I am _very_ familiar with Claudish, and to some extent, the other AIs' writing styles. This article is human-written and features human writing quirks.
The very first sentence
> Back in 2022 and 2023 there were two big branches of machine learning happening at Meta.
is unmistakably human. That's not how a LLM would phrase this sentence, and if it did, it would have put a comma after 2023.
Leaving out that comma is not a grammar mistake. The comma slightly changes the feel of the sentence, but it's not wrong to include or omit it. But yeah, I agree with the other commenter that AI would be less likely than a human to omit the comma.
Adverbial clauses at the front of a sentence don't always have to have a comma. It's often taught this way to new English language learners or in style guides that want to keep things as unambiguous as possible because including the comma in that case is never wrong. But it's also sometimes optional.
One of the answers cites section 6.34 of the Chicago Manual of Style as follows: "Although an introductory adverbial phrase can usually be followed by a comma, it need not be unless misreading is likely. Shorter adverbial phrases are less likely to merit a comma than longer ones."
For the example we're discussing, "Back in 2022 or 2023", my personal instinct as a university-educated native US English speaker would be to include or omit the comma primarily depending on how much emphasis I want to put on the timing of my statement. I also know that I tend to write overly long sentences with too many commas, so sometimes I'd intentionally counteract my own tendency and limit the complexity of the sentence by removing a comma like that if the sentence still seems good without it. Other times I'd split the sentence into two, but my sentences rarely get as staccato as AI-written ones.
I am a professional writer and have been for over 30 years. (I do not use any form of LLM ever.) This means I read a lot. This also means that I have 30+ years of experience of readers not understanding what I wrote, or not getting further than the title, or not getting the main message, or inverting it in their heads, or inserting their own message and then complaining when I diverge, and an endless list of Ways People Do Not Get It.
I am also a trained TESOL teacher. Ability to capture gist is a skill we test for and measure, and many, maybe the majority, of native speakers don't have it and don't know.
In recent years I constantly see people going "this is written by AI" and I have yet to see a single of of them able to coherently prove their point. It's all just feelings and hunches.
So I am calling you on this:
How do you know? Show your working. Demonstrate your case.
Claude's writing style is at least as distinctive as any human's personal style. It has a long list of favorite words, verbal tics and common structures. On top of that, LLM writing is often bad in a very particular way: it's weak on actual things to say, but with an overheated style.
Some days, I spend over 4 hours a day reading walls of text written by Claude. If I couldn't recognize Claude's default "voice" by now, something would be wrong. It would be like a Hemingway fan not being able to recognize Hemingway. Except more so, because Claude's writing style is getting worse from version to version, descending into self parody.
On the statistical side, Pangram's model identifies AI-authored text with a 1-in-5,000 false positive rate, measured against hold-out texts from before 2022. My "ear" also agrees closely with Pangram. If I think something sounds AI written, Pangram virtually always comes back with "AI, confidence: high."
Claude's default voice, yes. But I'd assume a lot of people have learned to prompt it to something other than its default style. IMO it is good practice to have a style guide to feed in with the prompt.
> On top of that, LLM writing is often bad in a very particular way: it's weak on actual things to say, but with an overheated style.
This point is interesting because it raises the question of what "LLM writing" actually is. If it is expanding a smaller prompt into a larger article then yes, by construction the information density is low. But it can also be used to take a semi-coherent stream of consciousness and turn it into something readable and the people using it that way might already have started to slip under the radar.
This is a lot like how the criminals seem especially stupid because the ones who get caught are disproportionately the stupid ones. The easily detectable LLM writers are going to be the lazy ones.
> The easily detectable LLM writers are going to be the lazy ones.
To an extent, true. There are a lot of lazy ones though. And even for those who take steps, it sometimes leaves enough of a trace to at least raise the question.
One thing I think that helps is that for anything more worthwhile than message board posts like this I use Claude to review my text, make suggestions, and iterate on structure with me. But I'm the one writing the bulk of the text. I'll take some of its suggestions verbatim, but only if I genuinely like it better than anything I came up with myself.
The end product is something much more polished than anything I'd writ eon my own, but still comes off as being genuinely from me. At least that's what people have told me when I've asked.
I often run my writing through scans for AI tells. The number of things it flags that are just my own personal vocal tics, that I've had for 40+ years, is amazing.
But AI written pieces do have a certain feeling. A sort of saccatto in the succession of ideas that does not feel natural. They emphasize certain points, and you as a reader, you just wonder why is that. There is the “This thing, not just that thing”. There are also the three successive propositions (mostly in one sentences) to accentuate an idea and “Negation. Strong positive idea in the same direction”.
In general try reading one (vocally) to yourself and it will feel really weird.
It feels manipulative. The strong language always seems like it is trying to convince you it's right even when there is no need to, such as a factual presentation. Of course if you look at popular pre-AI blogs you can see strains of the same thing, so it's no surprise that llms blend together all of those "most persuasive" methods.
Just like em-dashes, some people have always done these though. Why are they penalized with immediate AI slop witch hunts? The LLMs didn't come up with these tics out of thin air.
I've realized that even when humans write that way, I also stop reading. Manipulative writing always shuts down my interest in reading it. At least when the LLM does it, it's a byproduct of training. When a human does it's intentional.
Yeah see that makes a lot more sense. If one doesn't like the writing style or content, just don't read it. The reflexive "this is AI slop!" complaints have themselves become a caricature. If something was written by AI but one liked the writing style and content, who cares? But regardless of the authorship if it's vapid and annoying, then that's different.
> ust like em-dashes, some people have always done these though.
Everytime someone claims that they have always written like this I grab a pre-2022 post of theirs and five both to a few SOTA chatbots and ask "did the same writer author both these texts".
Thus far I have never gotten a "likely" response.
If the author truly did not use an AI to write something, then it is more likely that theybhave spent so much time conversing with their LLM than reading human authored material that they now sound like an LLM.
This specific article, though, doesn't look anything like LLM output.
PS. Isn't it odd how all LLMs have converged on the same speech patterns, patterns which resemble almost no human authored material outside of high-pressure sales techniques?
I've used pseudo-em dashes going back to the 90s. In that I'll use hyphen (-) a lot. So for me, an actual em dash would be a tell, but if I just have a normal hyper in between two thoughts, it's not.
And yes, I agree that most people who light up on AI tell scans are indeed using AI. That's not my point.
>The LLMs didn't come up with these tics out of thin air.
LLMs were trained with a lot of synthetic data to transform them from a complete this text into a chatbot, I suspect that this tons of synthetic data that forces the LLM to answer questions into a specific ways also forced them to have this "synthetic/robotic" language. Claude users would have noticed the "belter and suspenders" phrase just started popping out after an update and I am sure is nto because lots of developers used it in their blogs and Anthropic scrapped those blogs in that update.
You have to start with the reason there’s a witch hunt. I love reading. I read books (novels) almost every day and I’m almost always perusing a textbook pr an article for my jobs and my hobbies. The signal/noise for entertainment or information was fairly high, then come LLM tools.
You start to be interested by the title of an article or a book cover, and then you start reading it and it’s just vapor. Nothing tangible to be gained. It’s like buying something expensive and finding out a cheap trinket under the wrapping.
After a couple of times, you will develop a certain kind of heuristics for this kind of texts. It will not be perfect and will have some false positives, but that’s the only way to keep your sanity.
I mean I get it. It bugs me too. And yes, obviously most of the text that people point at as being AI slop is in fact AI slop. At the same time, you also have the other set of asshats that just go respond with things like "AI slop!" to everything when they can't actually be certain. Like I said, some of these AI tells are things that actual real humans do too.
I like your vapor term. For me it's about the content anyways. If it's just some sort of vapid, pointless drivel then I'm not going to like it regardless of who or what wrote it. And in my experience text that strongly correlates with AI tells also strongly correlate with having sparse substance and lots of fluff.
In other words, if people don't like something, just don't read it. Shouting at people for AI generated text just makes you look foolish if the text is not in fact AI generated. And the person shouting "AI slop!" has no way to prove it other than vibes.
I want Scrabble rules for HN AI challenges. If someone finds an AI-generated comment, the commenter has violated HN guidelines and the comment should be deleted. But if the accusation is wrong, there should be a penalty for the often massive disruption the accuser has caused to the discussion.
(As of now, that four-word low-effort comment has generated over a thousand words in response, none of which improve this article's discussion.)
You need to start using LLMs a lot and then you will know how we know.
Edit: You know how you can recognise someone just from their gait while they walk towards you? I would struggle to describe that for an individual person but it doesn't mean I can't identify them from that alone.
> You need to start using LLMs a lot and then you will know how we know.
$SWEARWORD no!
I am an AI vegetarian: I won't touch them ever except for unavoidable ones (defaults on some search engines, e.g. when I use a web browser on its default settings – my own disable that junk) or for machine translation, which remains their only really valid use case in my book.
Amusing :) but I specifically mean in a SWE context. It’s used constantly by LLMs, Claude especially. Genuinely so curious to see certain phrases come up so often, beyond what I would’ve expected for how often they’re used by people
Genuine question from a place of ignorance: what in the silicon pipeline makes it take 2-4years to produce chips with a new model on them? Curious what the process bottleneck is.
Without being an insider, I imagine that most global fab capacity is contracted out several years in advance.
You might be interested in the tiny tape out project, which guides you through the process of getting your own design etched on silicon. If you only need larger features and not the next gen single digit nanometer stuff, you may not be so supply constrained.
I think you could get it down to three months between weight changes, if you can encode it in metal layers only. The remaining limits are the fab lead time, and the cost of a metal respin (hundreds of thousands to millions of dollars depending on process).
I think that comment meant it's 2-4 years until local models are good enough that it's worthwhile to burn an ASIC of them. Not that it takes 2-4 years to make an ASIC chip.
Why not have a bunch of SRAM and various operations like "Q4 matmul" in silicon? Model weights and even architectures could still evolve on a platform like that.
Doesnt "a bunch of SRAM" top out at maybe a few gigs per chip (with zero area used for logic)? You'd need an order of magnitude more to fit even a fairly weak general purpose LLM model.
If the silicon costs $200-300 and the company throws it away every two years that’s cheaper than a subscription.
Also, how many companies will just buy an M6/M7 MacBook Pro with 32GB+ of RAM in a couple of years and get “free” AI along with the workstation they were going to buy anyway?
This will be interesting. I can see some world where it’s used with consumers, but for the most part I think it will be in the cloud and that would make most sense
Zig isn’t so much on the blacklist because of the culture it carries from its maintainers, but because the ecosystem is no longer easily composed with other GitHub projects/GitHub Actions.
Powerful people figured out how to make suspicion work for them long ago. You have every right to be unconditionally suspicious, but it’s not a good way of accomplishing any change. Also their feelings are not hurt by what you or I think, they don’t care.
> Powerful people figured out how to make suspicion work for them long ago. You have every right to be unconditionally suspicious, but it’s not a good way of accomplishing any change.
How does one accomplish change? Even being a martyr doesn't get traction. As far as I can tell, you need to already be powerful. Nobody lets you into that group if you're not aligned with said group.
Protests (at least in their current form) don't work. Trying to assassinate someone doesn't move the needle (also not the play, I don't support murder), vocal grassroots leaders are no longer relevant at all, if they ever were.
Not by trading the same suspicions on the internet with fellow true believers over and over again, I think the past 10 years have proven that pretty conclusively. Maybe people should try some of the things previous social movements did, seemed to work pretty well even against a much more uniform media environment and a stronger hostile social consensus.
Protests don’t immediately solve everything, but I think looking at 2026 and concluding they don’t move the needle at all is a weird take. There are recent examples of protest movements (especially long-term ones) working all over the world.
I said assassinations don’t move the needle. Protests just give people the warm fuzzies. They don’t change anything. Tell me what has changed with all of the no kings protests? How well did it go for Iran when they protested?
The onus is on you to prove your point, not me to disprove it.
I’m not asking you to prove my point, but I think you’re being myopic.
* Vikor Orban was just ousted by a popular protest movement. This took years due to structural electoral issues but it did work eventually. And it wouldn’t if the people opposing him gave up because they didn’t change anything right away.
* The US ICE protests (and the federal government’s insane overreaction to them) let to the head of DHS being fired and a quantifiable drop in ICE activity (e.g. arrests and number of people currently detained).
* Nepal’s protests last year led to the resignation of the prime minister and a resounding electoral victory this year for their opponents.
Protests aren’t magic win buttons, especially because even the people protesting don’t fully agree on exactly what “winning” looks like. But they accomplish more than acting out your emotions on the internet.
Dario Amodei is not Elon Musk. Most powerful people do not internalize the emotions of others (though obviously some do). If you think powerful people are disproportionately sociopaths, you either agree with me or are not using the word correctly.
They generally care about what powerful/influential people think. I’m not powerful or influential, but I guess maybe the person I was talking to is. If so I apologize for the miscommunication.
Four days ago there was no intention to rewrite, now it's a simple desire to refactor. It's not adversarial conclusion, it's pointing out the clear hypocrisy.
Running an experiment, the experiment being more successful than you thought, and then deciding to put more effort into a bigger experiment is not hypocrisy. It’s engineering. If you think some of the objective facts they’re putting out (like test coverage and performance) are lies, go and prove it instead of appealing to emotion.
What do you think one would have to pay to have flesh-and-blood engineers get a cross-language port of a codebase of over half a million lines with a broad test suite to over 99% conformance? I think it would be astronomically high, especially given that for this specific project your hiring pool is going to be limited to people who can get up to speed with Zig and JavaScriptCore right away (or you’re going to have to pay them for low output for a while as you train them). Also it would be literally impossible to do in 6 days no matter how much money you paid, so unless they’re lying about that it’s still something that couldn’t have been done prior for any price.
More handwaving about the LLM hype machine is incredibly boring and enough of it is spewed everywhere that whatever social good it was going to accomplish must have already happened by now. If you want to inject reality into the situation, talk about reality (like Anthropic is at least pretending to).
The hype machine is real and we will talk about it as long as it pleases us. It took decades to get rid of smoking in public places and restaurants, and the clankers will eventually fall, too.
Did I say it wasn’t real? Or tell you that you couldn’t talk about it? No, I just pointed out that it’s all anybody talks about and it’s boring and doesn’t engage with anything specific about this stunt/project. And I can make melodramatic analogies too — like to the panic about global overpopulation that led to mass sterilizations in The Emergency. Panic is not an unalloyed good, and if you want to fight “the clankers” you should understand what they are and are not capable of.
Anti AI cope is unreal, the comparisons to smoking won't stop lol. The mental model of such people (like you) will be studied. LLM's won't go anywhere, keep dreaming.
Running an experiment and deciding based on the results is not hypocrisy, it's engineering, 100%.
Saying you have no intention of doing something then doing it is not engineering, it's being dishonest. He could have said "well decide when we see the results", why didn't he?
If he wasn’t willing to change his mind after he saw the results, then why would he do it at all? Can you explain the false motivation that you think he communicated in the original kerfuffle about this?
Why are you conflating "no intention of doing a rewrite" with his actual wording, "we haven’t committed to rewriting"? The latter does not at all indicate that there would definitely not be a rewrite.
Maybe he didn't think it would work. Maybe even if it does "work" they'll keep the zig version anyway. Maybe further study is needed beyond existing compiling/test-suite. Intentions and perspectives change over time, even only a few days, without dishonesty.
I'm guessing that if I said it ... that we have no intention of re-writing in rust ... that what I mean is "we have no intention of spending the extreme cost it would take to rewrite". When I discover the cost model is completely different that changes things.
Giving an opinion and making a commitment are different things, wording is important.
If you mean "we have no intention of spending the extreme cost it would take to rewrite" then say that, and it would be fine. If you instead say "we have no intention of re-writing in rust" you've said something very different, using a different set of words, which changes the meaning. Especially, if you say it directly in response to someone asking you whether you're going to rewrite or not like was the case here, and say that there's a high chance you'll just be throwing it away, to get the other person off your back. If then you go ahead and do it, expect them to call you out for it.
This is a very simple concept that can generally be understood by children at around age 4. Trying to cover it with vague terms and using the defence of "well I said I had no intention, and I probably won't do it but you see, I saw the results so I changed my mind so the chance was small but not zero", that's what a slightly older kid will try to do to see if they can get away with it, and as any kid discovers, that doesn't fly.
Okay, that's such a shallow take I'm going to try and explain it to you like you're 5 years old:
Changing your mind is okay, for example if someone said it was impossible to do the migration with current LLMs and it turns out they did it in four days, that person can and should admit they were wrong. That's not what he did though. What he did is say he had no intention of doing it, and then did it. That is lying. If he was testing and he didn't know if the change was going to be worth it, he could have said for example:
"This branch is a test, it's not a given it will work so until we see the results we won't decide if we'll be migrating or not."
He didn't say anything like that though, he basically said:
"We have no intention to migrate."
Why did he said the latter and not the former? Because he wasn't being honest, he was just trying to get people off his back, and so he didn't say what he was doing, the best for his own interest. We have a saying in my country: "it's easier to catch a liar than someone who's lame".
Also, before you come and say but he said he had no "intention" not that he wasn't gonna do it. A five year old might think that's a valid argument, but this person is an adult and we're all adults here, so it's not, it's equivocation and it's a logical fallacy.
> I am so sick of emotionally frail software engineers.
Then don't look in the mirror, you're probably being the biggest crybaby in this thread so far.
If experienced (in open source and corporate politics) developers would bet on Polymarket if the rewrite is going to be ultimately merged, which side would you bet on?
What would the emerging odds be? My guess is 19/20 in favor of ditching Zig.
I have followed many initial denials on a wide range of topics, not only rewrites, over the years. Like clockwork, most of them were lies.
Even if it passed the full test suite there are a ton of software qualities that are not captured by tests and I think it's unlikely the AI made the right trade-off in every such case.
* We haven't seen the benchmarks yet.
* It hasn't seen wide usage. Zig Bun has had tons of bugs ironed out, Rust Bun has a different set of bugs to iron out.
* The developers know the zig codebase well, they don't know the rust code base.
I see ChatGPT more akin to Wikipedia and Google than a calculator. We don't consider Wikipedia or Google as much of a tool as they are a knowledge reservoir of truthiness.
It's absolutely a knowledge base. The "training data" is its raw database which is then preprocessed to speed up the query engine. Then it does some additional processing to return amalgamated results from that database when presented with a query, which for some reason they call a "prompt." It's clever, but when you view the training data as part of the program source code, which it absolutely is, the information entropy of its output is miniscule.
- Google: How do I ...
- visit stack overflow
- copy/paste
- code generated.
There's more steps, but how is this different? Chat GPT though will go the extra mile and actually EXPLAIN what each bit does usually. It's not always accurate but neither is stackoverflow.
The only knowledge this system knows is whatever it can scrape and regurgitate that appears truthful. A whole bunch of positively true, but also many cases of falsely positive.
This system can’t prove its own work (at least not yet) before publishing the results. It just publishes, which is similar to Wikipedia and Google.
As part of a 2019 review of its ongoing decrees, the Department of Justice issued a two-year sunsetting notice for the Paramount Decree in August 2020, believing the antitrust restriction was no longer necessary as the old model could never be recreated in contemporary settings. [1]
