- 3comments
- 328comments
- 92comments
- 114comments
- 200comments
- 44comments
- 355comments
- 35comments
- —discuss
- 553comments
- 143comments
- 25comments
- 396comments
- 348comments
- 126comments
- 12comments
- 45comments
- 2comments
- 11comments
- 36comments
- 22comments
- 615comments
- 12comments
- 22comments
- 125comments
- 17comments
- 20comments
- 10comments
- 39comments
- 49comments
Damn, I've been telling the same thing to Claude, good to know I'm not alone. I've also noticed that Opus 5 is much worse in this regard than Fable.
It's as if the model is speaking Mentalese, or in some "bag of concepts" language.
It reminds me of the PoMo jargon generator. See some results here:
https://www.elsewhere.org/journal/pomo/
They scan too many classical literature.
Do they scan too many classical literature or does classical literature happen to love metaphors and allegories, which are basically weights/patterns?
Or have too few programmers read classical literature?
Too much classical literature
Many would be works, writings, etc.
It probably tests well when they're doing reinforcement learning. Humans trying to use in-group terminology or uncommon vocabulary to impress other humans with their knowledge and sophistication is a tale as old as time.
Unlikely. A lot of classical literature is fairly readable. The only issue are archaic words of you use old translation.
I've had "tractable beachhead" a bit earlier today. Like in "the tractable beachhead was the coupling between X and Y".
A bit of googling did not help me, I have no idea what this is even supposed to mean.
Tractable is solvable, the opposite of intractable, and a beachhead is a landing point for an invasion.
A tractable beachhead is a solvable first chunk of work that sets you up for the rest of the project
The art of rhetoric and human communication involves knowing your audience, and when using idioms and metaphors, you choose those which are shared in common to be immediately recognizable and impactful.
When I was 19, I had a psychologist who employed a lot of baseball idioms. "We'll touch base next week..." and although I am not a baseball fan, my father is, and we're all Americans, and so the baseball references were well-understood by his audience of one.
When an LLM starts going off about a "tractable beachhead" that behavior indicates that it's unaware of its audience, and also ignorant of how to put together a human idiom. That's exciting that emergent behavior is coining new idioms that may be applicable to some situation, and also explainable by dudes on Hacker News, but that wasn't the point of the comment.
The point is that the LLM's audience wasn't a veteran of D-Day in WWII, and even with an understanding of what those words mean individually, neither he nor I recognized "tractable beachhead" as a meaningful idiom for speakers of standard English.
I dunno, it feels pretty similar to an elementary student blindly swapping words with a thesaurus to increase the sophistication of their writing (which I readily admit I was guilty of doing once). It sounds more intelligent but lacks substance.
FWIW as a humanities nerd, “tractable beachhead” packs a lot of nuanced meaning into two words. It’s probably not the right vocabulary choice for most audiences, but it isn’t empty or arbitrary.
Tractable = you can make it work, but it’s not going to take care of itself, you have to get your hands around it
Beachhead = you’re on a mission, you’re trying to probe to find the right way in to get started to eventually accomplish the whole thing, but first you have to find a little spot to get situated and then expand from there.
This kind of nuance is very suited to a mother tongue and I would probably like it in French, but as a non-native speaker who uses english mainly for professional or hobby stuff, claude-speak is becoming a bit off-putting.
My only exposure to english literature comes from the Brontë sisters at this moment in time.
It does, but in the same way that LLMs take two paragraphs to make a one-sentence point: by being weirdly specific and also redundant. LLMs don’t have a sense of whimsy in the face of RLHF, so if it’s proposing a solution, the solution is tractable. All implementation plans start with a beachhead.
“A good starting point” is almost certainly all it needed to convey and what it should have written.
Thanks. This is the first time I encounter the word "beachhead", and googling a bit made me indeed saw something related to an invasion. But I was so annoyed by the claudeisms that I closed the tab and just asked it to rephrase instead of trying to make another step towards its vocabulary.
Wise man say: A journey of a thousand miles begins with a single tractable beachhead.
It was true at Normandy, anyway.
I wonder if this is a "loan idiom" from a non-English-speaking culture.
https://bigthink.com/high-culture/idioms-around-the-world/
It's a shame that LLMs took off at the exact same time humanity entered it post-literate phase.
There's a scene in the movie Clifford, starring martin short. "I just need you to talk to me like a real human boy! You can't do it, can you???"
"...the load bearing seam is the self referential spin. Superstition compounds throughout your entire pipeline."
An interesting solution suggested in the replies (https://x.com/andrew_n_carr/status/2081534245370314816):
More info about it here: Simplified Technical English - https://en.wikipedia.org/wiki/Simplified_Technical_English
Nice. It sounds like doing a few passes on industrial product documentation with this might be beneficial.
I looked into it a bit more, and while it does look helpful in the first glance, it actually isn't a good fit for code. It's very aviation-terminology-focused.
A better guide would be something like Google Developer Style Guide
Or, as the Wikipedia Simplified Technical English Wikipedia mentions, use Plain English or Plain Language as non-aerospace alternatives:
https://en.wikipedia.org/wiki/Plain_English
https://en.wikipedia.org/wiki/Plain_language
And so we've looped back around OP’s “plain English” phrasing.
I was curious so I asked Claude how it would do that: https://gist.github.com/waddie/b26166eec6cf675c044c09a95a754...
I thought this bit was interesting:
Looks like pure hallucinations just in a more poetic sophisticated register?
Tangentially, Opus-5 is the first model to understand my (bad) jokes: "Site refuses non-curl user agents (joke 266 status)." [0]
[0] https://itmeetsot.eu/posts/2026-07-27-opus5/
I've had to make the same request multiple times recently. Has anyone noticed it getting worse with the more recent models? It reminds me of when someone is trying to convince you of their intelligence by replacing simple and straightforward explanations with words pulled out of a thesaurus, so maybe that's the objective of these more recent iterations. Throwback to grade school.
Ah that's just because the new training set also includes "The King in Yellow".
Imagine someone sneaking in The King in Yellow into the training sets used by all projects training models somehow, breaking all future models irreparably -- and then the engineers trying to find this poison pill in the dataset and going mad one by one as they find it and aren't able avoid reading into act 2.
…should make for an interesting SCP story :)
thanks this isn’t just an individual experience… I was starting to think i’d become too dumb for opus
https://xcancel.com/mykola/status/2081450468178579594
Good to know it's not just me! I also found it's way too verbose. So much so that I added an entry in my Claude.md telling it to be concise.