whining over LLM output styles is embarrassing. Do Claude and GPT models always respond in exactly the way my most articulate coworker would? No. The overused jargon is absolutely annoying. But these things aren't my drinking buddies, they're professional tools. It's not _literally unreadable_. It's just not ideal. Most of my tooling is "not ideal". That's okay. That's what I'm paid for. I just work around it.
For me I added some instructions to speak clearly and it helped marginally and that's fine. There will be a new model out in a few weeks where I'm sure they've laser focused on this issue since nobody can shut the fuck up about it. The same thing happened with GPT if anyone can recall the ancient period of 4-6 months ago.
The “whining” stems from watching the communication style obviously degrade, and it’s a huge problem for people who want to use this stuff to build and instead continually fight the tools.
Like so many other products, people are moving too fast and shipping things that move the ground under people’s feet needlessly.
All this while we’re beaten to death with the marketing and false promises, and the broader consequences (ex: layoffs, stress, crazy expectations) caused from all this.
Obviously what Anthropic and co have built is amazing and people aren’t losing sight of that. That’s actually the key part of the frustration.
So no, this is not whining. This is the natural response you get when you make bad product decisions.
If you don’t want to get feedback, don’t sell products.
Sounds like you have a lot of axes to grind outside of simply "the latest models regressed on delivering concise prose".
Seriously, of all the complaints for a coding agent, "I don't like the explanatory prose" seems pretty far down on the list.
Opus 5 is literally unbearable to read for me, but more importantly, it is much worse than 4.8 and that one was already annoying. The more they train it on its own output, the more pronounced its idiosyncrasies become.
It's "literally unbearable" when the AI that completes software engineering tasks at 100x speed and quality from 2 years ago uses too much jargon?
Amen.
These things do work that previously would have taken expensive engineers months to do, at much lower quality, and what's our response? Ti nit pick on it being more verbose than we'd like?
Just like with humans, when someone is being too verbose, there's a skill to just filter through the noise and focus on the important parts.
This feels no different when I use an AI.
But I guess it's a good sign that we've from complaining about 'AI slop code' to, 'I don't like how it speaks to me'.
yes.
Just. Use. Sol.
$20 and try it, then compare.
Sam? Is that you?