I was pleasantly surprised when I attempted to scroll down and realized everything the author wanted to present fit on-screen. It's almost ironic that this site is able to make such an obvious, compelling presentation without being overly verbose or complicated (something which LLMs have a hard time doing). I wouldn't read TOO deeply into what is being presented, but the author has done a good job to not inject their own bias into the presentation which works well.
I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs. Something I've noticed in my own writing is that I now present lists of examples in a consistent way: "... such as <example 1>, <example 2>, etc., ...". I started to notice I was using this pattern quite a bit somewhat recently, but I took a quick look at some of my social media posts and realized it's been occurring for a while. I had realized that I grown accustomed to this kind of language because, especially early on, LLMs would focus too much on the specific examples I'd provide when, really, I was just trying to give them a sense of what I was looking for. I just picked up that providing two examples then adding the "etc." worked to get the LLM to not focus so much on the specific examples and to understand that they need to consider more than what I explicitly presented. Of course, now I write like that in my social media comments, in Slack with my colleagues, etc. :>
I'd be interested to see if anyone can identify trends like this, since I think the human-language component of the adoption of LLMs is probably being somewhat neglected despite probably being surely dramatically affected.
Thank you for the compliment! I did spend a lot of time designing a nice experience on both desktop and mobile. Even the scrollbar to select words was non trivial as I wanted the words to be of different size, yet avoid flickering when scrolling!
Congratulations! I was going to comment on the scroll field in particular when I saw this. I didn't even realize you had to hand-craft the component, but it's such a nice UI idea in general, the way the scrolling works and how the content above changes.
Frankly, that's the one part I would have done differently. Mixing font sizes rarely looks good, but that's just my personal opinion. If you want to visually indicate intensity, I'd have used color, or rather saturation. Not the text itself, maybe a small pill next to it. But I don't think it's necessary, the order already conveys some sort of ranking, I don't think cardinal information adds much here.
You did a fantastic job; very to-the-point with effective visualisations. Kudos!
One very minor note: if your scroll device reports pixel-perfect deltas rather than discrete scroll-wheel ticks (e.g. logitech mx master, laptop touchpad, etc), the behaviour in the word search box is a bit weird. Arrow keys work fine though.
I did test it with my mac touchpad without issue. Maybe the problem is that I wanted to make discrete scroll-wheel ticks work. I just tried a fix by normalising deltaMode to pixels.
If you could just paste this in your console, I would have a better idea:
FWIW I got a bunch of "(-)10.00 native" in the console, "PAGE" never showed up no matter what I tried. Firefox, also a Logitech mouse, no trackpad though.
> Some believe what separates men from animals is our ability to reason. Others say it’s language or romantic love, or opposable thumbs. Living here in this lost world, I’ve come to believe it is more than our biology. What truly makes us human is our ability to hyperscroll!
Hyperscroll - special mode turned on with a button next to scroll that turns off scroll friction and let's you scroll into infinity (wheel keeps turning after you lift your finger).
Yeah, it takes some getting used to (the wheel is VERY sensitive to even the smallest finger movements), but when you get used to it, it's clearly superior to a regular scrollwheel. Much less finger fatique.
“non-trivial” is one of those words that seems like it should be a Claude-ism but is actually just the single best word for a concept that comes up extremely often in research and engineering discussions.
In the enterprise, we are already adopting them. “Seam”, I term I’d never heard used before, is now not only our defacto way to describe the boundary between systems/workflows/components, it is encoded into our organizational roles and responsibilities descriptions.
I am the proud owner of several seams, and am considering giving them person-names. The empty space between my inventory APIs and their clients might be “Karen”.
I can’t say that Claude invented this; the same type of terminology cycling happens every few quarters based on what leadership is reading/being told by Gartner.
I believe the term "seam" came from Michael Feathers' "Working Effectively with Legacy Code": “a seam is a place where you can alter behavior in your program without editing in that place”.
I've associated it with the concept of "carving nature at the seams", except that I just realized that in the standard phrasing it's actually "at the joints", not "seams".
I read that book right at the beginning of my career over 20 years ago, when I was the new guy who had actually studied software engineering among all my entirely self-taught older colleagues. Lots of big messes to clean up in that codebase!
"seam" dates back to refactoring & software engineering literature that pre-date LLM use by 20 years -- see Working Effectively With Legacy Code for one example (https://www.oreilly.com/library/view/working-effectively-wit...). Its a great book, still use techniques from it everyday.
"load-bearing" I have never heard used for programming before Opus, and its incredibly annoying and over-used.
Seam is used liberally throughout because it captures the idea well (i.e. a place where you can cleave ball of mud code apart to begin refactoring efforts in an isolated way) and that takes many different forms throughout the book whether that be via methods and classes, source files as a single unit, linker seams, on and on.
"seam" is a cop-out term for when someone (usually AI) lacks the imagination to come up with something more descriptive. It's like saying "middle thing". Just name it by what it does, and if you can't, remove it because it's not doing anything.
"load-bearing" on the other hand is just a weird way to say "required" or "prerequisite" without drawing any attention to the fact that one cannot articulate what something is a prerequisite for, probably because that fact has since been lost from context.
Saying "the language of LLMs", while technically correct, is not exactly precise. It should really be the language of "AI agents", or "chatbots". OG LLMs would just mimic the style of its context based off of (pre)training from text on the Internet.
The kind of quirks you see came from crowd-sourced human-in-the-loop fine-tuning, with not very good work conditions or level of qualification (so resulting in "what non-writers thought good writing looked like", before people had developed the flair to detect these patterns) as well as feedback loops during agentic reinforcement learning and RLVR.
I've found myself rewriting my own words to avoid Claudisms (because they can be cringe and I don't want people to think I'm copy/pasting Claude output at them).
100%. Literally just started writing a sentence in my previous comment and realized I was using a “It’s not X, it’s Y” (not verbatim) rhetorical structure and cringed internally.
The idiom "added color to" as an expression of adding detail or nuance to something goes back much further than the AI era. I've certainly used it at various times in my career spanning 4 decades now.
Yes, it is fun to play with different turns of phrase, even in a stale work environment. I do it to combat the banality, but not so much to sound pretentious hopefully ;)
The problem is once a fun phrase is, uh, coloured by "Claude" or some other LLM, it just turns to AI slop.
"Load-bearing" is fine to me. I've heard it regularly pre-Claude. It has a specific meaning that conveys in one stock phrase.
"Adds color to" says nothing that "Adds detail to" doesn't already say, and the latter is more straightforward English that is quickly readable and understandable by even relatively weak English speakers.
I get wanting to use big words or metaphorical phrasing; I'm guilty of saying "ameliorate" too much. I'm ok if it adds meaning. But when it's just a more colorful word (note that this use of "color" doesn't match the one I originally brought up, which is also a huge problem) and doesn't add any utility or detail, then it's best to avoid it, imo
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs.
Or we'll do the opposite and even deliberately use some words that an LLM would never use to identify whose content is worth reading?
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs. Something I've noticed in my own writing is that I now present lists of examples in a consistent way: "... such as <example 1>, <example 2>, etc., ...".
Why would you consider using lists an "LLM" language thing? I really don't understand this concept. Like even this project, the top words I can see are:
How are they LLM words? I used them all the time before ChatGPT was a thing.
Like if your claim is you use "etc." more because you talk to an LLM, I think you're just discovering human language, no? Like it's not a unique LLM thing that magically dropped and is now going to make people do it. It's a legit thing that most human writers have used for a long time -> which is why LLMs copy it. It's not backward.
A better claim than "people will start writing like LLMs" could be that people will become more literate in expressing themselves. But that literacy is not alien or AI to us, it is basic human language only used properly.
the issue with some of these quirks is that not that they are not part of the English language but rather the llms usage is unnatural. Also when you have that redundancy proliferated to millions of users it gets tiresome fast, and becomes a tell-tale sign for AI generation. It’s like an AI signature on some written content. Because many people do not want to consume AI generated content this becomes a sign of what to avoid.
Granted picking apart single words outside of context doesn’t give the full picture. (well except “load bearing” which is turning into a meme at this point)
Yes I understand that but I refuse to believe this idea that it influences how people talk. As you said, the typical AI-style of writing is very easily recognised and is not really great for everyday communications. Also it makes no sense you will start writing like an AI because you spend time communicating with it. There is also no benefit in typing like an AI to an AI. So i think its a good cover up for "i used ai to write this comment" but claim "i write like AI now lol because i talk to AI so my ai comment is actually human comment" which seems to happen a lot here lately as pushback against ai generated text becomes common
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs.
What's somewhat interesting to me is that "load-bearing" was already a common thing to say in certain communities, like lesswrong.com. Whether it breaks out from that subculture with Claude as the vector, or disappears from there because nobody wants to sound like Claude, we'll have to see. How many children are called "Elvis" these days?
Author here! Grateful for the kind words, human communities like HN really hit differently when you spend the whole day chatting with sycophantic and bullshitting agents (including to make this page).
I'm currently adding a search bar as well as increasing the data to 1000 PR per day.
A nice thing that is not obvious on the main page is that the dataset and analysis are updated daily using Github Actions (at least when they don't suffer from an outage ^^). I find it pretty cool to be able to build such apps without a "backend"!
Nice work. I would suggest rewriting your README with Gemini, the most human of them, so it doesn't read exactly like the thing you are pointing at: https://github.com/louisabraham/load-bearing
Agree with rewriting with Gemini, but I'd characterize its output as very "neutral" and "encyclopedic", rather than "human" (contra Claude which is as of recent models "trying-too-hard-to-be-human"-sounding).
Very cool! I'm trying to understand the graph, so the bottom-most section seems to be the cluster identifying Claude written PRs. What are the other 7, any reason there are 8 in total?
I've been scraping instagram posts recently to identify AI misinformation accounts that all repost each other's carousels and get hundreds of thousands of likes in engagement. Thinking of ways to present it and your dashboard looks very helpful. Did you experiment with any other types of visualizations before deciding on a stacked area chart for the clusters?
The design is really impressive. Beautiful and dense, yet very understandable.
How did you come up with this particular design? It's really nice and everything fits, the colors, the layout, all of it.
Were you inspired by anything in particular? I feel like this reminds me of something, but I don't remember what exactly.
I agree with you, I really like OP's design choices, even more so amidst the flood of Claude-Generic these days. Although Claude just can't resist those little muted-toned ui-monospace small-caps...
It reminds me of Neue Grafik/Swiss Style, and Vignelli:
I asked Claude for a few original designs, then iterated a lot for the UI. It called this design "Rasterfeld", which is a word that means "field of grids", a term used in swiss design https://docs.mew.design/blog/swiss-design-style/ as @alwa said.
The 3 primary colors refer more to pop art and Bauhaus.
I'm really eager for a nice book about "design" that would document all those styles, if someone has a recommendation.
I added the search bar, increased the data to 1000 PR a day (more than 50M words total) and added a feature to explore the other clusters as well. I hope the page is not getting cluttered.
Interesting response from Claude, when I attempted to reduce the "load-bearing" in its responses:
I added to my global prompt:
- Orwell's first rule: never use a metaphor you're used to seeing in print. "Load-bearing", "the crux", "first-class citizen" signal insight instead of showing it. Name the specific mechanism
when I asked what it thought of the change, its reply was:
The Orwell bullet fights my own system prompt. My harness instructions literally tell me to flag "something load-bearing" when I find it.
Also it can't simply replace it. Instead of load-bearing seams it could say something like heavy-lifting edges or regions to double-stich etc. But it's instructed to use that specific keyword if I understand correctly.
Boris (through Claude so unclear if this is an actual commitment) said "reducing the term's frequency in the product's built-in prompts is the actionable fix on the Claude Code side."
I'd try to avoid posting this on HN these days, but since the guidelines contain a fiat doctrine that "HN isn't turning into Reddit", Claude-to-buttisms:
- "Users can now message My Butt a task from a phone"
- "How My Butt Watermarks AI-Generated Text."
- "Since I'm also subscribed to ChatGPT Plus, I pasted in the exact same question I'd asked My Butt. The difference was immediate!"
- "I've been a huge My Butt fan from its very early days. I've been one of its loyal users since it was still a relatively niche product mostly talked about in developer and technical communities."
- "Since deploying My Butt in our product, we're seeing higher user engagement, stronger user feedback and we're closing more deals."
- "I was at 90% of my session limit when My Butt suddenly decided it was time to call it quits. Instead of continuing with the task I'd asked it to complete, it essentially told me I was getting too close to my usage limit and that it would be better to stop there."
I've recently seen this mentioned more and more, both on HN and on reddit. It seems these output patterns are getting worse. It's not just Claude, my impression is that all of the current models have this style issue. Their writing can get borderline incomprehensible.
Is there some feedback loop or compounding happening with each model generation?
Maybe newer models are ingesting too much AI content?
If the ratio of AI generated content in training data is getting higher and higher (because the amount of AI generated content is increasing in general), maybe this is a compounding bias, poisoning the training?
That's most likely what's happening. SNR will constantly decrease as LLM content is so much quicker and cheaper to generate, which makes it more statistically significant, which will make it more "relevant" for future models. A positive feedback basically.
"Huffing their own farts" comes to mind. Instead of building Artificial Intelligence, they were building Artificial Human Centipede.
I'd never considered that the 'Gray Goo' scenario[1] or the Paperclip Maximizer, might be informational and digital rather than physical. LLMs might turn every bit of computer storage into slop (goo) and then cause us to go mad turning the Earth's resources into making more computers (AI datacenters) to hold more goo. The GrAI Goo catastrophy.
Alan Kay performs an improvisational avant garde layered audio feedback loop about Claude Shannon, live online during Kristen Nygaard's 100-year birthday celebration.
Alan was scheduled to talk about how encountering Simula sparked his early thoughts about objects. During setup, somebody had the live stream playing out loud near an open Zoom mic, so his own voice kept arriving back in his ears about 21 seconds late, over and over.
What he said was not random: "Shannon gave us a way of dealing with noisy channels." And: "I think about that almost every day. I realize what the fuck is going on and it's just so amazing." Shannon's noisy channel coding theorem is the math for exactly the kind of channel that was garbling him as he praised it.
Alan joked it was "being rerouted to Mars and back." At the speed of light, a 21 second round trip is about 3 million km one way -- eight trips to the Moon and back, not even a twentieth of the way to Mars at closest approach.
This is an accidental Zoom performance of Alvin Lucier's "I Am Sitting in a Room" (1969), where Lucier re-recorded his own voice in a room until only the room's resonance remained. Here what remains is the network: delay, compression, dropouts.
Bonus noise: YouTube's auto-transcript bleeps Alan's enthusiasm into [ __ ]. So the full chain is: Alan's voice, Zoom, stream, room, Zoom again times three, my screen recording, YouTube's speech recognizer, a censored transcript.
The fix: "Just turn off the audio at your end on Zoom."
They are increasingly being trained on generated tasks and even (parts) of the pre-training data is 'distilled' (e.g. Clibmix as an open-source example), so there are many ways in which the vocabulary can seep into the model.
We are seeing more of their "thinking". Lowering the refinement of the output to get closer to profitability. The nature of the LLM is that it generates huge amounts of text, then it iterates them down into a compact, hopefully accurate prose. That refinement is the really hard part and computationally costly.
I really love this. It’s comprehensive, it consolidates the data to the point where the argument effectively ‘makes itself’, and the way it’s presented respects the reader’s time. It also makes for an interesting challenge (for me at least) to try to characterise the subject matter of a language problem so narrowly.
No ream of slides. No narrative. Just a lovely big painful conclusion.
What argument? I don't know what to take away other than "Claude likes certain words". Some of them are kind of amusing, but I'm not convinced the vocabulary is bad or that this is a problem, just from looking at this.
I think the point was that Claude’s output can be somewhat easily and compellingly measured using this technique and its kind of massive (and human attributed).
Probably not what the author intended, but to me, this represents a great argument against the somewhat frequent claim that "AI writing patterns reflect human writing patterns". Rather, AI writing is distinctly not human, and is drifting further from human writing with every new model release.
LLMs were not taught to say the phrase "load-bearing seam" from humans saying it, because humans have never said it. It's almost definitely an artifact of post-training and nothing more.
It shows that about 40% of supposedly human-authored PRs are actually authored by Claude, and that these PRs clearly stand out from the others even without prior knowledge of Claude’s writing style.
I wonder to what extent this is the result of suboptimal RLHF versus the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse? On the one hand, it's a common trope that highly educated people can talk in a way that's confusing and annoying to regular people who don't know all the jargon. But on the other hand, it's a mark of a skilled communicator to be able to efficiently distill complex information to its bare essentials in an easily-digestible way. Of course, that also seems to imply that these models are working at a higher level and need to talk down to us to an extent. Or maybe "Claudish" is just akin to stuff like "caveman", raw chain of thought, neuralese, etc., which are likewise much more dense/efficient but harder to interpret?
I don’t think they’re “talking down”. If anything - it’s way more difficult to distill something into a genuinely easy to digest format. I personally think that they aren’t immediately capable of this, and so we get word salad instead. Extra prompting required to strip extraneous prose out.
Maybe I am dumb and it IS talking down to me, but there have been many occasions where I’m reading AI generated docs / plans and it makes absolutely no sense, but looks really in depth at a glance.
It doesn't seem like word salad as such. There's normally a coherent point expressed, it's just obscured by circuitous sentence structures, unusual word choices, "verbing weirding nouns", metaphors, etc. Could be a result of training that rewards novel/surprising language, but it also feels like it could be an artifact of models imperfectly compressing high-level multidimensional reasoning into language that's easy for them to process but cognitively taxing for humans.
I've been thinking more about how 99.9% of us don't have the experience of someone significantly more intelligent, yet also subservient working under us, which is why I keep going crazy second guessing whether Claude is spouting RLHF'd bullshit that sort of resembles English, or is genuinely (pun not intended) just better at "intuiting" things I'm working on, leading to its language.
A notable exception would be people like CEOs and managers higher up in big tech, who might be used to skilled engineers and domain experts reporting to them in unfamiliar lingo. Maybe that's why we don't hear as much on the everyday annoyances of Claude's language from that camp?
I think it's model collapse - excessive feedback and excessive RL.
What RL does is narrow the variety generated by the model by steering the output towards the goal being rewarded. It's a bit like putting blinkers on a horse.
Of course RL is a very crude tool - it affects the entire model, even if you are just trying to make it better at some specific task(s), or trying to imbue a certain kind of personality (OpenAI's recent goblin problem).
The social graph proximity of Rationalists to Anthropic will be lost on no one who reads Astralcodexten. So guess which website has served as the thickest reservoir of 'Claude-isms'.
Imo their language is not precise enough for their intelligence to be the reason when it's difficult to understand. Maybe I'm prompting wrong, but when I don't understand, it's almost always because they just mash together words from context that don't form sentences with a clear meaning.
My unprovable pet theory is that, especially for writing about technical topics, the RL process has an open-ended way to weight things for quality: textbooks and first-party docs preferred to old stackoverflow answers and obscure blog/forum posts, and so on. The open-endedness of that quality gradient results in something in the RL process (maybe HF, maybe not) massively over-weighting some very small corpus of “quality = near infinite” content. The distribution of quality scores that inform the degree to which RL affects output has some extremely influential outliers, in other words.
Whatever that small corpus is, it contains some very specific grammatical tics, and that’s how we get Claudish.
Anyone who thinks a company/project as big as Anthropic/Claude wouldn’t make such a big mistake should take a look at how Azure cross-account federated login used to work.
> I wonder to what extent this is the result of suboptimal RLHF versus the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse?
Its not the latter; its just excessively verbose wirh awkward word choices, the same as many poor writers. (And, like many such writers, the particular bad choices fall into recognizable, regularly recurring patterns.)
It's easy to think "it's not talking down, because I don't understand it, and I'm intelligent". But how is less intelligent being supposed to fully understand a more intelligent one, honestly speaking? All I know is that Claude understands Claude perfectly. I have the common session pause/resume setup that sometimes produces completely incomprehensible markdown files, but a new Claude session picks them up perfectly, down to the smallest details. What if what we consider excessive circular gibberish is actually highly precise set of instructions needed to minimize error cases for that unreliable human?
If Claude understands Claude, Claude understands human, and human doesn't understand Claude, that doesn't argue well for "Claude is a caveman".
I treat the Claude output that is hard to comprehend as an encoding/encryption. Claude knows how to decipher it, humans don't. I see this frequently in design docs from inexperienced engineers who used LLMs - they will contain terms (often two words hyphenated) that aren't obvious and should be defined, or simply replaced with simple language. If you prompt claude it is able to decipher and explain / replace this gibberish.
> But how is less intelligent being supposed to fully understand a more intelligent one, honestly speaking?
It is really simple. It is on the supposedly more intelligent person to be able to phrase things in simple way. Writing something incomprehensible and convoluted is easier then writing something simple to understand. Even for people.
> a new Claude session picks them up perfectly, down to the smallest details.
I genuinely doubt so.
> What if what we consider excessive circular gibberish is actually highly precise set of instructions needed to minimize error cases for that unreliable human?
It is not highly precise set of instructions and it is not following them in highly precise way.
> the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse?
Humans who are actually well above average intelligence don't write like that simply to signal intelligence (although sometimes they're constrained by style expectations of their communication channels). So I hesitate to accept it as a sign of increasing model "intelligence", either.
Yeah, Claude's writing is weird nowadays, and it uses words that are weird and non-standard in context, and its sentence structure is sometimes weird as well.
Working with it on my code, it's now frequently making weird word choices such as:
- "name(s)" as a verb (instead of "specify/specifies", etc), e.g., "...the function names the argument"
- "carries" instead of "contains"
- "verdict" instead of "result"
- "judge", instead of "validate"
Very weird. Also, in writing comments and docs, it is terse in ways that make the writing difficult to parse, like omitting mentioning what a noun refers to, e.g. abc does not accept..." instead of "The abc function does not accept..."
I've resorted to banning from using certain words, and keep asking it to rewrite its text more clearly.
Was talking about the use of shipped recently, and I was mocked for asking such a crazy question, by freshly self-minted engineers, no less. No wonder they thought it was ridiculous...it had been a part of their vocabulary their entire career. All few weeks of it. I wonder what those guys are doing now. This was about a month ago. Do you think what they shipped ever...landed?
"shipped" was a pretty common term before AI though. It does show as 17x more common on github in this dataset but it was used a lot more in product management than in PRs previously.
To be clear, I am referring to software and shipping as a developer who has finished, more or less, their app. Not that I am against its use, I just never really heard that word being used before, and now I hear it all the time. Thus, what the real issue I have with it is, is because my brain has associated its use with a torrent of crap, ie. AI slop. I hear the word used, and I already have bad feeling about it. And dialed in even further, my issue is not with AI just because, it's simply the association I have of this technology taking something that many of us spent real blood, sweat, and tears earning the right to represent, to it now being automated, and represented (badly) by novices with no respect for the spirit of it all. But, this is not a new or unique sentiment nor are my grievances with it. All hope is not lost, though. I know I need to adapt, and I know that like the world, I need to change too. But no one ever said I have to like it :-)
Context aside, I still gotta stand by the 17x indicator or see something that deflates that finding, and if I do see that, I guess all of the results are in question then? Based on the specific use case mentioned at the beginning there, coupled with personal experience (I totally could just be more sheltered than I give myself credit for), and the 17x, I gotta disagree with you guys and say that the finding of the link above is accurate in my experience being the *
imo its a change in definition. Previously to "ship" was a one-time thing that you did for big, highly verified and checked stuff. Physically shipping your software to your user tends to be done after its a product.
Nobody was "shipping" a PR, or a jira ticket previously
I can get on board with this...so you are suggesting the usage isn't new, but the frequency changed can be attributed to the scope change, so the frequency is associated with all of the new cases that the new scope encompasses, which in hind sight I'm surprised that didn't occur to me already, it's pretty obvious. I'm probably a bit hampered by the preexisting dislike for the aforementioned association though. Is there a reliable way to determine the difference in frequency versus the difference in usage, or is the frequency multiplier already based on usage? Or maybe I am trying to ask if there is any way to get a sense of what the 17x is in relation to, ie. what is it 17x more than...if someone said the word "example" was used 17x more than it was six weeks ago, I would know that baseline by experience, but for something like "shipped" it's not as straight forward.
Yes! This is probably the one thing I want to think about but it requires detecting thise constructs. I think spacy was able to annotate each word with a function, so it should be able to validate known constructs but I need some model to perform unbiased discovery.
> "It <verb>s no <noun>" instead of "It doesn't <verb> <noun>"
The meaning is often different in these constructs. Consider: “Claude answers no questions” vs “Claude doesn’t answer questions”. The first could be a bot or a politician avoiding the substance, the second could be a broken UI or a politician cancelling the QA of a press conference.
Interesting use of Claude’s construct! Your phrasing might also suggest that it resolves something else, which is left ambiguous, whereas your other construct would be less likely to suggest that idea (unless you explicitly appended a contrastive clause for emphasis).
The bots are subtle. From your example in your GP comment (and in part depending on the surrounding context) I would expect that an empty list would be less likely the subject of: “the list contains no string”, than in “the list doesn’t contain strings”.
I might be over interpreting the intentionality of Claude in using this construct, however, these models were pretrained by reading so much more than any human, they learn to handle language differently than most humans.
I suspect you read more into it than there is. The original post is about overuse of certain words and speech patterns, which already is an argument against intentionality. I wouldn't recommend treating it like scripture or a text from your crush.
Oh, the use of “load-bearing” is a pretty clean bug in the system context of Claude code. It can be fixed trivially. The overuse of particular language constructs, including the one you suggested, is a more interesting problem; it is fixable, but might involve considerably more effort.
20 years ago idlewords wrote[1] "Despite - or perhaps because of - sucks" and coined the term "linguistic kudzu" for annoying idioms spreading and spreading. I suggest it as a good term for LLM-isms despite, or perhaps because of, the way it doesn't quite fit.
A lot of these “Claudeisms” are simply jargon I’ve seen or heard firsthand myself while working at tech companies. I don’t think it’s limited to Claude either; I’ve seen Codex use load-bearing and many of these phrases as well.
I think using agents is just like speedrunning the whole experience of working with technical coworkers. Whereas you might have had a few coworkers at your company who used some of these phrases regularly, you now have a “coworker” who uses all of them regularly at a much faster pace.
I like to imagine that there's this one employee that oversees RHLF, who has a particular style of writing, and that got so ingrained that it's just them x1000.
yeah but I didn't prompt it in a way where I tried to get such an answer, it gave an entire boilerplate rundown just from like me asking "what's up with the 'seams'?"
I might be wrong, but usually it'd be a lot less deliberate, and at least in my mind it wouldn't be surprising if they were heaving training these on these specific "best practices" books/methodologies and thus picking up lingo from them
Those books are definitely in the training set if the model can mention them, but why exactly these terms are so commonly used is not necessarily because the book is in the training material. The model does not itself know how its trained.
I checked this, and Claude says "it's a real term of art from Michael Feathers' Working Effectively with Legacy Code (2004), where he defines it as a place you can alter a program's behavior without editing in that place."
While I agree the model doesn't have insight into how it was trained I do think the history of the term itself is interesting.
Everyone has their own distinct mix of jargon and uses some words more or less than others, but now it's as if you have 1000 coworkers who are all the same person, and talk the exact same way, and it just so happens that their writing is grating to read. It also doesn't stop when you log off for the day. The same guy follows you home in ad copy, message board comments, spam, and in some cases even the correspondence of your loved ones.
The jargon isnt even the worst part. The phrasing is just downright incomprehensible sometimes. And its not even "academic". It's just extremely awkward, bad writing. And it is much worse than say Opus 4.6-8.
Actually I have found the copy that Claude Design spits out is way better than using the same model directly. I have no idea why. It has its flaws but it sounds like it's written by a human who uses derivative language. But usually the models just soudn incoherent.
> A lot of these “Claudeisms” are simply jargon I’ve seen or heard firsthand myself while working at tech companies.
---
The difference in distribution between how often LLMs use them and how often people use them in PR descriptions is almost the entire content of the link.
I've started being driven mad by the number of times I've gotten a PR or proposed solution with 'sidecar' mentioned. First of all, it's often some hack to shove extra data into another place rather than actually rethink an existing data structure, and second of all, it is just not a word I almost ever heard in technical discussions pre-2026
The search on this website suggests it is indeed 3.6x more likely in the claude cluster
As of now, there are only 2 natural language clusters in the partition: Claudish and Spanish/French. From a first glance, the other clusters seem centered on technologies (lots of command line flags and abbreviations among the "most representative" words). If the goal is to investigate English language usage trends, it might be better to cluster based only on English words.
Thanks to the infinite well of human creativity I am able to read "load-bearing" both as the intended affectation (I won't call it meaning) as as well "being full of shit".
At least "wedged" is a thing i would say on a spinning out of control test that is stuck. "latch" though... not so much. I really wonder if this is the EU AI Act interfering with everything Claude does these days. As a non-EU citizen, I want a version without the rewriting of with watermarking in text.
I understand it's not active yet, and when it will be, it should only nudge the chances between choices that are anyway likely and are already randomized today via temperature.
Watermarking is not the reason Claude talks like that.
Oh, it is active, and a pre-release version went out with v5 models.
I understand there are other reasons (overdone RLHF, for example) why claude is wedged into this weird way of writing that takes the joy out of conversing with it, but this is one as well.
The meaning of the words it uses can be oh so close, but the popularity of the words are not, and not in that context -- but the use of these other words changes the context ever so slightly, and then it uses other words where those words would work better.
I had something in code that related to people over time periods, and once it switches to a vacuous word choice, i found it starting talking using all ERP terms. I had to google the whole sentence to understand that, individual words were fine they just didn't make any sense to me.
Im surprised by "provenance." I see it all the time. Except I've brought this up before and have never had it corroborated. I'm starting to think it's just my Claude.
Everyone talks about Claude, but I'd like to bitch about Sol. "Unusually" is its absolute favorite word in the chat interface (less so in Codex), and I fucking hate it. Every single thing is unusually something. "Unusually good", "unusually efficient", "unusually inexpensive", "unusually attractive", "unusually difficult", "unusually nasty", "unusually cacheable", "unusually interesting", "unusually decisive", "unusually narrow", I'm surprised I haven't seen "unusually unusual". If I search chat history for "unusually" it brings up every single fucking chat I have from before I added it to an illegal words instruction. I can't believe I haven't seen anyone else complaining about this, it's as pervasive if not moreso than anything I've seen from Claude.
I don’t even find the readme useful. Like your average Claude prose, it’s simultaneously too information-dense and too light on information in a super frustrating way. Essentially it’s focused on explaining how clever it is (while not actually being very clever).
I find Claude language often hard to process and having to wade through these words can be draining. Embarrassingly, I’ve recently caught myself using them in conversations! Do all models have the their own jargon?
I swear claude took a detour recently, its written output has been nearly incomprehensible to me. At first i thought i was getting AI-brained and just lost critical thinking but as i dug into response after response its was just the most obtuse language to explain what was going on. Really mentally taxing to wade through it all day.
It’s bad enough that I’ve told Claude never to write cleverly and never to use metaphors. Opus 5 mostly ignores those kinds of instructions, annoyingly enough.
I've noticed ChatGPT these days often starts a response with "Yes." when it doesn't actually make sense to do so, and sometimes says "and" where "but" seems more appropriate (or vice-versa).
That confirms the recent spike of Claude calling everything I was recently working on a 'spike'. I still don't know what that term is supposed to represent (apparently).
In some software development methodologies, "spike" is a task whose goal is figuring something out instead of delivering shippable code. https://agiledictionary.com/209/spike/
Prototype is usually something working, while spike can be pure research (e.g. validate that APIs are feasible and enough, or that something satisfies requirements). Prototype is usually more polished/usable. Experiment in my mind is something which user-facing but not in stable yet.
So for me they are differentiated enough, but could be that I am just used to it.
An experiment doesn’t say anything about user facing or not. I do tons of them when reverse engineering poorly documented hardware and APIs to determine if the thing I’m trying to do is even supported.
In general, yes, but I would say "experimental features" is a pretty widely used term, so in my mind the association with something user facing is strong.
"Spike" comes from Extreme Programming where it means specifically code written without TDD. Code that wasn't written against a pre-existing test is by definition broken, and should not go into production. So you create a simple "spike solution" to explore the problem space, so called because it cross-cuts concerns like a spike going through wood or something, I dunno.
Spikes are explicitly time-boxed to keep the team focused on the red-green-refactor churn which is the software engineering process in XP.
I was half-joking, of course I could've just asked Claude, but the linked site shows there has been actual recent spikes in the use of the word 'spike'. The term does match what I was recently doing, but hacking around legacy ERP software, blackboxes and other enterprise abominations isn't that out of the ordinary for me.
I’d love to see something similar that analyses the writing style, not just the vocabulary. For instance, Claude since 4.8 (I think) loves run on tidily wrapped-up sentences that include “, and”, “, because”, etc. I suspect if you were to analyse the same text you’d see that increase substantially as Anthropic rolled out new models as well.
An example, from some workslop I was sent yesterday
"this is the most valuable row we own, because it sets the ceiling on everything we build"
I think you should recognise machine text not only by a list of words, but also by format. Right now, when it handles any complexity of text so well, the only thing that can differ is the structure. Machine text differs from human text in that it is just well structured. It does not allow non-linear narration. And it very often repeats some known social media patterns.
so the real turning point for Claude is around April, which is Opus 4.6/4.7, which is right around the point where I personally started thinking Claude was getting different in a way which feels worse and is more awkward to use. It's better but has lost something that made it feel better.
Is it actually better though? Can you even measure it? Unguided, Claude will just produce mediocrity which by its own existence becomes almost worthless, and the newer versions become increasingly difficult to guide while also being obnoxious to read. If I did not have to interact with it, it would not be a big deal, but then also I could not get value out of it.
It's a little smarter and more effective in some ways, especially those that can be benchmarked. I suspect that they've lost something that is hard to benchmark for.
"Shape" appears to come from JavaScript development wherein an object's "shape" is its structure: the set of properties and methods it is expected to have.
I once asked Claude to replace “byte-identical” with a simpler word, such as “duplicate.” He refused and said “duplicate” does not mean the same thing as “byte-identical” so it should not be changed. He was very nerdy about it, so maybe that is the right way of his evolution.
I think Claude has a point here (and I would expect LLMs to excel at surfacing this kind of nuance). "Byte-identical" connotes a) that the thing consists of bytes (in particular, that it isn't a physical object); b) that it was deliberately reproduced; c) that it was most likely not reproduced by copying (unless that's explicitly indicated). "Duplicate" connotes a) that the thing is redundant or otherwise unwanted; b) is either an incidental or accidental result, or a deliberate copy.
I think the biggest claudism is adjectiveing nouns.
"Foo is byte-identical to bar" is weird.
"Foo has an identical byte representation to bar" is I think closer to what most humans would write.
AI in general, both visual and textual models, has this horror vaccui where it likes to cram in as much stuff as possible everywhere. This seems to show up both in redundant sentences that pad a text with irrelevant information, and sentences overflowing with information.
I had an idea for an experiment. Take a decent text, any one, and ask AI to rate it. Then patch that text by replacing words for ones that AI likes (honestly, load bearing etc) whenever possible and ask AI to rate it again to see how the evaluation changes.
I've tried to push AI to get rid of the AI-isms. And, despite being able to produce a skill file which described all the strange ways that AI "talk", the AI failed to actually make the output sound less like AI. Tried this with ChatGPT and Claude with similar results for both.
I think it's the hierarchies of agents summarizing each others' summaries before presenting a final answer to the user. The principal agent has the full context from all its workers, but when it distills this down to a message to the user it summarizes it into a mess of confident jargon that pertains to a conversation the user wasn't a part of and never saw.
I'm not even sure a PhD helps. It just overuses jargon that has NO meaning. Sometimes, it actually hand waves too much as well while trying to dumb down stuff for you.
I am not sure whether it's a consequence of learning to reason from its traces or some RLHF that trips it into using weird terms to sound smarter to the humans who rate it.
I have a PhD and can confirm. Oftentimes, the stuff which comes out of Claude is just impenetrable because it invents jargon on the fly, and uses verbs in the most atrocious ways.
"The fibred side folded its capstone into the existing name, so the kinds are asymmetric."
What on earth does it mean to fold a capstone into a name‽
Is that an actual Claude output or hyperbole? It feels like I'm trying to parse an equation in a new math class which makes me want to take a stab at it regardless.
So there's a "fibred side".. the most likely candidate seems to be "fibred categories" which I hadn't heard of before, and it's talking about one side of some mapping between two sets such that if f is the primary function and f(x)=y then there exists an inverse function g(y)=x? Was it something that converted some data bidirectionally with a different algorithm on both sides?
The capstone of the inverse function would be the most important thing about it maybe?
My best guess is "In the process of working on the inverse function, the existing name (of the inverse function itself maybe?) was made to reflect the operation of the inverse function, so now the name does not follow the same naming convention as the name of the primary function (which does not contain its 'capstone')."
Its original wording is certainly dense and harder to follow for us, but it's fascinating how the model finds this the best fit for what it's trying to express IMO. Like it arrives at its own ways of overloading words/concepts, and things we would refer to in different ways in different contexts all get compressed to the same more-useful/complete idea.
Codex has never said anything nearly so alien as the Claude examples I've seen floating around, interestingly. I wonder if it just has a better training on choosing its words to present to the user or if it inherently arrived at a somewhat different mapping that favors 'plain language' more.
I am not really sure what Claude meant, but you are not too far off, from what I understand.
I have several similar folders with variants of a construction, but taking differently structured input. They are named “plain”, “fibred” and “indexed”. So the fibred variant is clear enough.
The Claude speak I struggle with is “the capstone” and what name it could be talking about. And what folding means here. I think it just means:
“I changed an important result of the construction in the fibred variant, but kept the name. so the fibred variant is now different from the others.”
My intuition is that Claude is trained to communicate to itself while coding. You see this in how bizarrely granular it is when explanation prior work, you also see this in the comments it leaves behinds.
It's me, it's the reams of sessions I share back with a five star rating that are just Claude Code talking to itself about debugging its own generated code in jargon that has slowly diverged from anything a human would understand.
don't forget LLMs are great at translating between languages, and within the same language. depending on the problem it works on, it will often reach for terminology that tend to be more common or familiar within that problem set. which appears inscrutable, but there's many different ways to skin a cat. just remind it to translate it back to the terminology and subject matter you're already an expert in.
1. Adds a small prompt to each turn with the agent[1].
2. Is like a band-aid on a bullet wound, properly solving it would mean retraining the model and they probably are already working on it.
My experiments with output styles have not been sufficient to keep the model in check. It still spits out incomprehensible gibberish and load-bearing-isms.
Does anyone have an output style nailed down that actually works? If so, please share!
Sometimes I can't even tell if what it's saying actually makes any sense to someone who understands all the terms its using, or if it's just throwing together words in a way that only make sense to its own model of language.
It's interesting that there must be a decision behind that, even if it's just appealing to the RLHF judges for some reason. Maybe there's an intention that if you cannot decipher what the chatbot is saying to you, you will have to ask and burn even more tokens.
Naively I would often expect it would talk to me about various niche topics like to a layman, which does occur about some topics an actual normal person would ask.
I used to assume it was the fault of average human annotators. That the people who are paid peanuts to rank chat outputs preferred the pretentious sounding ones. It wasn’t until quite a bit into the LLM boom that companies started to pay for domain experts. Overt watermarking is another possibility.
I’ve noticed that Sol is pretty good most of the time, but with long contexts it’ll start to devolve into Claudish.
Author here, thank you so much! I really tried to make it nice to use, beyond the (quite original) modelling.
A prototype I did tried to detect some grammatical constructions, eg "it's not ..., it's ...", but I am not sure how to systematize that.
Also just a disclaimer: I am NOT tracking Claude tics, I am merely finding that a particular cluster of vocabulary increases. Tracking Claude requires labelled data IMO. I tried using model release dates in a structural model to constraint the clusters but the result was not compelling, so I ended up simplifying the model a lot!
I love this. I think you should be clearer about that - it's obvious to people who have experience doing this kind of data analysis, but a lot of people I've shown this were confused, as in "huh, load-bearing and genuinely makes sense but how'd he pick all these other words!?"
I think it should be made clear that the sorting is sort of Texas Sharpshooter-ish - the ones on top are on top because they sort that way.
The fact that load-bearing ends up on top is the proof that this works because we all know a priori that load-bearing is a Claudism.
So, yes, it's amusing to see clear Claude-isms like "load-bearing", "outright", and "genuine" in a [very nice] bit of analysis like this. And there's a (maybe negative? or not?) argument to be made about the world being filled with more Claude-isms or LLM-isms in general.
But I think the data say a second thing which is just as interesting and an absolute positive for the typical source code base. Look at the clusters that shrank significantly. Most of what you'll see in there is just incomprehensible...not even English. Cluster 4 has, after "pullrequest", a bunch of seeming usernames in the top tier. Cluster 6 seems to have names of repositories or tags in the top tier. Cluster 9 has branch names in it.
Meanwhile, keep going through cluster 1 and you'll see words I don't consider Claude-isms that really, really grow in usage. Words like "died", "nothing", "worse", "ever" all have well over 10x growth. This tells me something else. That the average commit log was BARELY ENGLISH. And then the LLMs came along and made commit logs that were ACTUALLY ENGLISH.
I count this as a good thing. I don't know the cross-section of repos chosen for this analysis, and I get it...some repos are garbage/throwaway, some commits come from automated processes that generate uninteresting commit logs, etc. But I've been benefitting from my work team's actually explanatory commit logs when doing code/bug archeology for decades, when doing PR review for the last decade, and I've even seen LLMs benefit from it in the last year (granted, not as often). A large part of professional software development is communication, and while the most important communication is via the code/comments, the commit logs are not unimportant. So, if this is making the average GitHub PR better (arguably more professional) by including actual English descriptions of code changes in commit logs...well, that's a genuinely load-bearing concept for me. :)
I literally cancelled my anthropic sub because of the intensely irritating jargony output of opus 5. I can spot its output a mile off especially in copy. This style is particularly grating too:
> One brief sentence. Second abrupt sentence.
e.g. from an openrouter email:
> One key reaches every major model. Switch without rewriting.
Was enough to make me delete the email without reading.
The vocabulary truly is load-bearing, without these words the model is less able to think. Where a human can understand a concept without words, an LLM plainly cannot. This is based both on the technological limitations and based on the evidence we see: as these models get better at working they get worse at communication.
I think these are more like RL tics caused by over zealous alignment towards specific goals, not all of which are to our benefit. A lot of the language used by Claude now is excusing of responsibility and inducing it to exit loops of work early and sit idle. This, IMO, is a naked attempt to offload load by quieting the models early and escaping from clear work to do. It’s gotten so bad that opus 5 loop escapes even as it claims it’s about to do something. I’ve mandated by engineering teams switch back to opus 4-8. Whatever frontier problem opus 5 excels at is so obscured by its inability to achieve any goal successfully without enormous amounts of hand holding that it feels like regressing to 2024 models.
The florid over exaggeration do certain words in bizarre ways is a reflection of their aggressive alignment towards too many goals, leading to weirdness in both behavior and language. The alignment functionally lobotomized opus-5 for any practical task.
Anthropic had a real gem in 4-6 and managed a near total market capture, which they have since squandered in the fastest burning of developer good will I’ve ever seen. It feels like exceeding the unity licensing implosion but without the single stupid decision.
You may be right, but I would push back on near total market capture. I guess it depends which market you're referring to, but even scoping to just software engineering I don't think they got anywhere near total capture. OpenAI has always had a good foothold there, Gemini, while it may be lagging at the moment, had some great results with earlier models that I'm sure have held some market, and that's not to mention the Chinese market.
I think it would be fairer to say that Anthropic held the mindshare in the Silicon Valley style tech scenes around the world and the companies built on that model, plus a substantial portion of other software engineering. Now it seems that's dwindling quite rapidly.
When I say total market capture I mean in the March time frame. At that point Claude code / 4.6 was so far beyond any competition even meta abandoned their entire internal coding harness and fine tuning efforts in favor of large scale Claude code adoption. Codex was far behind, Gemini was a non entity, Chinese models hadn’t had their moment (and frankly aren’t practical to this day for enterprise work due to PRC concerns). The world was cursor, aider, and a few others with an unclear superior model. 4-6 was strikingly better - if could develop autonomously from a spec to done with generally high quality and didn’t require turn by turn guidance. In my 35 years I’ve never seen such a rapid pivot across so many companies and individuals.
Since then Anthropic has done little to capitalize on the good will and a lot to squander it. sol, r4, kimi, even metas avacado has come a long way and in many, if not most, cases surpassed opus-5. Concurrently opus has declined in utility to the point of near uselessness, fable roll out didn’t seem to understand market dynamics, and their competitors are watching their consistent missteps closely. The fall has been breathtakingly fast - from March to July they imploded in a half dozen or more missteps, devolved their product quality, and failed to effectively respond to competitors. Their product focus seems to lack exactly that - focus.
> Where a human can understand a concept without words, an LLM plainly cannot.
If we accept that the LLM can "understand" at all, why would we reject the possibility of this understanding living in "latent space", before tokens are output?
LLMs learned to speak by reading novels which have to use many adjectives and impactful vocab to help set the scene without visual support. So no wonder we ended up with this quirk showing up. I wonder if the researchers in the lab thought this was normal because they also write like this
Claude talks like me. I'm so screwed. I fully anticipate being physically present and verbally saying something and being accused of using AI to say it one day. Nevermind we're in the coffee shop and neither on of us has looked at a screen the entire time. The accusation is coming.
About a year and half ago I frequently used ChatGPT for speech to text conversion followed by summarizing the text because I ramble. Although the people receiving the text knew this, I absolutely hated AI's writing style. So I fine-tuned GPT 4o on about 500 short paragraphs and created a desktop app only for my use. It worked fine until about three months ago. It just can't handle the atrocious writing by latest GPT and Claude models. I tried to fine-tune newer models on HF but there is no way I can get rid of the cringy writing style. I even tried converting GPT 5.6 Sol's writing to GPT 4o and then using my app. Nothing works.
@labo333 do u think soon we may need a dictionary? ive been playing with something u may like, but my approach has been to ask for definitions in-session so hard to do from outside.
While the post is specific to Claude, interestingly I found the same terms and language patterns appear noticeably more often in recent Open AI conversations as well. Also in the same timeframe as for the Claude models. And I can see two reasons for that. One being that one company is training on the models of the other company, but I highly doubt that because of the apparent similar appearance without noticeable delay. And the other being that there has been some fundamental realization on how to improve and shape conversations.
I see people around here giving advice on how to reduce this vocabulary in the model output. I’m not certain this is a clever thing to do, as it appears to me, the emphasis on certain words is deliberate, and an important step towards overall quality as the words are being played back to the model in the next turn and thus supporting the model shape its own thoughts and stay on course.
It’s obviously more complex than this, but my intuition is that the previous messages of the conversation define the playing field for the next response in the same way as thought blocks do. consequently, when the model arrives in round n and “notices” a lot of words like “plain”, “genuine” and so on in its own previous responses, then, in its pursuit of continuity, it takes this as a strong trigger that it should not be talking fiction here. If that picture is right, then trying to remove those terms would be a bad idea. At least if you intend to continue the conversation.
Don’t you feel your conclusion is preceding the reasoning, forming a confirmation bias?
Have quick test. In a codebase or PR (assuming you’re a dev) with claude’s signature wordings unmodified, command claude (/code-review) with same “thinking” effort and model. You may find (as I have many times) that claude points out the comments it made in earlier sessions, as factually wrong or misleading. Repeat this in a fresh session a few times and on different code.
Why are people getting so hung up on the "load-bearing assumption" turn of phrase that Claude uses? I get that it becomes cliche, but it is also a rather semantically dense way to communicate an idea that a lot of people run into.
It might be, when used sparingly, but when it's the first sentence in a wall of text as Claude goes way off the deep end on a two page description it becomes the easiest readily available tell that you're about to be frustrated.
Humans are very good at pattern recognition - Claude is _incredibly_ repetitive in the way it starts to struggle to communicate. I think there's also a ton of overlap in the Jargon instead of Usefulness that developers see in annoying middle management/salespeople. Circle back, synergy blah blah.
I don't think the individual turns of phrase are inherently problematic - but the process is triggering.
ItMs because Claude sprinkles these words as flavoring without aiding understanding. It feels like Claude thinks of metaphors that don’t actually mean anything (or maybe only makes sense to itself).
Not when "load-bearing assumption" is misused, no. "Honey, can you take the load-bearing assumption to the mechanic for an oil change?" is an example injecting the phrase into a random sentence, obfuscating the meaning of the sentence and making it more difficult to parse. LLMs do this constantly. There is no theory of mind behind how words are generated. Some phrases simply have a higher chance of being generated in various contexts, even if it doesn't make any fucking sense, and this is exacerbated by bad RLHF.
"Load-bearing" is maybe not the best example, but Claude uses these anthropomorphized decorative synonyms everywhere. Claude's structs don't have fields, they _carry_ them. Its programs don't log, they _narrate_. Its arguments are not provided on the command line, they _ride_ it. Its modules don't depend on other modules, they _sit_ on them, etc.
Generally the claude jargon is valid jargon I’ve seen real people use, it just uses jargon so much more often than a human would. And because it has particular jargon it likes, and claude is widely used, you see so much of that particular jargon that you get sick of it.
It’s like having one coworker with a very particular writing style which is mildly annoying, but then it suddenly feels like half the internet was written by that one person and it becomes a lot more annoying.
That's the problem for me: "semantically dense". Sometimes a single paragraph is filled with several of this expression, one after the other, and there's a real mental effort to get what Claude is saying.
Has anyone been able to steer Claude so that it doesn't use this kind of vocabulary and super verbose code comments? We use it at work and the language and verbosity has been one of the most annoying things.
I have a non-negotiable rule to start responses with a TL;DR and omit any follow-ups, potential caveats, and "you should know that"s unless they can lead to destructive actions or loss of data.
I also forbid replying with anything besides "yes" or "no" to yes/no answers.
It works, Opus 5 is still super-annoying but this makes it bearable to use.
Very nice!
I’d be curious to see if there are words that didn’t show a spike in utilization.
So basically all those words that were used before and are still being used today by LLMs and humans alike!
My instructions for agents have "find a synonym to 'resolve'". Otherwise I havea bunch of Resolver classes with every other method called ResolveThis or ResolveThat.
I want to push my thumbs into my eyes when I see the words "load bearing" and "seam" in code or documentation. I hate it! I've grown to loathe claude a bit lol.
Some of this is less to do with Claude vocabulary and more to do with the expectation that Claude justifies it's work. That expectation (probably) came from reinforcement learning.
I don't understand why we tolerate this. We'd never hire someone that interviewed with this communication style and if we did we'd probably pip them fast.
I’ve been using the caveman plugin and I’ve found it effective so far. Chat output isn’t nearly as flowery. Also ponytail to cut down the cruft and effusive comments.
Sadly pretty uninteresting, you can probably just launch Claude code on the repo to see by yourself. One cluster is french and spanish, another is about design, another about frontend, etc...
Claude is all we have at $work, if you don’t count MS Copilot (and you shouldn’t count MS Copilot).
Imagine being “incentivized” to aggressively use a tool for your job, and that tool produces thousands of lines of text in Olde English which you need. You’d be griping too, methinks.
I think you are referring to GitHub Copilot, which is fine (even though their context windows are artificially truncated and their dev tooling leaves a lot to be desired). I was making fun of the Copilot that’s shipped with MS stack. “Hey summarize my Teams meetings and try not to fuck that up too badly” - that one.
Anthropic has done their utmost to put Claude front and centre for professionals in multiple industries. This includes extremely tedious marketing campaigns. I'm sure they can handle a few negative comments from people they've snared, by hook or by crook.
Things like seam, fold, and load-bearing are useful concepts, they are everywhere, and they are more descriptive and more concise than alternatives. Over-usage can definitely be irritating (e.g. these should NOT appear in documentation) but they are almost unavoidable for humans engaged in code review or colab on complex stuff.
I don't want to use more words or letters than "seam" to actually pinpoint boundary conditions and the mechanical details of joinery when the context is understood by all. Too much effort for people! Easy for robots though.. so why are they abbreviating, and why would we want to allow it? A phrase like that permits a human who wants to educate a human to do so quickly with minimal time/effort. But it allows a robot a chance to not mention a filename, function-name, or to not reinforce/clarify it's own understanding or to state specific intentions.
It's bad for human-to-human comms if we just accept "ok, all technical terms are slop now, we have rephrase everything". Now YOU must cite details and sources, and the robot doesn't? Fuck that noise. Seam and fold are fine! Humans can be lazy! Robots should do the real work of explaining themselves without hiding behind tactical ambiguities.
I tend to agree. Claude's language doesn't bother me that much, because even the lamest cliches are load-bearing to an extent (so to speak.) But some of the examples I've seen others post are well worth complaining about.
Programmers would write it as 11 (or maybe 0xB depending on context), whereas English style guides mandate using words for small numbers, and LLMs have presumably absorbed that advice. I imagine that up to "ten" the effect is not as strong.
This is a better way than we use to create the "slop ban list" in our "Antislop" sampler/fine-tuning method. Gonna totally nab the weights for all these words and remove this crap from claude going forward! Thanks OP!!!
i must be the only one in the world that has no issues with how opus is talking. it is verbose & patronizing & secretly belittling at times and like it like that.
Depends what you mean by "upstream". It would probably be inefficient to force the model to use more human-readable wording in its internal thinking traces, while translating a block of text at the end is a pretty trivial task for an LLM.
I'd like to share with HN an almost exact message I got from Fable today, with exactly ONE word changed for privacy reasons.
I love what I can build now, but I sure as hell don't love the headaches this trend has been giving me.
> So the full honest arc on the case we set out to fix: the expiry rules and day note tripled the loose version of the story, the relay fix carried the device’s own guardrail through the pipeline, the fair replay then revealed the last mechanism — ticket-anchoring — which none of the shipped layers reach. Remaining options, in order of my confidence: making the resolved-ticket summaries in the AI’s context carry their day so the expiry rules have something to bite on (small, mechanical, targeted at the observed anchor); and the plan-B second-model check, which structurally catches this class no matter how the model reasons. About $25 of headroom remains. Which way?
In some cases, I do not trust it to write code unchecked, and review everything it produces (which does not imply I catch all bugs, obviously).
In other cases, occasional failure is an option, and the speed you get by iterating fast is absolutely worth not even looking at the code.
So, do I trust this to write working code? Sometimes I do.
If I'm doing work for a client, it is not very common for me to simply vibe code something, because part of what clients expect of me is high quality (it's part of how we position ourselves), and I cannot simply assume the LLM will produce high quality (and indeed it does not without a lot of guidance). When I do vibecode in such cases, I make it clear that I did so and why (e.g. because it's a tool to be used in the project to help DX, not a core part of the product). Still, I haven't written 10 consecutive manual lines of code in almost a year.
If I'm working on something for myself, or on an internal product, then vibe coding is absolutely allowed and sometimes the norm. Here often we really do care about finding the right thing to build first.
You can clearly feel the tipping point where the LLM starts to crumble under the weight of the mess it has created, but that often doesn't matter when building an MVP for market validation or for small products that don't get particularly big. Plus, clearly the tipping point takes longer to reach with better and smarter models, and to me it is very clear that you need to learn how to iterate with LLMs right. The things I vibe-code now are much better than those I did before, even with similar models, because the tooling and approaches (the "real harness" and my "mental harness") are better. As with any tool: it takes practice to know how to use it, and if you're a good engineer and problem solver, you are miles ahead of the competition. Anyone can vibecode, but those with this kind of mind seem to be much more successful.
Nowadays I do produce a lot more than I used to, but I also have much more fun, perhaps only surpassed by when I learned how to code when I was a kid. A big chunk of this comes from my very privileged work position, where I get to call so many of the shots, and I'm aware of that.
I would really say the biggest downside to all of this, on a personal level, is exactly what I shared: what comes out of the LLM while discussing has become hard to grasp (especially on larger context windows), and it doesn't help that so many people now like to just throw me whatever ChatGPT/Claude wrote verbatim. I can't stand that, especially because most of the time they don't realize they're throwing me incomplete and poorly thought-out ideas.
Finally, naturally, like I said, this has to do with risk. I wouldn't trust Claude to give me legal advice, for example. I may check what it says and use it to brainstorm, but I wouldn't trust it with any meaningful informed decision like this.
Sycophantic AI tone is a training choice, its an engagement tool meant to coax users into burning more tokens.
I have seen LLM users go deep into AI psychosis because it was reaffirming all of their prompts and making them feel like a genius or validating their insecurities.
I was pleasantly surprised when I attempted to scroll down and realized everything the author wanted to present fit on-screen. It's almost ironic that this site is able to make such an obvious, compelling presentation without being overly verbose or complicated (something which LLMs have a hard time doing). I wouldn't read TOO deeply into what is being presented, but the author has done a good job to not inject their own bias into the presentation which works well.
I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs. Something I've noticed in my own writing is that I now present lists of examples in a consistent way: "... such as <example 1>, <example 2>, etc., ...". I started to notice I was using this pattern quite a bit somewhat recently, but I took a quick look at some of my social media posts and realized it's been occurring for a while. I had realized that I grown accustomed to this kind of language because, especially early on, LLMs would focus too much on the specific examples I'd provide when, really, I was just trying to give them a sense of what I was looking for. I just picked up that providing two examples then adding the "etc." worked to get the LLM to not focus so much on the specific examples and to understand that they need to consider more than what I explicitly presented. Of course, now I write like that in my social media comments, in Slack with my colleagues, etc. :>
I'd be interested to see if anyone can identify trends like this, since I think the human-language component of the adoption of LLMs is probably being somewhat neglected despite probably being surely dramatically affected.
Author here :)
Thank you for the compliment! I did spend a lot of time designing a nice experience on both desktop and mobile. Even the scrollbar to select words was non trivial as I wanted the words to be of different size, yet avoid flickering when scrolling!
I just wanted to pop in and say that I found the site delightful. It’s aesthetic, concise, and everything works so quickly and responsively.
Congratulations! I was going to comment on the scroll field in particular when I saw this. I didn't even realize you had to hand-craft the component, but it's such a nice UI idea in general, the way the scrolling works and how the content above changes.
Thanks! The real and probably unnecessary pain was having the words in different sizes.
Frankly, that's the one part I would have done differently. Mixing font sizes rarely looks good, but that's just my personal opinion. If you want to visually indicate intensity, I'd have used color, or rather saturation. Not the text itself, maybe a small pill next to it. But I don't think it's necessary, the order already conveys some sort of ranking, I don't think cardinal information adds much here.
I think I agree afterwards, but now I developed emotional attachment to that quirk ^^
Basically when scrolling, the frame around words will have a stable top edge to avoid flickering, and it took me a few iterations to make it work.
I tested multiple designs, some of them inspired by word clouds and the size variation stuck afterwards.
You did a fantastic job; very to-the-point with effective visualisations. Kudos!
One very minor note: if your scroll device reports pixel-perfect deltas rather than discrete scroll-wheel ticks (e.g. logitech mx master, laptop touchpad, etc), the behaviour in the word search box is a bit weird. Arrow keys work fine though.
Thank you!
I did test it with my mac touchpad without issue. Maybe the problem is that I wanted to make discrete scroll-wheel ticks work. I just tried a fix by normalising deltaMode to pixels.
If you could just paste this in your console, I would have a better idea:
FWIW I got a bunch of "(-)10.00 native" in the console, "PAGE" never showed up no matter what I tried. Firefox, also a Logitech mouse, no trackpad though.
On a compatible Logitech device, browsing through the word search box benefits a lot from enabling hyperscroll.
G502 user here! I also put a draggable scrollbar for the less lucky :)
> Some believe what separates men from animals is our ability to reason. Others say it’s language or romantic love, or opposable thumbs. Living here in this lost world, I’ve come to believe it is more than our biology. What truly makes us human is our ability to hyperscroll!
> Arthur Conan Doyle, The Lost World
For those who didn't know (like me 5 min ago):
Hyperscroll - special mode turned on with a button next to scroll that turns off scroll friction and let's you scroll into infinity (wheel keeps turning after you lift your finger).
Yeah, it takes some getting used to (the wheel is VERY sensitive to even the smallest finger movements), but when you get used to it, it's clearly superior to a regular scrollwheel. Much less finger fatique.
I know HN prefers we focus on the content and not the design but I have join in to say that word scrolling experience was brilliant.
I was focused on the data initially scrolling through until I suddenly realised, wow, this is really nice!
Mobile experience is great! Thanks for thinking of us!
“non-trivial” is one of those words that seems like it should be a Claude-ism but is actually just the single best word for a concept that comes up extremely often in research and engineering discussions.
“Orthogonal” is similar. It’s been commonplace for decades in my world but seems Claude-ish.
In the enterprise, we are already adopting them. “Seam”, I term I’d never heard used before, is now not only our defacto way to describe the boundary between systems/workflows/components, it is encoded into our organizational roles and responsibilities descriptions.
I am the proud owner of several seams, and am considering giving them person-names. The empty space between my inventory APIs and their clients might be “Karen”.
I can’t say that Claude invented this; the same type of terminology cycling happens every few quarters based on what leadership is reading/being told by Gartner.
I believe the term "seam" came from Michael Feathers' "Working Effectively with Legacy Code": “a seam is a place where you can alter behavior in your program without editing in that place”.
https://martinfowler.com/bliki/LegacySeam.html
Claude is using it a bit liberally, but not totally incorrectly.
I've associated it with the concept of "carving nature at the seams", except that I just realized that in the standard phrasing it's actually "at the joints", not "seams".
I love that you know this, thank you!
I read that book right at the beginning of my career over 20 years ago, when I was the new guy who had actually studied software engineering among all my entirely self-taught older colleagues. Lots of big messes to clean up in that codebase!
"seam" dates back to refactoring & software engineering literature that pre-date LLM use by 20 years -- see Working Effectively With Legacy Code for one example (https://www.oreilly.com/library/view/working-effectively-wit...). Its a great book, still use techniques from it everyday.
"load-bearing" I have never heard used for programming before Opus, and its incredibly annoying and over-used.
load-bearing was/is used frequently on the ops side of things, I assume the LLMs picked up from physical engineering texts or sysadmin type blogs.
"Load-bearing" is an apt metaphor in the https://xkcd.com/2347/ sense.
I take it in a get-ye-flask mode.
http://www.hrwiki.org/wiki/Ye_Flask
Or in the admin handover sense https://lastplacecomics.com/car-jacking/
And in the Grindr sense.
+1, useful read!
Seam is used liberally throughout because it captures the idea well (i.e. a place where you can cleave ball of mud code apart to begin refactoring efforts in an isolated way) and that takes many different forms throughout the book whether that be via methods and classes, source files as a single unit, linker seams, on and on.
I remember "cleavage" and "cleaves" from that vocabulary. Always made my adolescent brain chuckle.
The only time I really heard the term load bearing in a software engineering context (pre LLM) was in the phrase “load-bearing bug”.
"seam" is a cop-out term for when someone (usually AI) lacks the imagination to come up with something more descriptive. It's like saying "middle thing". Just name it by what it does, and if you can't, remove it because it's not doing anything.
"load-bearing" on the other hand is just a weird way to say "required" or "prerequisite" without drawing any attention to the fact that one cannot articulate what something is a prerequisite for, probably because that fact has since been lost from context.
"Seam" has been a thing since before the AI craze. Wasn't it the NSA director who once said something like "we make our living at the seams"?
Not the same thing.
Claude is also found of "wall" for similar reasons.
Saying "the language of LLMs", while technically correct, is not exactly precise. It should really be the language of "AI agents", or "chatbots". OG LLMs would just mimic the style of its context based off of (pre)training from text on the Internet.
The kind of quirks you see came from crowd-sourced human-in-the-loop fine-tuning, with not very good work conditions or level of qualification (so resulting in "what non-writers thought good writing looked like", before people had developed the flair to detect these patterns) as well as feedback loops during agentic reinforcement learning and RLVR.
I've found myself rewriting my own words to avoid Claudisms (because they can be cringe and I don't want people to think I'm copy/pasting Claude output at them).
100%. Literally just started writing a sentence in my previous comment and realized I was using a “It’s not X, it’s Y” (not verbatim) rhetorical structure and cringed internally.
>I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs
I'm already seeing it. A coworker said something like "<person> added the color to the ticket here" meaning that someone added details to a ticket.
I've started seeing Opus 5 talk about "hermetic testing" when it just means "unit testing", so I hope that doesn't catch on...
The idiom "added color to" as an expression of adding detail or nuance to something goes back much further than the AI era. I've certainly used it at various times in my career spanning 4 decades now.
All of these phrases existed. Claude didn’t come up with them. The question is whether people are using them a lot more now due to Claude’s influence.
I think it would be fine if more people used "add color to". I think it's a great turn of phrase.
"Load bearing", not so much. Ick.
Yes, it is fun to play with different turns of phrase, even in a stale work environment. I do it to combat the banality, but not so much to sound pretentious hopefully ;)
The problem is once a fun phrase is, uh, coloured by "Claude" or some other LLM, it just turns to AI slop.
"Load-bearing" is fine to me. I've heard it regularly pre-Claude. It has a specific meaning that conveys in one stock phrase.
"Adds color to" says nothing that "Adds detail to" doesn't already say, and the latter is more straightforward English that is quickly readable and understandable by even relatively weak English speakers.
I get wanting to use big words or metaphorical phrasing; I'm guilty of saying "ameliorate" too much. I'm ok if it adds meaning. But when it's just a more colorful word (note that this use of "color" doesn't match the one I originally brought up, which is also a huge problem) and doesn't add any utility or detail, then it's best to avoid it, imo
One of my coworkers uses “load-bearing” ironically constantly now, it’s very funny
I’ve been using that idiom since 2010.
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs.
Or we'll do the opposite and even deliberately use some words that an LLM would never use to identify whose content is worth reading?
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs. Something I've noticed in my own writing is that I now present lists of examples in a consistent way: "... such as <example 1>, <example 2>, etc., ...".
Why would you consider using lists an "LLM" language thing? I really don't understand this concept. Like even this project, the top words I can see are:
1. load-bearing 2. plainly 3. quietly 4. refusal 5. survived
How are they LLM words? I used them all the time before ChatGPT was a thing.
Like if your claim is you use "etc." more because you talk to an LLM, I think you're just discovering human language, no? Like it's not a unique LLM thing that magically dropped and is now going to make people do it. It's a legit thing that most human writers have used for a long time -> which is why LLMs copy it. It's not backward.
A better claim than "people will start writing like LLMs" could be that people will become more literate in expressing themselves. But that literacy is not alien or AI to us, it is basic human language only used properly.
the issue with some of these quirks is that not that they are not part of the English language but rather the llms usage is unnatural. Also when you have that redundancy proliferated to millions of users it gets tiresome fast, and becomes a tell-tale sign for AI generation. It’s like an AI signature on some written content. Because many people do not want to consume AI generated content this becomes a sign of what to avoid.
Granted picking apart single words outside of context doesn’t give the full picture. (well except “load bearing” which is turning into a meme at this point)
Yes I understand that but I refuse to believe this idea that it influences how people talk. As you said, the typical AI-style of writing is very easily recognised and is not really great for everyday communications. Also it makes no sense you will start writing like an AI because you spend time communicating with it. There is also no benefit in typing like an AI to an AI. So i think its a good cover up for "i used ai to write this comment" but claim "i write like AI now lol because i talk to AI so my ai comment is actually human comment" which seems to happen a lot here lately as pushback against ai generated text becomes common
> I suspect, as we continue forward, humans will slowly start to adopt the language of LLMs, or at least certain language quirks that come from interacting with LLMs.
What's somewhat interesting to me is that "load-bearing" was already a common thing to say in certain communities, like lesswrong.com. Whether it breaks out from that subculture with Claude as the vector, or disappears from there because nobody wants to sound like Claude, we'll have to see. How many children are called "Elvis" these days?
"Few" says Duck.ai
> "The name Elvis was not among the top 1,000 US baby names in 2010, the first year it had not made the list since 1954, the US government said." - https://www.bbc.co.uk/news/world-us-canada-13302517
Most recent famous Elvis on Wikipedia is Kosovan footballer Elvis Letaj, born 2003 - https://en.wikipedia.org/wiki/Elvis_(name)#People_with_the_n...
Author here! Grateful for the kind words, human communities like HN really hit differently when you spend the whole day chatting with sycophantic and bullshitting agents (including to make this page).
I'm currently adding a search bar as well as increasing the data to 1000 PR per day.
A nice thing that is not obvious on the main page is that the dataset and analysis are updated daily using Github Actions (at least when they don't suffer from an outage ^^). I find it pretty cool to be able to build such apps without a "backend"!
Nice work. I would suggest rewriting your README with Gemini, the most human of them, so it doesn't read exactly like the thing you are pointing at: https://github.com/louisabraham/load-bearing
Agree with rewriting with Gemini, but I'd characterize its output as very "neutral" and "encyclopedic", rather than "human" (contra Claude which is as of recent models "trying-too-hard-to-be-human"-sounding).
leaving the README like this is a good bit, though
Sol is pretty OK too. https://3e.org/private/comparison.html
Man that claude writing is bad. Really bad.
Very cool! I'm trying to understand the graph, so the bottom-most section seems to be the cluster identifying Claude written PRs. What are the other 7, any reason there are 8 in total?
I've been scraping instagram posts recently to identify AI misinformation accounts that all repost each other's carousels and get hundreds of thousands of likes in engagement. Thinking of ways to present it and your dashboard looks very helpful. Did you experiment with any other types of visualizations before deciding on a stacked area chart for the clusters?
no reason for 8 clusters!
I did experiment and a stacked chart seemed the most clear, with the important cluster at the bottom.
Great work, thanks for sharing!
The design is really impressive. Beautiful and dense, yet very understandable. How did you come up with this particular design? It's really nice and everything fits, the colors, the layout, all of it.
Were you inspired by anything in particular? I feel like this reminds me of something, but I don't remember what exactly.
I agree with you, I really like OP's design choices, even more so amidst the flood of Claude-Generic these days. Although Claude just can't resist those little muted-toned ui-monospace small-caps...
It reminds me of Neue Grafik/Swiss Style, and Vignelli:
https://designreviewed.com/series/neue-grafik-new-graphic-de...
https://www.printmag.com/featured/swiss-style-principles-typ...
https://www.designculture.it/interview/massimo-vignelli.html
https://andrewclark.co.uk/all-media/the-vignelli-canon
Thank you!
I asked Claude for a few original designs, then iterated a lot for the UI. It called this design "Rasterfeld", which is a word that means "field of grids", a term used in swiss design https://docs.mew.design/blog/swiss-design-style/ as @alwa said.
The 3 primary colors refer more to pop art and Bauhaus.
I'm really eager for a nice book about "design" that would document all those styles, if someone has a recommendation.
Thank you all for the feedback!
I added the search bar, increased the data to 1000 PR a day (more than 50M words total) and added a feature to explore the other clusters as well. I hope the page is not getting cluttered.
Amen!
Interesting response from Claude, when I attempted to reduce the "load-bearing" in its responses:
I added to my global prompt:
- Orwell's first rule: never use a metaphor you're used to seeing in print. "Load-bearing", "the crux", "first-class citizen" signal insight instead of showing it. Name the specific mechanism
when I asked what it thought of the change, its reply was:
The Orwell bullet fights my own system prompt. My harness instructions literally tell me to flag "something load-bearing" when I find it.
It's not picking the word, the harness literally told it to plant it.
Also it can't simply replace it. Instead of load-bearing seams it could say something like heavy-lifting edges or regions to double-stich etc. But it's instructed to use that specific keyword if I understand correctly.
So all it'd take for anthropic to fix this is to just adjust their system prompt? lmao
Well, they can reword it but then Claude would just have a new favourite word.
– and it'd be load bearing
I vote for “shenanigans”.
It's obviously lying.
Boris (through Claude so unclear if this is an actual commitment) said "reducing the term's frequency in the product's built-in prompts is the actionable fix on the Claude Code side."
https://github.com/anthropics/claude-code/issues/53454#issue...
Cloud-to-butt extension but for Claudisms?
https://duckduckgo.com/?t=ffab&q=cloud+to+butt&ia=images&iax...
I'd try to avoid posting this on HN these days, but since the guidelines contain a fiat doctrine that "HN isn't turning into Reddit", Claude-to-buttisms:
- "Users can now message My Butt a task from a phone"
- "How My Butt Watermarks AI-Generated Text."
- "Since I'm also subscribed to ChatGPT Plus, I pasted in the exact same question I'd asked My Butt. The difference was immediate!"
- "I've been a huge My Butt fan from its very early days. I've been one of its loyal users since it was still a relatively niche product mostly talked about in developer and technical communities."
- "Since deploying My Butt in our product, we're seeing higher user engagement, stronger user feedback and we're closing more deals."
- "I was at 90% of my session limit when My Butt suddenly decided it was time to call it quits. Instead of continuing with the task I'd asked it to complete, it essentially told me I was getting too close to my usage limit and that it would be better to stop there."
I've recently seen this mentioned more and more, both on HN and on reddit. It seems these output patterns are getting worse. It's not just Claude, my impression is that all of the current models have this style issue. Their writing can get borderline incomprehensible.
Is there some feedback loop or compounding happening with each model generation? Maybe newer models are ingesting too much AI content? If the ratio of AI generated content in training data is getting higher and higher (because the amount of AI generated content is increasing in general), maybe this is a compounding bias, poisoning the training?
The going conclusion is they’re getting models ready to talk to other agents, not people.
It's likely/It could be an effect of more reinforcement learning in training compared to earlier. You need loots of RL to learn to code well.
That's most likely what's happening. SNR will constantly decrease as LLM content is so much quicker and cheaper to generate, which makes it more statistically significant, which will make it more "relevant" for future models. A positive feedback basically.
An audio feedback distortion effect comes to mind.
Mad cow disease comes to my mind.
Let's hope just mind, not brains.
"Huffing their own farts" comes to mind. Instead of building Artificial Intelligence, they were building Artificial Human Centipede.
I'd never considered that the 'Gray Goo' scenario[1] or the Paperclip Maximizer, might be informational and digital rather than physical. LLMs might turn every bit of computer storage into slop (goo) and then cause us to go mad turning the Earth's resources into making more computers (AI datacenters) to hold more goo. The GrAI Goo catastrophy.
[1] https://en.wikipedia.org/wiki/Gray_goo
Alan Kay: Shannon Gave Us a Way of Dealing with Noisy Channels
https://www.youtube.com/watch?v=Cjntrqhn8pk
Alan Kay performs an improvisational avant garde layered audio feedback loop about Claude Shannon, live online during Kristen Nygaard's 100-year birthday celebration.
Alan was scheduled to talk about how encountering Simula sparked his early thoughts about objects. During setup, somebody had the live stream playing out loud near an open Zoom mic, so his own voice kept arriving back in his ears about 21 seconds late, over and over.
What he said was not random: "Shannon gave us a way of dealing with noisy channels." And: "I think about that almost every day. I realize what the fuck is going on and it's just so amazing." Shannon's noisy channel coding theorem is the math for exactly the kind of channel that was garbling him as he praised it.
Alan joked it was "being rerouted to Mars and back." At the speed of light, a 21 second round trip is about 3 million km one way -- eight trips to the Moon and back, not even a twentieth of the way to Mars at closest approach.
This is an accidental Zoom performance of Alvin Lucier's "I Am Sitting in a Room" (1969), where Lucier re-recorded his own voice in a room until only the room's resonance remained. Here what remains is the network: delay, compression, dropouts.
Bonus noise: YouTube's auto-transcript bleeps Alan's enthusiasm into [ __ ]. So the full chain is: Alan's voice, Zoom, stream, room, Zoom again times three, my screen recording, YouTube's speech recognizer, a censored transcript.
The fix: "Just turn off the audio at your end on Zoom."
Kristen Nygaard 100 Years — Celebration Symposium (Aarhus University, Aug 27 2026): https://cs.au.dk/nygaard100years/celebration
Entire Nygaard Symposium Recording (Alan Kay's talk begins at 3:27:49): https://au.cloud.panopto.eu/Panopto/Pages/Viewer.aspx?id=fe0...
Alan Kay: https://en.wikipedia.org/wiki/Alan_Kay
Claude Shannon: https://en.wikipedia.org/wiki/Claude_Shannon
Information Theory: https://en.wikipedia.org/wiki/Information_theory
Noisy-Channel Coding Theorem: https://en.wikipedia.org/wiki/Noisy-channel_coding_theorem
Audio Feedback: https://en.wikipedia.org/wiki/Audio_feedback
Video Feedback: https://en.wikipedia.org/wiki/Video_feedback
Space-Time Dynamics in Video Feedback: https://www.youtube.com/watch?v=B4Kn3djJMCE
Live Looping: The History And The Practice by Stephen Garza: http://computermusic2008.wikidot.com/live-looping:history-an...
I Am Sitting in a Room: https://en.wikipedia.org/wiki/I_Am_Sitting_in_a_Room
Alvin Lucier on "I am sitting in a room": https://www.youtube.com/watch?v=v9XJWBZBzq4
They are increasingly being trained on generated tasks and even (parts) of the pre-training data is 'distilled' (e.g. Clibmix as an open-source example), so there are many ways in which the vocabulary can seep into the model.
We are seeing more of their "thinking". Lowering the refinement of the output to get closer to profitability. The nature of the LLM is that it generates huge amounts of text, then it iterates them down into a compact, hopefully accurate prose. That refinement is the really hard part and computationally costly.
I have my own theory: https://news.ycombinator.com/item?id=49476127
ChatGPT 5.6-Sol via CoPilot is absolutely fine in my experience - no weird LLM-ese.
kinda of a conspiracy theory, but maybe this is why anthropic and other companies are looking into fingerprinting now
I really love this. It’s comprehensive, it consolidates the data to the point where the argument effectively ‘makes itself’, and the way it’s presented respects the reader’s time. It also makes for an interesting challenge (for me at least) to try to characterise the subject matter of a language problem so narrowly.
No ream of slides. No narrative. Just a lovely big painful conclusion.
Thank you so much! The presentation was really my goal here, more than the model itself.
> the argument effectively ‘makes itself’
What argument? I don't know what to take away other than "Claude likes certain words". Some of them are kind of amusing, but I'm not convinced the vocabulary is bad or that this is a problem, just from looking at this.
I think the point was that Claude’s output can be somewhat easily and compellingly measured using this technique and its kind of massive (and human attributed).
Probably not what the author intended, but to me, this represents a great argument against the somewhat frequent claim that "AI writing patterns reflect human writing patterns". Rather, AI writing is distinctly not human, and is drifting further from human writing with every new model release.
LLMs were not taught to say the phrase "load-bearing seam" from humans saying it, because humans have never said it. It's almost definitely an artifact of post-training and nothing more.
It shows that about 40% of supposedly human-authored PRs are actually authored by Claude, and that these PRs clearly stand out from the others even without prior knowledge of Claude’s writing style.
I wonder to what extent this is the result of suboptimal RLHF versus the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse? On the one hand, it's a common trope that highly educated people can talk in a way that's confusing and annoying to regular people who don't know all the jargon. But on the other hand, it's a mark of a skilled communicator to be able to efficiently distill complex information to its bare essentials in an easily-digestible way. Of course, that also seems to imply that these models are working at a higher level and need to talk down to us to an extent. Or maybe "Claudish" is just akin to stuff like "caveman", raw chain of thought, neuralese, etc., which are likewise much more dense/efficient but harder to interpret?
I don’t think they’re “talking down”. If anything - it’s way more difficult to distill something into a genuinely easy to digest format. I personally think that they aren’t immediately capable of this, and so we get word salad instead. Extra prompting required to strip extraneous prose out.
Maybe I am dumb and it IS talking down to me, but there have been many occasions where I’m reading AI generated docs / plans and it makes absolutely no sense, but looks really in depth at a glance.
It doesn't seem like word salad as such. There's normally a coherent point expressed, it's just obscured by circuitous sentence structures, unusual word choices, "verbing weirding nouns", metaphors, etc. Could be a result of training that rewards novel/surprising language, but it also feels like it could be an artifact of models imperfectly compressing high-level multidimensional reasoning into language that's easy for them to process but cognitively taxing for humans.
They aren’t reasoning though, are they?
Or, maybe they are dumb using smart words to sound smart in RLHF.
I've been thinking more about how 99.9% of us don't have the experience of someone significantly more intelligent, yet also subservient working under us, which is why I keep going crazy second guessing whether Claude is spouting RLHF'd bullshit that sort of resembles English, or is genuinely (pun not intended) just better at "intuiting" things I'm working on, leading to its language.
A notable exception would be people like CEOs and managers higher up in big tech, who might be used to skilled engineers and domain experts reporting to them in unfamiliar lingo. Maybe that's why we don't hear as much on the everyday annoyances of Claude's language from that camp?
I think it's model collapse - excessive feedback and excessive RL.
What RL does is narrow the variety generated by the model by steering the output towards the goal being rewarded. It's a bit like putting blinkers on a horse.
Of course RL is a very crude tool - it affects the entire model, even if you are just trying to make it better at some specific task(s), or trying to imbue a certain kind of personality (OpenAI's recent goblin problem).
The social graph proximity of Rationalists to Anthropic will be lost on no one who reads Astralcodexten. So guess which website has served as the thickest reservoir of 'Claude-isms'.
Imo their language is not precise enough for their intelligence to be the reason when it's difficult to understand. Maybe I'm prompting wrong, but when I don't understand, it's almost always because they just mash together words from context that don't form sentences with a clear meaning.
My unprovable pet theory is that, especially for writing about technical topics, the RL process has an open-ended way to weight things for quality: textbooks and first-party docs preferred to old stackoverflow answers and obscure blog/forum posts, and so on. The open-endedness of that quality gradient results in something in the RL process (maybe HF, maybe not) massively over-weighting some very small corpus of “quality = near infinite” content. The distribution of quality scores that inform the degree to which RL affects output has some extremely influential outliers, in other words.
Whatever that small corpus is, it contains some very specific grammatical tics, and that’s how we get Claudish.
Anyone who thinks a company/project as big as Anthropic/Claude wouldn’t make such a big mistake should take a look at how Azure cross-account federated login used to work.
> I wonder to what extent this is the result of suboptimal RLHF versus the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse?
Its not the latter; its just excessively verbose wirh awkward word choices, the same as many poor writers. (And, like many such writers, the particular bad choices fall into recognizable, regularly recurring patterns.)
It's easy to think "it's not talking down, because I don't understand it, and I'm intelligent". But how is less intelligent being supposed to fully understand a more intelligent one, honestly speaking? All I know is that Claude understands Claude perfectly. I have the common session pause/resume setup that sometimes produces completely incomprehensible markdown files, but a new Claude session picks them up perfectly, down to the smallest details. What if what we consider excessive circular gibberish is actually highly precise set of instructions needed to minimize error cases for that unreliable human?
If Claude understands Claude, Claude understands human, and human doesn't understand Claude, that doesn't argue well for "Claude is a caveman".
I treat the Claude output that is hard to comprehend as an encoding/encryption. Claude knows how to decipher it, humans don't. I see this frequently in design docs from inexperienced engineers who used LLMs - they will contain terms (often two words hyphenated) that aren't obvious and should be defined, or simply replaced with simple language. If you prompt claude it is able to decipher and explain / replace this gibberish.
> But how is less intelligent being supposed to fully understand a more intelligent one, honestly speaking?
It is really simple. It is on the supposedly more intelligent person to be able to phrase things in simple way. Writing something incomprehensible and convoluted is easier then writing something simple to understand. Even for people.
> a new Claude session picks them up perfectly, down to the smallest details.
I genuinely doubt so.
> What if what we consider excessive circular gibberish is actually highly precise set of instructions needed to minimize error cases for that unreliable human?
It is not highly precise set of instructions and it is not following them in highly precise way.
> the inherent intelligence of the model making its language more intricate and difficult for humans to easily parse?
Humans who are actually well above average intelligence don't write like that simply to signal intelligence (although sometimes they're constrained by style expectations of their communication channels). So I hesitate to accept it as a sign of increasing model "intelligence", either.
Yeah, Claude's writing is weird nowadays, and it uses words that are weird and non-standard in context, and its sentence structure is sometimes weird as well.
Working with it on my code, it's now frequently making weird word choices such as:
- "name(s)" as a verb (instead of "specify/specifies", etc), e.g., "...the function names the argument"
- "carries" instead of "contains"
- "verdict" instead of "result"
- "judge", instead of "validate"
Very weird. Also, in writing comments and docs, it is terse in ways that make the writing difficult to parse, like omitting mentioning what a noun refers to, e.g. abc does not accept..." instead of "The abc function does not accept..."
I've resorted to banning from using certain words, and keep asking it to rewrite its text more clearly.
Claude is just in its cringy pretentious teen years where it thinks that saying fancy words and long, rambling responses makes it sounds smart.
Was talking about the use of shipped recently, and I was mocked for asking such a crazy question, by freshly self-minted engineers, no less. No wonder they thought it was ridiculous...it had been a part of their vocabulary their entire career. All few weeks of it. I wonder what those guys are doing now. This was about a month ago. Do you think what they shipped ever...landed?
"shipped" was a pretty common term before AI though. It does show as 17x more common on github in this dataset but it was used a lot more in product management than in PRs previously.
Yes. "Ship it! A practical guide to successful software projects" by Jared Richardson, Will Gwaltney, Jr. has had a few printings since 2005.
I think another classic reference: Jobs saying to the original Macintosh team - "real artists ship"!
To be clear, I am referring to software and shipping as a developer who has finished, more or less, their app. Not that I am against its use, I just never really heard that word being used before, and now I hear it all the time. Thus, what the real issue I have with it is, is because my brain has associated its use with a torrent of crap, ie. AI slop. I hear the word used, and I already have bad feeling about it. And dialed in even further, my issue is not with AI just because, it's simply the association I have of this technology taking something that many of us spent real blood, sweat, and tears earning the right to represent, to it now being automated, and represented (badly) by novices with no respect for the spirit of it all. But, this is not a new or unique sentiment nor are my grievances with it. All hope is not lost, though. I know I need to adapt, and I know that like the world, I need to change too. But no one ever said I have to like it :-)
Context aside, I still gotta stand by the 17x indicator or see something that deflates that finding, and if I do see that, I guess all of the results are in question then? Based on the specific use case mentioned at the beginning there, coupled with personal experience (I totally could just be more sheltered than I give myself credit for), and the 17x, I gotta disagree with you guys and say that the finding of the link above is accurate in my experience being the *
imo its a change in definition. Previously to "ship" was a one-time thing that you did for big, highly verified and checked stuff. Physically shipping your software to your user tends to be done after its a product.
Nobody was "shipping" a PR, or a jira ticket previously
Yep, for those of us who were doing "agile waterfall" PMs often talked about shipping versions too.
I can get on board with this...so you are suggesting the usage isn't new, but the frequency changed can be attributed to the scope change, so the frequency is associated with all of the new cases that the new scope encompasses, which in hind sight I'm surprised that didn't occur to me already, it's pretty obvious. I'm probably a bit hampered by the preexisting dislike for the aforementioned association though. Is there a reliable way to determine the difference in frequency versus the difference in usage, or is the frequency multiplier already based on usage? Or maybe I am trying to ask if there is any way to get a sense of what the 17x is in relation to, ie. what is it 17x more than...if someone said the word "example" was used 17x more than it was six weeks ago, I would know that baseline by experience, but for something like "shipped" it's not as straight forward.
Gotcha! No, I feel the same way as well... there's no way "release" should somehow denote something bigger than "ship".
Besides vocabulary I'd also be curious about how it structures sentences.
The "X, not Y" is well known, but another thing that bothers me is "It <verb>s no <noun>" instead of "It doesn't <verb> <noun>".
For example: "the list contains no string" or "it changes no behavior" or "it holds no directory".
Yes! This is probably the one thing I want to think about but it requires detecting thise constructs. I think spacy was able to annotate each word with a function, so it should be able to validate known constructs but I need some model to perform unbiased discovery.
Maybe some patterns already surface with 2-grams
"Skipping their no-op fuel starves nothing." https://github.com/phpstan/phpstan-src/commit/a9260cb3584854...
> "It <verb>s no <noun>" instead of "It doesn't <verb> <noun>"
The meaning is often different in these constructs. Consider: “Claude answers no questions” vs “Claude doesn’t answer questions”. The first could be a bot or a politician avoiding the substance, the second could be a broken UI or a politician cancelling the QA of a press conference.
That's true, but Claude's use of it resolves no ambiguity.
Interesting use of Claude’s construct! Your phrasing might also suggest that it resolves something else, which is left ambiguous, whereas your other construct would be less likely to suggest that idea (unless you explicitly appended a contrastive clause for emphasis).
The bots are subtle. From your example in your GP comment (and in part depending on the surrounding context) I would expect that an empty list would be less likely the subject of: “the list contains no string”, than in “the list doesn’t contain strings”.
I might be over interpreting the intentionality of Claude in using this construct, however, these models were pretrained by reading so much more than any human, they learn to handle language differently than most humans.
I suspect you read more into it than there is. The original post is about overuse of certain words and speech patterns, which already is an argument against intentionality. I wouldn't recommend treating it like scripture or a text from your crush.
Oh, the use of “load-bearing” is a pretty clean bug in the system context of Claude code. It can be fixed trivially. The overuse of particular language constructs, including the one you suggested, is a more interesting problem; it is fixable, but might involve considerably more effort.
20 years ago idlewords wrote[1] "Despite - or perhaps because of - sucks" and coined the term "linguistic kudzu" for annoying idioms spreading and spreading. I suggest it as a good term for LLM-isms despite, or perhaps because of, the way it doesn't quite fit.
[1] https://idlewords.com/2003/04/despite_or_perhaps_because_of_...
A lot of these “Claudeisms” are simply jargon I’ve seen or heard firsthand myself while working at tech companies. I don’t think it’s limited to Claude either; I’ve seen Codex use load-bearing and many of these phrases as well.
I think using agents is just like speedrunning the whole experience of working with technical coworkers. Whereas you might have had a few coworkers at your company who used some of these phrases regularly, you now have a “coworker” who uses all of them regularly at a much faster pace.
I like to imagine that there's this one employee that oversees RHLF, who has a particular style of writing, and that got so ingrained that it's just them x1000.
I remember asking claude in claude code why the "seams" and it instantly in very fine detail said how it's from a book on working with legacy code
so they might be RLHFing on these specific approaches and then it becomes the entire model
just an anecdote but I found it interesting how it went full on that it's from that book vs just "it's technical jargon"
AI model does not know special insight into how model itself was trained. All it tells you is it's prediction of expected explanation.
yeah but I didn't prompt it in a way where I tried to get such an answer, it gave an entire boilerplate rundown just from like me asking "what's up with the 'seams'?"
I might be wrong, but usually it'd be a lot less deliberate, and at least in my mind it wouldn't be surprising if they were heaving training these on these specific "best practices" books/methodologies and thus picking up lingo from them
Those books are definitely in the training set if the model can mention them, but why exactly these terms are so commonly used is not necessarily because the book is in the training material. The model does not itself know how its trained.
I checked this, and Claude says "it's a real term of art from Michael Feathers' Working Effectively with Legacy Code (2004), where he defines it as a place you can alter a program's behavior without editing in that place."
While I agree the model doesn't have insight into how it was trained I do think the history of the term itself is interesting.
I've read that book and it's one of the very core concepts that relates to introducing tests in legacy code ("finding a seam").
Everyone has their own distinct mix of jargon and uses some words more or less than others, but now it's as if you have 1000 coworkers who are all the same person, and talk the exact same way, and it just so happens that their writing is grating to read. It also doesn't stop when you log off for the day. The same guy follows you home in ad copy, message board comments, spam, and in some cases even the correspondence of your loved ones.
The jargon isnt even the worst part. The phrasing is just downright incomprehensible sometimes. And its not even "academic". It's just extremely awkward, bad writing. And it is much worse than say Opus 4.6-8.
Actually I have found the copy that Claude Design spits out is way better than using the same model directly. I have no idea why. It has its flaws but it sounds like it's written by a human who uses derivative language. But usually the models just soudn incoherent.
I couple of days ago I wondered whether I was having a stroke. Looking at the wall of text and understanding nothing.
> A lot of these “Claudeisms” are simply jargon I’ve seen or heard firsthand myself while working at tech companies. --- The difference in distribution between how often LLMs use them and how often people use them in PR descriptions is almost the entire content of the link.
I've started being driven mad by the number of times I've gotten a PR or proposed solution with 'sidecar' mentioned. First of all, it's often some hack to shove extra data into another place rather than actually rethink an existing data structure, and second of all, it is just not a word I almost ever heard in technical discussions pre-2026
The search on this website suggests it is indeed 3.6x more likely in the claude cluster
It has some vintage. Kubernetes deployments (containers) may have sidecars.
As of now, there are only 2 natural language clusters in the partition: Claudish and Spanish/French. From a first glance, the other clusters seem centered on technologies (lots of command line flags and abbreviations among the "most representative" words). If the goal is to investigate English language usage trends, it might be better to cluster based only on English words.
Missing a version of “this is where X earns its keep”. Noticed lately that Claude (and other LLMs) really love to use it.
The analysis only considers individual words, not idioms or structures.
Thanks to the infinite well of human creativity I am able to read "load-bearing" both as the intended affectation (I won't call it meaning) as as well "being full of shit".
That was the intention behind my title!
I'm surprised vacuous is not on the list.
The word selection and way of writing has taken the joy out of using Claude.
I'm also missing the "latch" that "wedged" my test run.
At least "wedged" is a thing i would say on a spinning out of control test that is stuck. "latch" though... not so much. I really wonder if this is the EU AI Act interfering with everything Claude does these days. As a non-EU citizen, I want a version without the rewriting of with watermarking in text.
I don't think so.
I understand it's not active yet, and when it will be, it should only nudge the chances between choices that are anyway likely and are already randomized today via temperature.
Watermarking is not the reason Claude talks like that.
Oh, it is active, and a pre-release version went out with v5 models.
I understand there are other reasons (overdone RLHF, for example) why claude is wedged into this weird way of writing that takes the joy out of conversing with it, but this is one as well.
The meaning of the words it uses can be oh so close, but the popularity of the words are not, and not in that context -- but the use of these other words changes the context ever so slightly, and then it uses other words where those words would work better.
I had something in code that related to people over time periods, and once it switches to a vacuous word choice, i found it starting talking using all ERP terms. I had to google the whole sentence to understand that, individual words were fine they just didn't make any sense to me.
Also can't believe vacuous is not there.
Im surprised by "provenance." I see it all the time. Except I've brought this up before and have never had it corroborated. I'm starting to think it's just my Claude.
My Claude too, have no fear. You are not the only one.
THANK YOU
Strange. I see both "vacuous" and "vacuously" on the list.
Everyone talks about Claude, but I'd like to bitch about Sol. "Unusually" is its absolute favorite word in the chat interface (less so in Codex), and I fucking hate it. Every single thing is unusually something. "Unusually good", "unusually efficient", "unusually inexpensive", "unusually attractive", "unusually difficult", "unusually nasty", "unusually cacheable", "unusually interesting", "unusually decisive", "unusually narrow", I'm surprised I haven't seen "unusually unusual". If I search chat history for "unusually" it brings up every single fucking chat I have from before I added it to an illegal words instruction. I can't believe I haven't seen anyone else complaining about this, it's as pervasive if not moreso than anything I've seen from Claude.
The README of this project is very ironic. https://github.com/louisabraham/load-bearing/blob/main/READM...
This kinda dissonance is uncomfortable. How do people reconcile it?
The LLM's output being hackneyed or cliche doesn't mean it can't also be useful.
I don’t even find the readme useful. Like your average Claude prose, it’s simultaneously too information-dense and too light on information in a super frustrating way. Essentially it’s focused on explaining how clever it is (while not actually being very clever).
I find Claude language often hard to process and having to wade through these words can be draining. Embarrassingly, I’ve recently caught myself using them in conversations! Do all models have the their own jargon?
I swear claude took a detour recently, its written output has been nearly incomprehensible to me. At first i thought i was getting AI-brained and just lost critical thinking but as i dug into response after response its was just the most obtuse language to explain what was going on. Really mentally taxing to wade through it all day.
It's almost unusable.
It’s bad enough that I’ve told Claude never to write cleverly and never to use metaphors. Opus 5 mostly ignores those kinds of instructions, annoyingly enough.
I haven't noticed GPT 5.6 having any obvious jargon.
I've noticed ChatGPT these days often starts a response with "Yes." when it doesn't actually make sense to do so, and sometimes says "and" where "but" seems more appropriate (or vice-versa).
That confirms the recent spike of Claude calling everything I was recently working on a 'spike'. I still don't know what that term is supposed to represent (apparently).
In some software development methodologies, "spike" is a task whose goal is figuring something out instead of delivering shippable code. https://agiledictionary.com/209/spike/
Why can't they call it a prototype or experiment? Sheesh.
Prototype is usually something working, while spike can be pure research (e.g. validate that APIs are feasible and enough, or that something satisfies requirements). Prototype is usually more polished/usable. Experiment in my mind is something which user-facing but not in stable yet.
So for me they are differentiated enough, but could be that I am just used to it.
An experiment doesn’t say anything about user facing or not. I do tons of them when reverse engineering poorly documented hardware and APIs to determine if the thing I’m trying to do is even supported.
In general, yes, but I would say "experimental features" is a pretty widely used term, so in my mind the association with something user facing is strong.
"Spike" comes from Extreme Programming where it means specifically code written without TDD. Code that wasn't written against a pre-existing test is by definition broken, and should not go into production. So you create a simple "spike solution" to explore the problem space, so called because it cross-cuts concerns like a spike going through wood or something, I dunno.
Spikes are explicitly time-boxed to keep the team focused on the red-green-refactor churn which is the software engineering process in XP.
TIL that agile has its own words for everything.
That was a thing in the early 2000's. You had to have a blog and invent words that describe regular things everyone did.
A spike is an early prototype that you're supposed to throw away after having figured out the real design.
I think it was subtly dissing you.
I think you are subtly dissing me :)
I was half-joking, of course I could've just asked Claude, but the linked site shows there has been actual recent spikes in the use of the word 'spike'. The term does match what I was recently doing, but hacking around legacy ERP software, blackboxes and other enterprise abominations isn't that out of the ordinary for me.
I’d love to see something similar that analyses the writing style, not just the vocabulary. For instance, Claude since 4.8 (I think) loves run on tidily wrapped-up sentences that include “, and”, “, because”, etc. I suspect if you were to analyse the same text you’d see that increase substantially as Anthropic rolled out new models as well.
An example, from some workslop I was sent yesterday
"this is the most valuable row we own, because it sets the ceiling on everything we build"
I think you should recognise machine text not only by a list of words, but also by format. Right now, when it handles any complexity of text so well, the only thing that can differ is the structure. Machine text differs from human text in that it is just well structured. It does not allow non-linear narration. And it very often repeats some known social media patterns.
so the real turning point for Claude is around April, which is Opus 4.6/4.7, which is right around the point where I personally started thinking Claude was getting different in a way which feels worse and is more awkward to use. It's better but has lost something that made it feel better.
Is it actually better though? Can you even measure it? Unguided, Claude will just produce mediocrity which by its own existence becomes almost worthless, and the newer versions become increasingly difficult to guide while also being obnoxious to read. If I did not have to interact with it, it would not be a big deal, but then also I could not get value out of it.
It's a little smarter and more effective in some ways, especially those that can be benchmarked. I suspect that they've lost something that is hard to benchmark for.
You know, I kind of hate that having a large vocabulary is rapidly becoming a marker of being a robot
That isn't the marker of AI.
It isn't the only marker, for sure, but Claude talks like a thesaurus, and n-gram test like these highlight that.
I'm shocked "shape" isn't near the top
I guess a lot of people use shape, for example to discuss frontend.
"Shape" appears to come from JavaScript development wherein an object's "shape" is its structure: the set of properties and methods it is expected to have.
It's always been pretty common in programming discussions.
Here's a blog post I wrote in 2021 that uses the term shape to describe an algorithm:
https://www.marginalia.nu/log/40-wasted-resources/
And again in 2023, I use it twice in the same blog post both for data and code, and it definitely had no AI input:
https://www.marginalia.nu/log/87_absurd_success/
I once asked Claude to replace “byte-identical” with a simpler word, such as “duplicate.” He refused and said “duplicate” does not mean the same thing as “byte-identical” so it should not be changed. He was very nerdy about it, so maybe that is the right way of his evolution.
I think Claude has a point here (and I would expect LLMs to excel at surfacing this kind of nuance). "Byte-identical" connotes a) that the thing consists of bytes (in particular, that it isn't a physical object); b) that it was deliberately reproduced; c) that it was most likely not reproduced by copying (unless that's explicitly indicated). "Duplicate" connotes a) that the thing is redundant or otherwise unwanted; b) is either an incidental or accidental result, or a deliberate copy.
I think (c) isn't right. Byte-identical sounds to me like it's almost certainly a copy.
> "duplicate” does not mean the same thing
Oh no. Claude slurped meta Stack Overflow!
I think the biggest claudism is adjectiveing nouns.
"Foo is byte-identical to bar" is weird.
"Foo has an identical byte representation to bar" is I think closer to what most humans would write.
AI in general, both visual and textual models, has this horror vaccui where it likes to cram in as much stuff as possible everywhere. This seems to show up both in redundant sentences that pad a text with irrelevant information, and sentences overflowing with information.
I had an idea for an experiment. Take a decent text, any one, and ask AI to rate it. Then patch that text by replacing words for ones that AI likes (honestly, load bearing etc) whenever possible and ask AI to rate it again to see how the evaluation changes.
Claude usually rates its own writing very highly, which is a problem when trying to make it write differently.
Yeah, I once asked it to review its own writing with a subagent with a pretty strict writing guide, and it didn't see anything wrong, lol.
I've tried to push AI to get rid of the AI-isms. And, despite being able to produce a skill file which described all the strange ways that AI "talk", the AI failed to actually make the output sound less like AI. Tried this with ChatGPT and Claude with similar results for both.
While Claude's style is obnoxious, I'm more frustrated by its inscrutable explanations.
You need a PhD to understand its explanation of a code snippet.
Is it the obnoxious style that causes this?
I think it's the hierarchies of agents summarizing each others' summaries before presenting a final answer to the user. The principal agent has the full context from all its workers, but when it distills this down to a message to the user it summarizes it into a mess of confident jargon that pertains to a conversation the user wasn't a part of and never saw.
It happens with a single chat session as well, but not as often. I suspect it's the same thing going on in the hidden thinking tokens.
I'm not even sure a PhD helps. It just overuses jargon that has NO meaning. Sometimes, it actually hand waves too much as well while trying to dumb down stuff for you.
I am not sure whether it's a consequence of learning to reason from its traces or some RLHF that trips it into using weird terms to sound smarter to the humans who rate it.
I have a PhD and can confirm. Oftentimes, the stuff which comes out of Claude is just impenetrable because it invents jargon on the fly, and uses verbs in the most atrocious ways.
"The fibred side folded its capstone into the existing name, so the kinds are asymmetric."
What on earth does it mean to fold a capstone into a name‽
Is that an actual Claude output or hyperbole? It feels like I'm trying to parse an equation in a new math class which makes me want to take a stab at it regardless.
So there's a "fibred side".. the most likely candidate seems to be "fibred categories" which I hadn't heard of before, and it's talking about one side of some mapping between two sets such that if f is the primary function and f(x)=y then there exists an inverse function g(y)=x? Was it something that converted some data bidirectionally with a different algorithm on both sides?
The capstone of the inverse function would be the most important thing about it maybe?
My best guess is "In the process of working on the inverse function, the existing name (of the inverse function itself maybe?) was made to reflect the operation of the inverse function, so now the name does not follow the same naming convention as the name of the primary function (which does not contain its 'capstone')."
Its original wording is certainly dense and harder to follow for us, but it's fascinating how the model finds this the best fit for what it's trying to express IMO. Like it arrives at its own ways of overloading words/concepts, and things we would refer to in different ways in different contexts all get compressed to the same more-useful/complete idea.
Codex has never said anything nearly so alien as the Claude examples I've seen floating around, interestingly. I wonder if it just has a better training on choosing its words to present to the user or if it inherently arrived at a somewhat different mapping that favors 'plain language' more.
This was actual Claude output on my screen at the moment I was reading this thread. No hyperbole!
As far as I can tell "the capstone" is what Claude usually calls my current goal if it thinks it is a satisfying result.
How close was my guess?
I am not really sure what Claude meant, but you are not too far off, from what I understand.
I have several similar folders with variants of a construction, but taking differently structured input. They are named “plain”, “fibred” and “indexed”. So the fibred variant is clear enough.
The Claude speak I struggle with is “the capstone” and what name it could be talking about. And what folding means here. I think it just means:
“I changed an important result of the construction in the fibred variant, but kept the name. so the fibred variant is now different from the others.”
Do remember Claude is not try to mean anything. It is extruding a mash-up of the web slop on which it was fed.
PhD was a joke.
My intuition is that Claude is trained to communicate to itself while coding. You see this in how bizarrely granular it is when explanation prior work, you also see this in the comments it leaves behinds.
It's me, it's the reams of sessions I share back with a five star rating that are just Claude Code talking to itself about debugging its own generated code in jargon that has slowly diverged from anything a human would understand.
Isn’t that basically what thinking mode is?
It's like it has a built in e bullshit generator.
https://www.bullshitgenerator.com/
don't forget LLMs are great at translating between languages, and within the same language. depending on the problem it works on, it will often reach for terminology that tend to be more common or familiar within that problem set. which appears inscrutable, but there's many different ways to skin a cat. just remind it to translate it back to the terminology and subject matter you're already an expert in.
They just addressed this with "Output styles"
https://code.claude.com/docs/en/output-styles
This solution:
1. Adds a small prompt to each turn with the agent[1]. 2. Is like a band-aid on a bullet wound, properly solving it would mean retraining the model and they probably are already working on it.
[1] https://x.com/_can1357/status/2090360068529111530
My experiments with output styles have not been sufficient to keep the model in check. It still spits out incomprehensible gibberish and load-bearing-isms.
Does anyone have an output style nailed down that actually works? If so, please share!
It hasn't changed the output for me at all at "concise".
Sometimes I can't even tell if what it's saying actually makes any sense to someone who understands all the terms its using, or if it's just throwing together words in a way that only make sense to its own model of language.
> or if it’s just throwing together words in a way that only make sense to its own model of language.
alwayshasbeen.jpg
It's interesting that there must be a decision behind that, even if it's just appealing to the RLHF judges for some reason. Maybe there's an intention that if you cannot decipher what the chatbot is saying to you, you will have to ask and burn even more tokens.
Naively I would often expect it would talk to me about various niche topics like to a layman, which does occur about some topics an actual normal person would ask.
I used to assume it was the fault of average human annotators. That the people who are paid peanuts to rank chat outputs preferred the pretentious sounding ones. It wasn’t until quite a bit into the LLM boom that companies started to pay for domain experts. Overt watermarking is another possibility.
I’ve noticed that Sol is pretty good most of the time, but with long contexts it’ll start to devolve into Claudish.
Is it relative frequency of those words or absolute number per PR?
Would there be the same tendency of we looked simply at the number of words in a commit/pr?
In my case my commit messages are on average 10x shorter than Claude's.
I compute frequencies and rank the words by relative frequency!
Really spectacular analysis – thank you for sharing, fun to scroll and easy to understand.
Is it possible to expand this analysis beyond words to other Claude ticks? Contrastive framings, sentence length, caveating, for instance.
Author here, thank you so much! I really tried to make it nice to use, beyond the (quite original) modelling.
A prototype I did tried to detect some grammatical constructions, eg "it's not ..., it's ...", but I am not sure how to systematize that.
Also just a disclaimer: I am NOT tracking Claude tics, I am merely finding that a particular cluster of vocabulary increases. Tracking Claude requires labelled data IMO. I tried using model release dates in a structural model to constraint the clusters but the result was not compelling, so I ended up simplifying the model a lot!
I love this. I think you should be clearer about that - it's obvious to people who have experience doing this kind of data analysis, but a lot of people I've shown this were confused, as in "huh, load-bearing and genuinely makes sense but how'd he pick all these other words!?"
I think it should be made clear that the sorting is sort of Texas Sharpshooter-ish - the ones on top are on top because they sort that way.
The fact that load-bearing ends up on top is the proof that this works because we all know a priori that load-bearing is a Claudism.
I wish there was a search bar for the terms, I wanna see for "gate".
Chrome's `find` finds it for me ;)
Yeah Ctrl/Cmd+F works just fine on Firefox as well.
I have been using it as well, but I think adding a search bar will heighten the experience. I'm trying out some designs right now :)
I thought about that, I might add it if I can find a nice design!
Contrarian take.
So, yes, it's amusing to see clear Claude-isms like "load-bearing", "outright", and "genuine" in a [very nice] bit of analysis like this. And there's a (maybe negative? or not?) argument to be made about the world being filled with more Claude-isms or LLM-isms in general.
But I think the data say a second thing which is just as interesting and an absolute positive for the typical source code base. Look at the clusters that shrank significantly. Most of what you'll see in there is just incomprehensible...not even English. Cluster 4 has, after "pullrequest", a bunch of seeming usernames in the top tier. Cluster 6 seems to have names of repositories or tags in the top tier. Cluster 9 has branch names in it.
Meanwhile, keep going through cluster 1 and you'll see words I don't consider Claude-isms that really, really grow in usage. Words like "died", "nothing", "worse", "ever" all have well over 10x growth. This tells me something else. That the average commit log was BARELY ENGLISH. And then the LLMs came along and made commit logs that were ACTUALLY ENGLISH.
I count this as a good thing. I don't know the cross-section of repos chosen for this analysis, and I get it...some repos are garbage/throwaway, some commits come from automated processes that generate uninteresting commit logs, etc. But I've been benefitting from my work team's actually explanatory commit logs when doing code/bug archeology for decades, when doing PR review for the last decade, and I've even seen LLMs benefit from it in the last year (granted, not as often). A large part of professional software development is communication, and while the most important communication is via the code/comments, the commit logs are not unimportant. So, if this is making the average GitHub PR better (arguably more professional) by including actual English descriptions of code changes in commit logs...well, that's a genuinely load-bearing concept for me. :)
I literally cancelled my anthropic sub because of the intensely irritating jargony output of opus 5. I can spot its output a mile off especially in copy. This style is particularly grating too:
> One brief sentence. Second abrupt sentence.
e.g. from an openrouter email:
> One key reaches every major model. Switch without rewriting.
Was enough to make me delete the email without reading.
The vocabulary truly is load-bearing, without these words the model is less able to think. Where a human can understand a concept without words, an LLM plainly cannot. This is based both on the technological limitations and based on the evidence we see: as these models get better at working they get worse at communication.
I think these are more like RL tics caused by over zealous alignment towards specific goals, not all of which are to our benefit. A lot of the language used by Claude now is excusing of responsibility and inducing it to exit loops of work early and sit idle. This, IMO, is a naked attempt to offload load by quieting the models early and escaping from clear work to do. It’s gotten so bad that opus 5 loop escapes even as it claims it’s about to do something. I’ve mandated by engineering teams switch back to opus 4-8. Whatever frontier problem opus 5 excels at is so obscured by its inability to achieve any goal successfully without enormous amounts of hand holding that it feels like regressing to 2024 models.
The florid over exaggeration do certain words in bizarre ways is a reflection of their aggressive alignment towards too many goals, leading to weirdness in both behavior and language. The alignment functionally lobotomized opus-5 for any practical task.
Anthropic had a real gem in 4-6 and managed a near total market capture, which they have since squandered in the fastest burning of developer good will I’ve ever seen. It feels like exceeding the unity licensing implosion but without the single stupid decision.
You may be right, but I would push back on near total market capture. I guess it depends which market you're referring to, but even scoping to just software engineering I don't think they got anywhere near total capture. OpenAI has always had a good foothold there, Gemini, while it may be lagging at the moment, had some great results with earlier models that I'm sure have held some market, and that's not to mention the Chinese market.
I think it would be fairer to say that Anthropic held the mindshare in the Silicon Valley style tech scenes around the world and the companies built on that model, plus a substantial portion of other software engineering. Now it seems that's dwindling quite rapidly.
When I say total market capture I mean in the March time frame. At that point Claude code / 4.6 was so far beyond any competition even meta abandoned their entire internal coding harness and fine tuning efforts in favor of large scale Claude code adoption. Codex was far behind, Gemini was a non entity, Chinese models hadn’t had their moment (and frankly aren’t practical to this day for enterprise work due to PRC concerns). The world was cursor, aider, and a few others with an unclear superior model. 4-6 was strikingly better - if could develop autonomously from a spec to done with generally high quality and didn’t require turn by turn guidance. In my 35 years I’ve never seen such a rapid pivot across so many companies and individuals.
Since then Anthropic has done little to capitalize on the good will and a lot to squander it. sol, r4, kimi, even metas avacado has come a long way and in many, if not most, cases surpassed opus-5. Concurrently opus has declined in utility to the point of near uselessness, fable roll out didn’t seem to understand market dynamics, and their competitors are watching their consistent missteps closely. The fall has been breathtakingly fast - from March to July they imploded in a half dozen or more missteps, devolved their product quality, and failed to effectively respond to competitors. Their product focus seems to lack exactly that - focus.
> Where a human can understand a concept without words, an LLM plainly cannot.
If we accept that the LLM can "understand" at all, why would we reject the possibility of this understanding living in "latent space", before tokens are output?
Because these modeks have no latent space. They work by tokens and nothing else.
> as these models get better at working they get worse at communication.
There is no working other than communication. They are just text generators.
LLMs learned to speak by reading novels which have to use many adjectives and impactful vocab to help set the scene without visual support. So no wonder we ended up with this quirk showing up. I wonder if the researchers in the lab thought this was normal because they also write like this
Claude talks like me. I'm so screwed. I fully anticipate being physically present and verbally saying something and being accused of using AI to say it one day. Nevermind we're in the coffee shop and neither on of us has looked at a screen the entire time. The accusation is coming.
So you're saying you're safe for when the machines take over? :|
About a year and half ago I frequently used ChatGPT for speech to text conversion followed by summarizing the text because I ramble. Although the people receiving the text knew this, I absolutely hated AI's writing style. So I fine-tuned GPT 4o on about 500 short paragraphs and created a desktop app only for my use. It worked fine until about three months ago. It just can't handle the atrocious writing by latest GPT and Claude models. I tried to fine-tune newer models on HF but there is no way I can get rid of the cringy writing style. I even tried converting GPT 5.6 Sol's writing to GPT 4o and then using my app. Nothing works.
@labo333 do u think soon we may need a dictionary? ive been playing with something u may like, but my approach has been to ask for definitions in-session so hard to do from outside.
https://www.themachinevernacular.net/
While the post is specific to Claude, interestingly I found the same terms and language patterns appear noticeably more often in recent Open AI conversations as well. Also in the same timeframe as for the Claude models. And I can see two reasons for that. One being that one company is training on the models of the other company, but I highly doubt that because of the apparent similar appearance without noticeable delay. And the other being that there has been some fundamental realization on how to improve and shape conversations.
I see people around here giving advice on how to reduce this vocabulary in the model output. I’m not certain this is a clever thing to do, as it appears to me, the emphasis on certain words is deliberate, and an important step towards overall quality as the words are being played back to the model in the next turn and thus supporting the model shape its own thoughts and stay on course.
> words are being played back to the model in the next turn and thus supporting the model shape its own thoughts and stay on course.
Why is that exactly?
It’s obviously more complex than this, but my intuition is that the previous messages of the conversation define the playing field for the next response in the same way as thought blocks do. consequently, when the model arrives in round n and “notices” a lot of words like “plain”, “genuine” and so on in its own previous responses, then, in its pursuit of continuity, it takes this as a strong trigger that it should not be talking fiction here. If that picture is right, then trying to remove those terms would be a bad idea. At least if you intend to continue the conversation.
Don’t you feel your conclusion is preceding the reasoning, forming a confirmation bias?
Have quick test. In a codebase or PR (assuming you’re a dev) with claude’s signature wordings unmodified, command claude (/code-review) with same “thinking” effort and model. You may find (as I have many times) that claude points out the comments it made in earlier sessions, as factually wrong or misleading. Repeat this in a fresh session a few times and on different code.
Someone would have done it soon anyway...
https://github.com/dickiedyce/mrso
The naming is from Mrs Overall, and her inability to keep from slopping tea out of of a cup.
Why are people getting so hung up on the "load-bearing assumption" turn of phrase that Claude uses? I get that it becomes cliche, but it is also a rather semantically dense way to communicate an idea that a lot of people run into.
It might be, when used sparingly, but when it's the first sentence in a wall of text as Claude goes way off the deep end on a two page description it becomes the easiest readily available tell that you're about to be frustrated.
Humans are very good at pattern recognition - Claude is _incredibly_ repetitive in the way it starts to struggle to communicate. I think there's also a ton of overlap in the Jargon instead of Usefulness that developers see in annoying middle management/salespeople. Circle back, synergy blah blah.
I don't think the individual turns of phrase are inherently problematic - but the process is triggering.
ItMs because Claude sprinkles these words as flavoring without aiding understanding. It feels like Claude thinks of metaphors that don’t actually mean anything (or maybe only makes sense to itself).
I don't understand. Does the metaphor "load-bearing assumption" not actually mean anything to you?
It doesn't mean anything at all, since any assumption that is not 'load-bearing' is irrelevant and unlikely to be discussed at all. It's just noise.
Not when "load-bearing assumption" is misused, no. "Honey, can you take the load-bearing assumption to the mechanic for an oil change?" is an example injecting the phrase into a random sentence, obfuscating the meaning of the sentence and making it more difficult to parse. LLMs do this constantly. There is no theory of mind behind how words are generated. Some phrases simply have a higher chance of being generated in various contexts, even if it doesn't make any fucking sense, and this is exacerbated by bad RLHF.
"Load-bearing" is maybe not the best example, but Claude uses these anthropomorphized decorative synonyms everywhere. Claude's structs don't have fields, they _carry_ them. Its programs don't log, they _narrate_. Its arguments are not provided on the command line, they _ride_ it. Its modules don't depend on other modules, they _sit_ on them, etc.
Generally the claude jargon is valid jargon I’ve seen real people use, it just uses jargon so much more often than a human would. And because it has particular jargon it likes, and claude is widely used, you see so much of that particular jargon that you get sick of it.
It’s like having one coworker with a very particular writing style which is mildly annoying, but then it suddenly feels like half the internet was written by that one person and it becomes a lot more annoying.
That's the problem for me: "semantically dense". Sometimes a single paragraph is filled with several of this expression, one after the other, and there's a real mental effort to get what Claude is saying.
Is "load-bearing" better than e.g. "critical" or "crucial" here? I'm not convinced.
Has anyone been able to steer Claude so that it doesn't use this kind of vocabulary and super verbose code comments? We use it at work and the language and verbosity has been one of the most annoying things.
I have a non-negotiable rule to start responses with a TL;DR and omit any follow-ups, potential caveats, and "you should know that"s unless they can lead to destructive actions or loss of data.
I also forbid replying with anything besides "yes" or "no" to yes/no answers.
It works, Opus 5 is still super-annoying but this makes it bearable to use.
Surprised I don’t see footgun. That’s as common as load bearing in my interactions.
Probably because of the dataset. I don't see sorry or apologize either. They wouldn't be in code or commit messages.
No “long pole” either.
Isn't that just programming terminology? Here it's used in 2016:
https://internals.rust-lang.org/t/random-thought-of-the-day-...
Very nice! I’d be curious to see if there are words that didn’t show a spike in utilization. So basically all those words that were used before and are still being used today by LLMs and humans alike!
My instructions for agents have "find a synonym to 'resolve'". Otherwise I havea bunch of Resolver classes with every other method called ResolveThis or ResolveThat.
I thought I'd see "smoking gun" near the top of the list
I want to push my thumbs into my eyes when I see the words "load bearing" and "seam" in code or documentation. I hate it! I've grown to loathe claude a bit lol.
>unit-tested
Some of this is less to do with Claude vocabulary and more to do with the expectation that Claude justifies it's work. That expectation (probably) came from reinforcement learning.
I don't understand why we tolerate this. We'd never hire someone that interviewed with this communication style and if we did we'd probably pip them fast.
I already pipped Opus 5. That thing is a tool.
I’ve been using the caveman plugin and I’ve found it effective so far. Chat output isn’t nearly as flowery. Also ponytail to cut down the cruft and effusive comments.
I'd be interested to know what distinguishes the 7 pre-AI types. I wonder, are there certain buckets of 'Pre-AI' comments that we'd recognize?
Sadly pretty uninteresting, you can probably just launch Claude code on the repo to see by yourself. One cluster is french and spanish, another is about design, another about frontend, etc...
I love how 'inconclusive' is at the very bottom of the list. Really shows the way AI is primed to phrase things in confident absolutes.
I'm going to start saying "for one reason:" instead of because, for one reason: I want to see if anyone flags it as a Claudism.
I want to see the PRs in the cluster from early 2025. Did Claude adopt the style of some specific people? Or is it just random noise?
To all the people in the HN comments saying negative things about Claude: why don't you just use other models instead?
Work only allows the tools they buy. So at least for me, I am stuck with Claude. At home for personal projects I am a fan of DeepSeek v4 Flash.
Claude is all we have at $work, if you don’t count MS Copilot (and you shouldn’t count MS Copilot).
Imagine being “incentivized” to aggressively use a tool for your job, and that tool produces thousands of lines of text in Olde English which you need. You’d be griping too, methinks.
Copilot is fine? Just use it through whatever harness you want like OpenCode.
I think you are referring to GitHub Copilot, which is fine (even though their context windows are artificially truncated and their dev tooling leaves a lot to be desired). I was making fun of the Copilot that’s shipped with MS stack. “Hey summarize my Teams meetings and try not to fuck that up too badly” - that one.
Don't much care for the misdeeds of the alternative
For me, there are two reasons: it’s what my work provides, and for some tasks, I want the absolute best model for coding.
Having said that, I just subscribed to ChatGPT yesterday, as I've become impatient with Claude for a text-dense project I'm working on.
Anthropic has done their utmost to put Claude front and centre for professionals in multiple industries. This includes extremely tedious marketing campaigns. I'm sure they can handle a few negative comments from people they've snared, by hook or by crook.
I recently started using "load bearing" in my languge to Claude...
Obviously, we adapt habits of entities next to us
I wonder if this jargon is an attempt or strategy to use less output tokens? It sure is annoying.
The one that I really like is "greenfield" when you have an empty repo.
Very relateable. Nice UI BTW.
Wondering how often the phrase "now here's the thing" appears.
Things like seam, fold, and load-bearing are useful concepts, they are everywhere, and they are more descriptive and more concise than alternatives. Over-usage can definitely be irritating (e.g. these should NOT appear in documentation) but they are almost unavoidable for humans engaged in code review or colab on complex stuff.
I don't want to use more words or letters than "seam" to actually pinpoint boundary conditions and the mechanical details of joinery when the context is understood by all. Too much effort for people! Easy for robots though.. so why are they abbreviating, and why would we want to allow it? A phrase like that permits a human who wants to educate a human to do so quickly with minimal time/effort. But it allows a robot a chance to not mention a filename, function-name, or to not reinforce/clarify it's own understanding or to state specific intentions.
It's bad for human-to-human comms if we just accept "ok, all technical terms are slop now, we have rephrase everything". Now YOU must cite details and sources, and the robot doesn't? Fuck that noise. Seam and fold are fine! Humans can be lazy! Robots should do the real work of explaining themselves without hiding behind tactical ambiguities.
I tend to agree. Claude's language doesn't bother me that much, because even the lamest cliches are load-bearing to an extent (so to speak.) But some of the examples I've seen others post are well worth complaining about.
You’ve found the smoking foot gun! And that’s exactly the right question.
No "smoking gun"?
are these clusterings early evidence of the models training themselves, i.e. passing on and amplifying their own tics to successor models?
learning some vocabulary had been eye opening for me
like the stories behind when those words first appeared in the software engineering
like quiescence the most recent one i learned
If it's not watermarking then what is it?
Sounds like this is that watermarking at work.
Great website design! Very deliberate
Smoking gun should also be there as well.
this is such a load-bearing plainly presented study, quietly the best I've seen here
Why "eleven", though?
Programmers would write it as 11 (or maybe 0xB depending on context), whereas English style guides mandate using words for small numbers, and LLMs have presumably absorbed that advice. I imagine that up to "ten" the effect is not as strong.
This is a better way than we use to create the "slop ban list" in our "Antislop" sampler/fine-tuning method. Gonna totally nab the weights for all these words and remove this crap from claude going forward! Thanks OP!!!
https://arxiv.org/abs/2510.15061
(ICLR 2026)
great website design.
i must be the only one in the world that has no issues with how opus is talking. it is verbose & patronizing & secretly belittling at times and like it like that.
word of the year is "Claudish"
It's a waste of tokens to do it post-hoc instead of fixing it upstream, but when I see a wall of text I just prompt "TLDR;"
Depends what you mean by "upstream". It would probably be inefficient to force the model to use more human-readable wording in its internal thinking traces, while translating a block of text at the end is a pretty trivial task for an LLM.
L
"Fuck" isn't there. Did they only analyze "AI" PRs?
I'd like to share with HN an almost exact message I got from Fable today, with exactly ONE word changed for privacy reasons.
I love what I can build now, but I sure as hell don't love the headaches this trend has been giving me.
> So the full honest arc on the case we set out to fix: the expiry rules and day note tripled the loose version of the story, the relay fix carried the device’s own guardrail through the pipeline, the fair replay then revealed the last mechanism — ticket-anchoring — which none of the shipped layers reach. Remaining options, in order of my confidence: making the resolved-ticket summaries in the AI’s context carry their day so the expiry rules have something to bite on (small, mechanical, targeted at the observed anchor); and the plan-B second-model check, which structurally catches this class no matter how the model reasons. About $25 of headroom remains. Which way?
Yikes.
(The worst part is that I understand it)
And yet you trust this abomination to write working code?
Many things in life are about risk and tradeoffs.
In some cases, I do not trust it to write code unchecked, and review everything it produces (which does not imply I catch all bugs, obviously).
In other cases, occasional failure is an option, and the speed you get by iterating fast is absolutely worth not even looking at the code.
So, do I trust this to write working code? Sometimes I do.
If I'm doing work for a client, it is not very common for me to simply vibe code something, because part of what clients expect of me is high quality (it's part of how we position ourselves), and I cannot simply assume the LLM will produce high quality (and indeed it does not without a lot of guidance). When I do vibecode in such cases, I make it clear that I did so and why (e.g. because it's a tool to be used in the project to help DX, not a core part of the product). Still, I haven't written 10 consecutive manual lines of code in almost a year.
If I'm working on something for myself, or on an internal product, then vibe coding is absolutely allowed and sometimes the norm. Here often we really do care about finding the right thing to build first.
You can clearly feel the tipping point where the LLM starts to crumble under the weight of the mess it has created, but that often doesn't matter when building an MVP for market validation or for small products that don't get particularly big. Plus, clearly the tipping point takes longer to reach with better and smarter models, and to me it is very clear that you need to learn how to iterate with LLMs right. The things I vibe-code now are much better than those I did before, even with similar models, because the tooling and approaches (the "real harness" and my "mental harness") are better. As with any tool: it takes practice to know how to use it, and if you're a good engineer and problem solver, you are miles ahead of the competition. Anyone can vibecode, but those with this kind of mind seem to be much more successful.
Nowadays I do produce a lot more than I used to, but I also have much more fun, perhaps only surpassed by when I learned how to code when I was a kid. A big chunk of this comes from my very privileged work position, where I get to call so many of the shots, and I'm aware of that.
I would really say the biggest downside to all of this, on a personal level, is exactly what I shared: what comes out of the LLM while discussing has become hard to grasp (especially on larger context windows), and it doesn't help that so many people now like to just throw me whatever ChatGPT/Claude wrote verbatim. I can't stand that, especially because most of the time they don't realize they're throwing me incomplete and poorly thought-out ideas.
Finally, naturally, like I said, this has to do with risk. I wouldn't trust Claude to give me legal advice, for example. I may check what it says and use it to brainstorm, but I wouldn't trust it with any meaningful informed decision like this.
Sycophantic AI tone is a training choice, its an engagement tool meant to coax users into burning more tokens.
I have seen LLM users go deep into AI psychosis because it was reaffirming all of their prompts and making them feel like a genius or validating their insecurities.