Author failed to explain the dash universe and their semantic assignments.
- em dash: adjoining thought (Hey, so — <declaration>)
- en dash: expresses a range (1–100)
- dash: compounds word-parts (Sarbanes-Oxley)
I'd also like to see more concise philosophy around em v. ellipsis, "limerance" is poetic (and ironically antithetical to this "essay" … and what of parens?). In short, em is communication and ellipsis is personal.
I offer this feedback as a humorously offended fan of all sidecar prose (dash, ellipsis, parens, quotes, and meta typographical treatments) pre-AI era.
Finally, I'd love to hear from this author regarding modern use of period in modern comunication, with canonical use-case of bulleted list as example.
That's a hyphen, and (to be pedantic) style guides usually do not consider it a dash.
There's also the 2-em dash in some style guides for redacting words, and since care needs to be taken to avoid a line wrap between the em dashes, Unicode has a character for it: ⸺
Finally,
> - em dash: adjoining thought (Hey, so — <declaration>)
This use is sometimes discouraged by style guides, and in many cases, it can be replaced with a colon, comma, or semicolon.
In general, even before AI, the em dash is often considered overused.[0] My guess is that the em dash is perceived as sounding educated, so novice writers who learn about it end up using it recklessly, and LLMs could've picked up this habit from them.
Fun fact! Where a compound term is formed by joining two nouns of equal importance — such as Sarbanes–Oxley or attorney–client privilege — an en-dash, not a hyphen, should be used.
> The first known documentation of the hyphen is in the grammatical works of Dionysius Thrax. At the time, hyphenation was joining two pieces of text that might otherwise be read as two separate words by a low tie mark. This was necessary due to the ambiguity inherent in the lack of word spacing. In Greek these marks were known as enotikon, officially romanized as a hyphen.
Furthermore, the conventions for the first (parenthetical) use case, as far as I understand, are either em-dash without spaces (Anglosphere convention), or en-dash with adjoining spaces (Continental convention: German, French, etc.), but rarely em-dash with spaces (as in the article).
I just know that MS-Word will replace a hyphen with an em dash in practical use, which is a big part of why I didn't really pick it out as an AI thing in general. Though AI, in particular Chat GPT, does tend to use a lot of them.
With HTML, I can do everything as a <div> and the eyeball reader doesn't know/care. Where it matters is that I'm stripping semantic meaning, which hurts computers and possible accessibility devices.
Most of my life, the same has been true about dashes. I use - for every one of the kinds you provided because - lives on my keyboard.
Is there real value in caring? Does it have the same semantic—stripping effect for computers/accessibility devices? The part of me wondering if I really should care is trying to suss out if this is a real problem or just an argument for semantics pedantics.
I can't think of any case where a human gets confused by some text because a hyphen character was used instead of an em-dash, and maybe that's what you mean by real value.
But I see you using an em-dash to hyphenate "semantic-stripping" as a joke, which is funny. Is that real value? Is conveying to others that you're educated real value?
Know the rules, fight the rules, I guess. More readable, but I’ve since regressed as tooling comes into play. Maybe it’s a good way to still use ems and differentiate from AI.
If you want to be official about things, use the en dash to replace a hyphen in compound adjectives when at least one of the elements is a two-word compound.
the pre–Websterburg Bakery era
The thinking is that using a hyphen here, as in “the pre-Websterburg Bakery era,” risks the suggestion that pre attaches only to Websterburg. It’s unlikely, though, that a reader would truly be confused.
n.b. I don't know how to code an endash here. Would love to learn!
One of the more memorable and influential library finds for me was an article in _Advertising Age_ which enumerated all of the special characters which should be set so as to accurately communicate, so in addition to en and em dashes it noted:
- − (minus symbol) for use in math equations or to indicate a negative
- I believe it also noted that there were two forms for the hyphen, normal and discretionary
- × (multiplication symbol) for use in dimensions and when actually indicating multiplication
- ′ and ″ (single and double primes) for use when indicating feet and inches, or minutes and seconds
- it also made explicit the difference between º (ordinal o) and ° (degree symbol) and noted when to make use of the other ordinals
- … while available as a special character, it is often better to construct this using suitably sized non-breaking spaces
- the various spaces en and em space, non-breaking flex space, &c.
I _think_ I still have a photocopy somewhere, but really should find a copy of the actual magazine....
The far more egregious and reprehensible thing is that many books now are published using quite pedestrian typing of the near equivalents from the keyboard as opposed to actually typing the correct signs.
The big reason for manually constructing an ellipsis is to ensure consistent spacing for the 3 and 4 dot forms --- some typefaces will include both, but may not fit well if the type specs have the text set tight or loose.
My apologies, should I have typed \thinspace.\thinspace.\thinspace.\thinspace.?
"It is (and, increasingly, was) used to signify that a bibliography entry has the exact same author(s), editor(s), translator(s), or corporate author(s) as the previous bibliography entry."
Absolutely does not deserve a new character just for that(especially one that is so unusually large).
Whoever introduced that is absolutely drunk on power.
The only reason that exists is because of a power discrepancy where the people with powers to create new characters are also the ones who do enough writing that they feel such a specific and niche pain.
Did you notice the examples were really old? Like late 1700s old?
Now, imagine you have a wooden box of lead or pewter letter forms. They're small, heavy and backwards. You spend your days lining up those forms in rows and then pressing them to ink and paper to print books and documents.
So you're lining them up, slowly running out of 'e's and 'f's and 'r's. You decide to save yourself some time and more importantly some of your precious few letters that you need to make last the whole page. So you omit some of those words, names in this case, and establish a convention with other printers and authors about how and when to replace words with shorthand. This lets you and your fellow printers save time by not having to work to arrange those admittedly duplicate letters and stretch those letters.
Fun fact, that's why old 's' are sometimes written with an 'f' missing the bar. Because they were re-using forms where otherwise a letter was so common they'd run out!
Just because it’s a problem you’ve never encountered doesn’t make it stupid.
You’ve clearly never had to deal with a lot of references. Author lists can cluster. If you’re reading, it can get tedious if the last 5+ references are to the same three authors. Let alone 20+ references.
This happens. In some writing it happens _a lot_.
It’s a common enough that it’s even explicitly mentioned in the Chicago Manual [0], one of the main style guides for academic writing in the US:
> Em dashes also substitute for something missing. For example, in a bibliographic list, rather than repeating the same author over and over again, three consecutive em dashes (also known as a 3-em dash) stand in for the author’s name.
Hmm, like in the References section for a paper? Personally I don’t see any point to typing that sort of thing out (bibtex handles it more accurately that I could anyway).
For the reader… I dunno, do people actually read through a citation list, in a manner where the redundancy would be relevant? When I look, I’m usually specifically looking for the single entry that corresponds to the citation on some claim in the text.
Although if someone really wanted these triple-dashes for their journal for some reason, I bet some tex could be coded up that would generate them. Could look neat, maybe.
If not, I doubt some typesetter in the 19th century (or earlier) was so drunk with power when setting out the lead type to make the 3-em dash.
EDIT: Additionally, unicode primarily exists to encode the typographic symbols that exist—that is to say, it's a catalog of what already is. Take Ogham[1] for example.
I do miss the old PHP manuals on Hurricaine Electric with the comment sections where anyone could discuss every function, ask stupid questions, and post totally bogus answers, which everyone would then believe without skepticism, then copy and paste into their code in production.
It was like pre-AI poor man's artisnal LLM hallucinations.
There need to be chat discussions like that for every unicode character.
I mean... would it really be that big of a crime against humanity to insert two or three em dashes if the situation seriously needs it? I already have enough trouble distinguishing an en dash from an em dash.
I hadn't thought of it as named that way, but super long dash is a reality of ye olde typesetting, e.g. bibliographies with one author appearing many times, but named only with the first work, or in indexes where there are subordinate concepts etc.
Here's a googelensed bit of the index of a philosophical work from the thirties. The entries after 'Aristotle' are all megadashes, but google lens reads them as hyphens. Under 'Kant' the subheadings have subheadings so the typesetter doubles the triple-dashes up without however employing a six-em-dash. Current taste I think would just use indentation for all the purposes mentioned they are all line-beginnings.
Only know the discussion superficially, but from what I understood, the problem was as follows: Chinese Hanzi, Japanese Kanji and Korean Hanja scripts have the same historical roots (Han) and still look very similar when you write them — however, they are fully separate scripts with different use of letters and optical differences that are important.
However, when encoding them into Unicode, the consortium suddenly discovered the importance of being frugal with character IDs and only defined a single "Han" sequence of characters for all three scripts. You need to select a matching font to actually display the text in the script that's intended.
This makes using those characters an unnecessary hassle and also came over as a cultural faux-pas to some (essentially like encoding Greek Eta, Latin H and Cyrillic En as the same character, because they look similar and the scripts have a common history)
It also looks a bit hypocritical when Unicode otherwise goes the opposite way and has the space to encode dozens of variants of the Latin alphabet or — as we see here — several categories of dashes.
Three-em dashes are multinational conglomerate punctuation.
They can be used as adhesives, abrasives, laminates, passive fire protection, personal protective equipment, window films, paint-protection film, electrical insulation, car-care products, electronic circuits, optical films, Scotch tape, Post-it Notes, labels, dispensers, and magnetic tape.
In German, nouns have gender, so why not more inclusive punctuation? The him dash, her dash, it dash, and the nonbinary them dash. It would really infuriate MAGA if you could apply emoji gender and skin color.
And of course if you really want to make a statement and stick it to those anti-AI em-dash nit-pickers, there's always the Zen-like struck-out no-break space (U+00A0 NO-BREAK SPACE + U+0336 COMBINING LONG STROKE OVERLAY) ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ you can combine them to any length!
It's a shame Hacker News filters out the really good ones, like the üm dash: — (U+2014 EM DASH + U+0308 COMBINING DIAERESIS) and the over-accentuated üm-á-ñå-ḍ-ç dash: — (EM DASH + DIAERESIS + ACUTE + TILDE + RING ABOVE + DOT BELOW + CEDILLA).
@dang I know @mentions don't work so you'll never read this, so screw you for filtering out my favorite glitchy punctuation diacritics, you jerk! I feel like my creativity is stifled and freedom of speech suppressed!
Test:
'There is no greater joy in life than typing a sentence, realising it needs a clarification, a caveat, an addendum, but in the same sentence, not a new one, and, guess what, the comma is right there for you. You want to package things neatly together, and the comma lets you do exactly this. Commas make writing feel like you are assembling lego bricks.'
I like using commas for spacing though em dashes do the same thing and avoid the run-on police?
"There is no greater joy in life than typing a sentence, realising it needs a clarification, a caveat, an addendum (in the same sentence, not a new one) and, guess what, parentheses are right there for you."
> There is no greater joy in life than typing a sentence and realising it needs a clarification, a caveat, an addendum (in the same sentence, not a new one) and guess what? Parentheses are right there for you.
That's what I'd do. Note the conversion of the first comma to "and," as well.
Your first sentence is a perfect example of why emdashes need to exist. The comma is overloaded, since it can be used both as a list separator and for parenthetical asides. But that first sentence starts using them as a list separator, and then inserts a parenthetical aside into the list. Emdash wins this test.
I sometimes prefer parenthesis for very short asides—say, three words or fewer.
I find m-dashes easier to read for most other purposes. They’re nicer in longer asides because they’re a little more visually prominent, while also (for some reason) not breaking my mental flow quite so much. They’re far nicer looking when an aside comes at the end of a sentence, as you don’t need to close an m-dash. That always looks like crap with parenthesis, as if the author should have thought of a way to avoid doing that, while with the m-dash there’s no problem at all.
I just thought of the most annoying blog in the world. Individual spacing between words, depending on how thoughtful you want the pause to be. And by "individual" I mean a different floating point number denoted amount of em each. em > em-dashes. Or put differently, em dashes em-dashes.
It's a bit like 2 space indention versus 4 or 8 space. I find the em-dash example pretty compelling because visually relaxes my eye more than commas do — similar to how prefer at least 4 space indention over 2.
(Unrelated: The alt-code for em-dash on Windows is Alt+0151. I just typed my first manual em-dash wow exciting.)
I'm unreasonably fond of the numero symbol (№), although I've never quite settled on whether it should be followed immediately by the number or whether there should be a (non-breaking, potentially thin) space between them. Although it does tend to look awful in monospace fonts.
Ah, so my constant interrupting myself is a sign of intelligence? That's great news.
But seriously, trying to express the spider web of thoughts in my head as a linear string of words is really difficult for me and that doesn't make me feel very smart.
Funny - the ChatGPT app on mac uses a font that doesn't differentiate between any dash.
I use dashes in place of commas, semicolons, colons, and brackets all the time - I just use the standard keyboard dash. Why the heck should I bother learning some stupid key combination to make the dash longer? I highly doubt that the vast majority of readers even know that there are different dashes with different uses. When I learned about it, I decided to be a conscientious objector because using - vs — and assuming I can tell the difference is inane (because it is font based).
(ironically, the HN text editing dialog shows - as ~.8 of the length of — whereas on the site it shows as ~.25 of —)
Wrong analogy. Why should I bother to differentiate between the length of a dash when the vast majority of readers don't have a clue what it means (and don't need it to determine meaning).
Because the impact of something can be felt even when the object doesn't appreciate they are feeling it. People choose fonts for the particular impact they will have on the reader, even though that reader won't think about it.
That a book about a perfectly made tool cabinet suffers from so many typos (the maker's name, Henry, is mis-spelled on the inside cover) and such pedestrian composition embodies much of what went wrong when the typesetting unions all bargained for sinecures rather than active participation in the Desktop Publishing Revolution.
>Because it looks better and is a guide to communication.
Personally, I think em-dashes (when in a font that shows the damn difference) look exaggerated and silly. They certainly don't provide any important guidance to its meaning that I can't derive from context.
I love the author's attitude. In my writing, I'll probably consider using en dashes ("-") in place of em dashes ("—") in the way he describes.
While it technically won't be grammatically correct, it'll get the benefits he describes in the article while also not being accused of being AI-generated.
I love ellipses, it really captures that even when writing Rust documentation my mood is existential and disaffected... like a new grad in the pool, slowly burning under the noon sun on an inflatable lounger, too dazed and decadent to tap even a finger for their next paloma.
Since before LLMs were a thing I've thought em dashes (at least outside of fiction) have a kind of irritating, business-casual feel about them, since they are simultaneously informal but also fussy in a way. That also describes exactly what bothers me about the classic ChatGPT writing style, which is obsequiously friendly and casual but also perfectly organized and precise, and which I imagine is based on a large dataset of middlebrow American news blogs and similar sources. So I'm not surprised that the em dash has stuck out to people as such a prominent feature of its style. It definitely has its uses but I can't say I would mind at all if the controversy diminished its use in real writing.
If you love em dashes and commas you will also love this https://archive.ph/0dxpW from the newyorker by Mary Norris.
I was never a big reader as a kid, but I fell in love with the New Yorker, much later in life, where I kind of adopted their idea of: "Once you understand them, you can read through long sentences very easily". That is the newyorker style of writing. One falls into two buckets, you love that kind of writing or you hate it. If its the latter, you should adopt em-dashes and commas to improve your writing style.
My problem with AI use of em-dashes is that it covers up sloppy writing. I simply collected about six examples of horrible use of em-dashes and how I would have written the text. This became part of my SKILL for general writing (along with a lot of other specific examples).
Nothing wrong with CMOS/APA/MLA dashes, I quite like them but usage in LLM prose has ruined it for me. AP dashes are weird, too much horizontal spacing, looks horrible and can really erode typographic color sometimes. "International" Oxford/Cambridge spaced dashes is nicest.
The funny thing is at both extremes, either none or indiscriminate use of dashes like in TFA, is a canary of "human writing".
People (including comments here) often attribute LLM use to prevalence in training data but it is probably far less common in the wild than we are led to believe, the more likely scenario is that whatever text-to-text normalization tools used ended up inserting whatever their idiosyncratic typographic styles were.
> on Mac, for example, it’s simply Option + Shift + Hyphen
And on Windows it's Alt + 0151 on the numeric keypad, which means you a) must have a numeric keypad and remember the number, or b) must use the emoji search. And on Linux it's weird too.
Em dashes are therefore Mac user garbage. The correct replacement is --
Same. It is really annoying to write something and have someone say, “just so you know, em dashes makes it obvious it was written by AI.” You know how else uses em dashes? People who know how to write.
> It is really annoying to write something and have someone say (...)
It's worse than that. What they are saying is the equivalent to saying that knowing how to do something properly, and having the attention to detail to do so, makes it obvious it was written by AI. It sounds like a prelude to human incompetence, to ensuring that AI products are better than human ones.
About the only semi-convincing argument that I've heard is that AI over uses these tells. But let's face reality here: a lot of people griping about AI generated or AI assisted writing are not looking at over use. Plus a lot of people tend to over use the said tells while they learn how to use them properly.
So... fuck 'em? Hit 'em with a: "You're absolutely right — em dashes are a common signifier of AI" and go about your business. Anyone creating on the internet[1] is running laps around the class of losers who run from post to post dropping useless comments.
[1]: I'd even go so far as to extend that "anyone" to: "anyone creating, even with AI assistance!"
I'm going back to em dashes after this post, and anybody who calls me out is getting a link to it -- they can suck it up.
That said I've only been on Linux for 25 years so I still haven't figured out how to actually get em dashes and have to resort to the double minus like up there.
Regarding the AI-perspective of the emdash being commiserated here...
As a few have pointed out, many writers have been using it usefully since long before the blight.
The current state of revolt is probably a phase though, as models will learn to account and adjust for the litany of literary crimes they've been committing. In some ways we are fortunate to have such easy signals to work with while, already, false positives are becoming almost as common as actual positives, indicating it might not be long before we just can't tell anymore. Which would suck, but then we'd have to adjust the nature of our complaints to something more profound, something more... insidious, something.... Maybe Woodrow Wilson? [0]
It's not only the emdash that threatens the quality of our dialog and linguistic welfare — it's many principles, such as the the Rebuttle>Reframe>Prescribe structure of this very sentence. Once is a lot, but multiple instances of this structure in a single comment would be worse than swallowing a double hyphen sideways.
The LLM is methodical, deliberate, unhurried — in a way that rewards the vigilance, attention and patience required to spot compulsions, such as the Rule-of-Three seen here.
It's important to catch and worth noting each instance of filler-transition that adds little information while working toward word-counts.
Undoubtedly, if we want to make our world safe for the emdash once again, we certainly should not overlook overused intensifiers posing in places where real work should be.
While such examples can be found in legitimate human-written texts, they can also be found in LLM generated texts, and this Reflexive false balance style sentence is suggesting we just can't always be sure.
Consider a person suffering generic example syndrome, where instead of naturally recalling a situation from actual memory, an example might be synthesized from some statistical category, seeming a bit out of place.
And do not forget sentence length clustering around a mean, which humans have a tendency to naturally avoid.
Frictionless surfaces might also hint the involvement of LLMs, where the final structure may be suspiciously clean of normal human quirks.
Of course, the ultimate question here isn't even being asked, and the emdash is not a registered trademark of LLMs. It's not about whether we should fear using them, or even their value — it's about this Symmetrical closing that despite being written by a human (maybe subhuman) has strong patterns of an LLM.
0. “Since I entered politics, I have chiefly had men's views confided to me privately. Some of the biggest men in the United States, in the field of commerce and manufacture, are afraid of something. They know that there is a power somewhere so organized, so subtle, so watchful, so interlocked, so complete, so pervasive, that they better not speak above their breath when they speak in condemnation of it.” "
I did not know option + shift + hyphen produced an em dash. I've always used a double hyphen as a stand in -- as I did just now. LLMs don't do it and I haven't been accused yet. Though, more likely, I'm not popular enough for people to notice!
The overemphasis by critics on punctuation and "this not that" is tiring. I don't really know how to describe LLM writing but it has the same qualities as flat soda. Some authors you feel as though you fall through their work with little effort; unable to stop until gravity slams you into their conclusion. LLM writing can feel like an uphill hike to a sewage treatment plant. That is, its easy to convince yourself to stop at any time. There's a positive quality to that. I hope AI labs don't find a way to remove it.
I think Em dashes are fine in fiction where you're listening to someone's train of thought or something sudden happens, but they are really annoying in non-fiction.
has anyone used/seen em dashes in file/dir names? just wondered — I have't.
It's certainly easier to read than a hyphen/minus...but i guess it may become problematic cross-platform (on windows/linux it isn't as straight forward as on macos).
I used to have ripped music that had em dashes in the filenames. But several music players displayed a question mark for it. So I replaced them all with a hyphen.
I ripped music from both CDs and Internet radio (not Spotify). I remember there were definitely em dashes in those from Internet radio, but I'm not sure whether they also occurred in those from CDs — those filenames probably came from an online DB. Of course the Internet radios might have ripped the music in the same way.
This post is so self-consciously dependent on inserting em dashes everywhere that it kinda supports its own counterargument. If you're going to make a case for em dashes enabling more fluid writing and more distinctive voice, then you ought to actually use them that way.
> ...on Mac, for example, it’s simply Option + Shift + Hyphen.
Fedora with XFCE here, but last I checked on Windows, it worked exactly the same over there: With a Chinese IME installed, tap shift to switch input language, shift+hyphen produces a double em-dash: ——, likewise, shift+6 a double elipsis: ……
You encounter those a lot in simple, machine translated Chinese, along with literal translations of "not X but Y", "not just ... but also" and similar sentence constructs that are common in Chinese. I suspect that's how a disproportionate amount of that ended up in training material in the first place.
I'm all for Unicode if you want to say something like 你好 but separate nearly similar characters, like the different kinds of dashes, feel somehow wasteful. I prefer to stick with things that align with ASCII.
Typesetting aside, the em-dash as a grammatical construct is indeed pretty great.
LLMs use em- dashes because they appear in the training set. In other words, humans use them a lot. Such a weird issue to be adopted as an indicator of robot text.
I didn't know that AP style recommended spaces either side of the em dash. Here in the UK we generally use en dashes instead, and I thought em dashes always had no spaces. Apparently the wire machines that AP sent their news out by in olden times couldn't always do en dashes.
I just use hyphens all the time and always have. I don’t like that AI loves em dashes so much that it looks like everything I write now is AI generated, but if you were to deep dive you’d notice I’m using hyphens - never em dashes - so my writing can’t be AI. Granted you’d have to be pretty pedantic to care in the first place but it’s 2026 and here we are.
i typically enjoy a good profanity-laden rant, but something about this particular one feels very forced. this sentence, for example: "[...] What if someone points one out in my writing?”. Bitch, please.". i cant really put my finger on why, but it almost feels like the author wrote their rant, then came back afterwards with a quota of fuck/shit/bitch they wanted to meet, placing them wherever they could.
in any case, i mostly agree with the author that ai has co-opted too many things to be concerned about it anymore. i remember being linked a list of "1,000 words not to use" because ai used them a lot (e.g. "delve", "absolutely", etc.) and the venn diagram with the 1,000 most common english words was practically a circle. you would be unintelligible if you rigorously avoided them all. now i just write without care of my ai-similarity score.
Same. On Google Docs it auto-corrects to an en-dash (–), which is annoying, but I find the double dash (--) more easy to spot and read on computer screens.
Also, cosmetically I don't really like how em-dashes are used without spacing (e.g. "this—then that" instead of "this — then that"). Feels like we're trying to import publishing styles from newspapers and books heavy on serif font to places where monospace and sans-serif is the norm and it doesn't quite fit well.
> Em dashes can be used instead of brackets in a sentence — like this one for example — and you can just type your shit and keep going. Brackets are for people who raise their hand in meetings and say: “this might be a stupid question”. A parenthetical is whispering: “Sorry, don’t mind me, I’ll be quick”. Em dash kicks the door open and announces itself. Use brackets when you are embarrassed by your own thought.
Literally had me laughing out loud.
And then this bit:
> As for semicolons, semicolons are neat, they are like the coolest punctuation mark after the em dash — the fundamental indecision compressed into a single punctuation mark — you are neither starting a new sentence nor you are, exactly, continuing an existing one. You are just… piling your stuff together. I am always happy when I find a place for the little guy semicolon. But, again, most cases where you could use one are better off served by the em dash as well.
couldn't help but remind me of the following Kurt Vonnegut quote:
> Here is a lesson in creative writing. First rule: Do not use semicolons. They are transvestite hermaphrodites representing absolutely nothing. All they do is show you've been to college.
in order to inform people that im not a bot/llm i -- outside of giving up on autocorrect -- use two dashes instead of one solid line because LLMs were using them a lot.
unfortunately microsoft (specifically outlook) and libreoffice have chosen to replace them with a proper em dash even if you turn off autocorrect which is bonkers-levels of infuriating.
if i turn off something it /should be turned off/, not just be kinda turned off; y'know?
> But also, maybe consider not giving a fuck about what someone on the internet think?
If I'm writing something, typically I want to share my thoughts with other people and get their feedback. It won't work if they dismiss my writing because of some superficial stylistic choice, without even trying to engage with the ideas that I'm trying to express.
I feel like people only look at the form and ignore the substance of the content. Honestly, as long as the content itself is well written and the em-dash is used correctly, I don't think it really matters.
When I first learned about em-dashes, I was told they were a mark of sophisticated, intellectual writing. But now people say they sound like AI. Is that really a bad thing?
It's strange to see sophistication become low-class just because AI uses it. It makes me think that form is really about who the author is, not the form itself.
In that sense, I'm not sure what it even means to be 'human-like.' Being human-like just feels like an illusion that people who want to believe in it keep insisting on
I too find it a pity that the em-dash has been co-opted by AI. I use a double-hyphen ("--", the old school way of writing it), but fear that this will also become used by AI seeking a higher form of 'sophistication' as you put it.
We're experiencing broad scale changes in the sentiment of style in written English. I don't know what the best approach to take here is necessarily, but the more I see language evolving over time (especially over the past two decades) I think I'd prefer to shift my habits in line with whatever the "modern standard" is in order to avoid falling into pitfalls, i.e. being accused of being an AI writer.
Good point, but I think it's a little more complex than that.
To my knowledge (and what I meant by "old school"), it was common to type a "--" dash on typewriters[0] because most of them lacked en- and em-dashes, which are obviously not digital documentation devices.
But yes, typesetters of old (old, old school!) had myriad glyph blocks available to them for pressing!
The double-hyphen is by no means "old school": it's a kludge for twentieth-century systems like typewriters and word processors that did not have em dashes or an easy way to insert them. The very names "em dash" and "en dash" should illustrate their age: they go back to the fifteenth century, and "em" and "en" were standard printing widths in moveable type. People who write CSS still recognize em as a width.
The idea that a double-hyphen is "old school" is on par with the false notion that double-spacing between sentences is a product of the typewriter; one need only look at a typeset book from before the twentieth century to see the truth, or look at a decent image of the Declaration of Independence or other hand-written document from prior centuries to see the extra space left between sentences.
Many of the changes we see in written English style are the product of ignorance of the past beyond two generations or so. Avoiding em-dashes to keep up with the latest anti-AI fad is just reacting mindlessly to the stimulus of popular ignorance instead of deliberately situating one's style in the history of writing conventions.
As per sibling comment, I think my "old school" wording is getting misconstrued here. I did mean 20th-century "old school", i.e. typewriters... Is the Late Middle Ages considered "old school"? :)
Agreed that in terms of precedent, typesetters had moveable type dashes for their presses.
> the fifteenth century, and "em" and "en" were standard printing widths in moveable type. People who write CSS still recognize em as a width.
I'm going to write my response using vernacular English so as to avoid AI claims, so pardon my French.
> We're experiencing broad scale changes in the sentiment of style in written English.
Lol no we're not. The only people who give a fuck about this em dash AI bullshit are losers on Tumblr and commenters on fanfiction sites who are obsessed with the idea that AI is the current Great Evil.
This is the orthographical equivalent of the New York Times running an article that soy-glazed avocado bouillon is the latest trend among Millennials because, like, one of the editors caught their college dropout son eating it one morning at the family table. Son said "duh dad everyone's doing it" to fuck with his dad, and next thing you know, a hundred think pieces on Mexican-Japanese breakfast being trendy.
I've never met an adult in my professional or personal life who actually gives a shit about this. We're talking a fraction of a percent of people who don't want you to use AI, and a fraction of that who waste their life trying to detect it.
Write the way you want. If people can read it, that's all that matters.
Apologies for the French. Blame Canada.
*Edit* Also the em dash is based. Using a double dash should be a prison sentence.
> people only look at the form and ignore the substance of the content.
This is largely because most people are incapable of thinking critically and evaluating substance. So they regress to indirect markers of good writing, and the whole thing becomes a game of signaling.
The rot is so deep, many popular resources on writing focus more on style and presentation than focusing on the substance of writing. And even when best-intentioned, all this does is make poor writing a bit more palatable by giving it a veneer of polish.
This is generally the way it's always been. As evidence, observe the plethora of guides like Strunk & White for the entirety of the 20th century. This time period not only included the moon landing, but also the final decades of the third European Renaissance. You'd be hard-pressed to identify a time period that was jam-packed with brilliance.
Guides such as S&W focus entirely on style. Then we have the dearth of guides on composing a good sentence.
Of course, S&W is also filled with insanely stupid advice. "Don't split an infinitive" immediately comes to mind. Ask yourself which sounds more powerful, oh fellow geeks:
* To go boldly where no one has gone before
* To boldly go where no one has gone before
Split infinitive form is so much better. "Don't split an infinitive" is one more smooth-brained 19th century rule invented by Latin-obsessed weirdos who thought it was the perfect language, and that they should impose Romance grammar on Germanic languages. (See also "don't end a sentence with a preposition" and "don't start a sentence with a conjunction.")
I agree that post-COVID we have a glaring issue with the problem you highlight, but I don't think the focus on orthography is a symptom. Long ago, my dad criticized my résumé because I didn't put two sentences after periods despite me using a variable-spaced font. The man's nearly eighty. He's not exactly some 21st-century miseducation case.
One caveat I'll add at the end is that focusing on form isn't always bad. Form is one of the great joys of reading both poetry and prose. Word selection is often form rather than substance. The opening lines of Lolita are pure form. Nabokov could've easily just said "I'm a pedophile and Lolita is attractive to me." Instead, he chose words that have you overuse your tongue when you sub-vocalize, making the text feel uncomfortably erotic to a normal, non-pedo reader.
> This is largely because most people are incapable of thinking critically and evaluating substance.
No, with AI it's because it's MUCH faster than evaluating substance, and fast ways to weed out AI slop are desparately needed, since it can be produced so quickly.
It's also infinitely more error-prone. All form-based heuristics are horrible for detecting AI because the one thing AI is damn near perfect at is writing like an actual human.
It's the thing it was trained to do. The entire decades-old architecture has exited for specifically that purpose. That second "L" in "LLM" is "language" for a reason.
As of recently, there is a strong correlation between use of em-dashes and a verbose, bloated style of writing that uses many words to not quite get to multiple repeated points. Not well written, in other words.
The combination of both I usually find in text generated for me by LLM, and so far, it serves as a good signal for density of information and, inversely, not-quite-right ideas and their connections. (That's not to say it's pervasive, but below the surface, the quality of LLM text as a medium is often not great.)
There may be well-written text of non-trivial length generated by LLMs, but so far I haven't knowingly encountered any, with or without em-dashes. Human editing remains very necessary.
The problem is that the substancelessness of AI writing nowadays tends to be subtle and well-hidden (but huge). Focusing on superficial "tells" is basically an attempt of a quick, cheap defense against the imbalance of human effort, where someone can spend a few dollars and a few minutes and e.g. generate dozens of PRs that would take many hours of real human effort to process and fix the deeper shortcomings of.
The em-dash-as-AI trope is nothing new. (Apart from the fact that MS Word etc convert them automatically.)
If I post some information, then I have had people accusing me of cribbing it off Wikipedia. When I have translated from some languages, people assume I'm using Google. Even though in one case, the language in question wasn't even available on Google Translate. (I went and checked.)
I also love them - makes spotting AI written content easier. People debated implementing complex algorithms how to watermark AI content - but luckily AI just settled on it by itself.
I've used dashes in typeset works ever since I learnt how to do them in (La)TeX (---). You don't usually put spaces around them, though.
There are actually four different "dash" type things in TeX: a hyphen, an en dash (for ranges like "10--20 pages"), em dash (---), and the minus symbol ($-$).
Isn't english known for not having an official set of gramatical rules? There are conventions and rules from universities i.e Cambridge vs Oxford comma. But there is no single body that is commonly accepted by all english-speaking countries to be codifying the official rules of the language.
I am personally a fan of space padded em-dash. Visually, semantically, it makes sense on all modern planes. There's an argument for line-break parsing.
Of course I recommend do-not-adopt, think Roman-cart-wheel-axle-width reasons.
It annoys me that recently people seem to propagate the wrong use of em dashes. There is no space—an em dash is practically never typeset with spacing. You use an en dash in European typesetting – they have spaces, which your typesetting system hopefully adjusts to sane values. They're not like the spaces between words.
At least, that's how I learned it. Everyone seems to disagree and write em dashes with spaces nowadays.
- em dash: adjoining thought (Hey, so — <declaration>)
- en dash: expresses a range (1–100)
- dash: compounds word-parts (Sarbanes-Oxley)
I'd also like to see more concise philosophy around em v. ellipsis, "limerance" is poetic (and ironically antithetical to this "essay" … and what of parens?). In short, em is communication and ellipsis is personal.
I offer this feedback as a humorously offended fan of all sidecar prose (dash, ellipsis, parens, quotes, and meta typographical treatments) pre-AI era.
Finally, I'd love to hear from this author regarding modern use of period in modern comunication, with canonical use-case of bulleted list as example.
That's a hyphen, and (to be pedantic) style guides usually do not consider it a dash.
There's also the 2-em dash in some style guides for redacting words, and since care needs to be taken to avoid a line wrap between the em dashes, Unicode has a character for it: ⸺
Finally,
> - em dash: adjoining thought (Hey, so — <declaration>)
This use is sometimes discouraged by style guides, and in many cases, it can be replaced with a colon, comma, or semicolon.
In general, even before AI, the em dash is often considered overused.[0] My guess is that the em dash is perceived as sounding educated, so novice writers who learn about it end up using it recklessly, and LLMs could've picked up this habit from them.
[0]: https://slate.com/human-interest/2011/05/em-dashes-why-write...
> The first known documentation of the hyphen is in the grammatical works of Dionysius Thrax. At the time, hyphenation was joining two pieces of text that might otherwise be read as two separate words by a low tie mark. This was necessary due to the ambiguity inherent in the lack of word spacing. In Greek these marks were known as enotikon, officially romanized as a hyphen.
https://en.wikipedia.org/wiki/Hyphen
my favourite (and I hope LLMs never converge to use it)
Konnichiwaー
Most of my life, the same has been true about dashes. I use - for every one of the kinds you provided because - lives on my keyboard.
Is there real value in caring? Does it have the same semantic—stripping effect for computers/accessibility devices? The part of me wondering if I really should care is trying to suss out if this is a real problem or just an argument for semantics pedantics.
But I see you using an em-dash to hyphenate "semantic-stripping" as a joke, which is funny. Is that real value? Is conveying to others that you're educated real value?
I've been doing doing it that way for so long I still often use the "--" construct from way back in ascii email days.
At least I've gotten past the two-spaces after a period I learned from the typewriter. Most of the time, anyway.
If you want to be official about things, use the en dash to replace a hyphen in compound adjectives when at least one of the elements is a two-word compound.
The thinking is that using a hyphen here, as in “the pre-Websterburg Bakery era,” risks the suggestion that pre attaches only to Websterburg. It’s unlikely, though, that a reader would truly be confused.n.b. I don't know how to code an endash here. Would love to learn!
I grew up reading novels and plays
Inspector Lestrade and Holmes were speaking:
— What’s up?
— Not much
— Cool.
"I was thinking—"
"Well, there's your first mistake"
https://www.compart.com/en/unicode/U+2212
- − (minus symbol) for use in math equations or to indicate a negative
- I believe it also noted that there were two forms for the hyphen, normal and discretionary
- × (multiplication symbol) for use in dimensions and when actually indicating multiplication
- ′ and ″ (single and double primes) for use when indicating feet and inches, or minutes and seconds
- it also made explicit the difference between º (ordinal o) and ° (degree symbol) and noted when to make use of the other ordinals
- … while available as a special character, it is often better to construct this using suitably sized non-breaking spaces
- © ® ™ ℠ ℗ (copyright and so forth) to indicate such legal constructs as appropriate
- the various spaces en and em space, non-breaking flex space, &c.
I _think_ I still have a photocopy somewhere, but really should find a copy of the actual magazine....
The far more egregious and reprehensible thing is that many books now are published using quite pedestrian typing of the near equivalents from the keyboard as opposed to actually typing the correct signs.
I wrote a bit on this at:
https://tug.org/texshowcase/onetype.pdf
(note that it intended to be printed, folded, cut in the center, then re-folded so as to make a "hot dog" booklet: https://www.makingbooks.com/hotdog.shtml )
EDIT: That said, folks should be sure to get:
https://www.goodreads.com/book/show/41600.The_Mac_is_Not_a_T...
or
https://www.goodreads.com/book/show/358248.The_PC_is_Not_a_T...
https://wincompose.info/
I'm curious why? Kerning? Regardless, it sounds like a lot of work
> but really should find a copy of the actual magazine....
Uh, "..." ;)
My apologies, should I have typed \thinspace.\thinspace.\thinspace.\thinspace.?
Why this thing exists, I do not know. Unicode people seem to like making bad decisions, like Han unification and emoji sprawl.
> He lies buried in the corner of his churchyard, in the parish of ⸻, under a plain marble slab
If we were to use three em dashes in a row, the renderer typically puts a 1-2px gap between them. Using U+2E3B leaves no gaps.
Likewise, we use a 2-em dash, U+2E3A, for partially obscured words, for example this line from Gogol's short fiction:
> The town of B⸺ had become very lively since a cavalry regiment had taken up its quarters in it.
https://danieljtortora.com/blog/3-em-dashes
Absolutely does not deserve a new character just for that(especially one that is so unusually large).
Whoever introduced that is absolutely drunk on power.
The only reason that exists is because of a power discrepancy where the people with powers to create new characters are also the ones who do enough writing that they feel such a specific and niche pain.
Now, imagine you have a wooden box of lead or pewter letter forms. They're small, heavy and backwards. You spend your days lining up those forms in rows and then pressing them to ink and paper to print books and documents.
So you're lining them up, slowly running out of 'e's and 'f's and 'r's. You decide to save yourself some time and more importantly some of your precious few letters that you need to make last the whole page. So you omit some of those words, names in this case, and establish a convention with other printers and authors about how and when to replace words with shorthand. This lets you and your fellow printers save time by not having to work to arrange those admittedly duplicate letters and stretch those letters.
Fun fact, that's why old 's' are sometimes written with an 'f' missing the bar. Because they were re-using forms where otherwise a letter was so common they'd run out!
You’ve clearly never had to deal with a lot of references. Author lists can cluster. If you’re reading, it can get tedious if the last 5+ references are to the same three authors. Let alone 20+ references.
This happens. In some writing it happens _a lot_.
It’s a common enough that it’s even explicitly mentioned in the Chicago Manual [0], one of the main style guides for academic writing in the US:
> Em dashes also substitute for something missing. For example, in a bibliographic list, rather than repeating the same author over and over again, three consecutive em dashes (also known as a 3-em dash) stand in for the author’s name.
[0]: https://www.chicagomanualofstyle.org/qanda/data/faq/topics/H...
For the reader… I dunno, do people actually read through a citation list, in a manner where the redundancy would be relevant? When I look, I’m usually specifically looking for the single entry that corresponds to the citation on some claim in the text.
Although if someone really wanted these triple-dashes for their journal for some reason, I bet some tex could be coded up that would generate them. Could look neat, maybe.
If not, I doubt some typesetter in the 19th century (or earlier) was so drunk with power when setting out the lead type to make the 3-em dash.
EDIT: Additionally, unicode primarily exists to encode the typographic symbols that exist—that is to say, it's a catalog of what already is. Take Ogham[1] for example.
1: https://en.wikipedia.org/wiki/Ogham_(Unicode_block)
I mean the triple-em-dash has been around since Unicode 6.1; The Tironian Sign Capital Et wasn't added until Unicode 13.
It was like pre-AI poor man's artisnal LLM hallucinations.
There need to be chat discussions like that for every unicode character.
Imagine a discussion on how many em dashes should be used that mimics the discussions on story points.
Here's a googelensed bit of the index of a philosophical work from the thirties. The entries after 'Aristotle' are all megadashes, but google lens reads them as hyphens. Under 'Kant' the subheadings have subheadings so the typesetter doubles the triple-dashes up without however employing a six-em-dash. Current taste I think would just use indentation for all the purposes mentioned they are all line-beginnings.
Activity, 34-6, 99, 157.
Actuality, 343-4, 348.
Aesthetic experience, 15, 156, 339, 346.
Anselm, 232, 233-4.
An sich, 46, 84.
Appetite, 252-3.
Aristotle, vi, 3, 26.
- Activity, 35, 157.
- Being, 33-4.
- Categories, 34.
- Cause, 149-50, 155, 249.
- Demonstration, 186, 209-10, 279, 282, 354.
- Division, 279.
– ἐπαγωγή, 212, 225-6.
...
However, when encoding them into Unicode, the consortium suddenly discovered the importance of being frugal with character IDs and only defined a single "Han" sequence of characters for all three scripts. You need to select a matching font to actually display the text in the script that's intended.
This makes using those characters an unnecessary hassle and also came over as a cultural faux-pas to some (essentially like encoding Greek Eta, Latin H and Cyrillic En as the same character, because they look similar and the scripts have a common history)
It also looks a bit hypocritical when Unicode otherwise goes the opposite way and has the space to encode dozens of variants of the Latin alphabet or — as we see here — several categories of dashes.
They can be used as adhesives, abrasives, laminates, passive fire protection, personal protective equipment, window films, paint-protection film, electrical insulation, car-care products, electronic circuits, optical films, Scotch tape, Post-it Notes, labels, dispensers, and magnetic tape.
In German, nouns have gender, so why not more inclusive punctuation? The him dash, her dash, it dash, and the nonbinary them dash. It would really infuriate MAGA if you could apply emoji gender and skin color.
And of course if you really want to make a statement and stick it to those anti-AI em-dash nit-pickers, there's always the Zen-like struck-out no-break space (U+00A0 NO-BREAK SPACE + U+0336 COMBINING LONG STROKE OVERLAY) ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ you can combine them to any length!
It's a shame Hacker News filters out the really good ones, like the üm dash: — (U+2014 EM DASH + U+0308 COMBINING DIAERESIS) and the over-accentuated üm-á-ñå-ḍ-ç dash: — (EM DASH + DIAERESIS + ACUTE + TILDE + RING ABOVE + DOT BELOW + CEDILLA).
@dang I know @mentions don't work so you'll never read this, so screw you for filtering out my favorite glitchy punctuation diacritics, you jerk! I feel like my creativity is stifled and freedom of speech suppressed!
https://youtu.be/ZEJBBlRlhAU?t=251
I like using commas for spacing though em dashes do the same thing and avoid the run-on police?
"There is no greater joy in life than typing a sentence, realising it needs a clarification, a caveat, an addendum (in the same sentence, not a new one) and, guess what, parentheses are right there for you."
That's what I'd do. Note the conversion of the first comma to "and," as well.
I find m-dashes easier to read for most other purposes. They’re nicer in longer asides because they’re a little more visually prominent, while also (for some reason) not breaking my mental flow quite so much. They’re far nicer looking when an aside comes at the end of a sentence, as you don’t need to close an m-dash. That always looks like crap with parenthesis, as if the author should have thought of a way to avoid doing that, while with the m-dash there’s no problem at all.
(Unrelated: The alt-code for em-dash on Windows is Alt+0151. I just typed my first manual em-dash wow exciting.)
I definitely prefer it un-spaced, though, personally.
> And also occasionally — occasionally — one just needs a random dramatic pause.
> And also occasionally (occasionally) one just needs a random dramatic pause.
I know which one I'd rather read. Though, frankly, I'd prefer neither and I'm trying to remove this kind of thing from my own writing.
In a first example, the second "occasionally" is stressed, when in second example it feels almost whispered.
One advantage of my ESL english is that the text I produce doesn't look that much AI generated anyway…
They help convey that I am so smart—and I think so fast—that I have to constantly interrupt myself.
But—and I think this is important—it’s really easy to overuse them.
But seriously, trying to express the spider web of thoughts in my head as a linear string of words is really difficult for me and that doesn't make me feel very smart.
I use dashes in place of commas, semicolons, colons, and brackets all the time - I just use the standard keyboard dash. Why the heck should I bother learning some stupid key combination to make the dash longer? I highly doubt that the vast majority of readers even know that there are different dashes with different uses. When I learned about it, I decided to be a conscientious objector because using - vs — and assuming I can tell the difference is inane (because it is font based).
(ironically, the HN text editing dialog shows - as ~.8 of the length of — whereas on the site it shows as ~.25 of —)
For the poster child of a naïvely set/composed text which suffers thereby see:
https://blog.lostartpress.com/wp-content/uploads/2025/10/vir...
That a book about a perfectly made tool cabinet suffers from so many typos (the maker's name, Henry, is mis-spelled on the inside cover) and such pedestrian composition embodies much of what went wrong when the typesetting unions all bargained for sinecures rather than active participation in the Desktop Publishing Revolution.
Personally, I think em-dashes (when in a font that shows the damn difference) look exaggerated and silly. They certainly don't provide any important guidance to its meaning that I can't derive from context.
That was a cool toolbox BTW :)
Yes, the Studley tool cabinet is amazing, which makes the infelicities of the book all the more enraging.
https://blog.nawaz.org/posts/2025/Dec/a-proclamation-regardi...
While it technically won't be grammatically correct, it'll get the benefits he describes in the article while also not being accused of being AI-generated.
//! This is mission critical, don't remove...
As a reader, I dislike seeing em dashes almost everywhere.
I was never a big reader as a kid, but I fell in love with the New Yorker, much later in life, where I kind of adopted their idea of: "Once you understand them, you can read through long sentences very easily". That is the newyorker style of writing. One falls into two buckets, you love that kind of writing or you hate it. If its the latter, you should adopt em-dashes and commas to improve your writing style.
The funny thing is at both extremes, either none or indiscriminate use of dashes like in TFA, is a canary of "human writing".
People (including comments here) often attribute LLM use to prevalence in training data but it is probably far less common in the wild than we are led to believe, the more likely scenario is that whatever text-to-text normalization tools used ended up inserting whatever their idiosyncratic typographic styles were.
And on Windows it's Alt + 0151 on the numeric keypad, which means you a) must have a numeric keypad and remember the number, or b) must use the emoji search. And on Linux it's weird too.
Em dashes are therefore Mac user garbage. The correct replacement is --
Great accessibility.
huh?
Oh - you meant iOS. I don't see a need to differentiate dashes there either...
It's worse than that. What they are saying is the equivalent to saying that knowing how to do something properly, and having the attention to detail to do so, makes it obvious it was written by AI. It sounds like a prelude to human incompetence, to ensuring that AI products are better than human ones.
About the only semi-convincing argument that I've heard is that AI over uses these tells. But let's face reality here: a lot of people griping about AI generated or AI assisted writing are not looking at over use. Plus a lot of people tend to over use the said tells while they learn how to use them properly.
[1]: I'd even go so far as to extend that "anyone" to: "anyone creating, even with AI assistance!"
That said I've only been on Linux for 25 years so I still haven't figured out how to actually get em dashes and have to resort to the double minus like up there.
As a few have pointed out, many writers have been using it usefully since long before the blight.
The current state of revolt is probably a phase though, as models will learn to account and adjust for the litany of literary crimes they've been committing. In some ways we are fortunate to have such easy signals to work with while, already, false positives are becoming almost as common as actual positives, indicating it might not be long before we just can't tell anymore. Which would suck, but then we'd have to adjust the nature of our complaints to something more profound, something more... insidious, something.... Maybe Woodrow Wilson? [0]
It's not only the emdash that threatens the quality of our dialog and linguistic welfare — it's many principles, such as the the Rebuttle>Reframe>Prescribe structure of this very sentence. Once is a lot, but multiple instances of this structure in a single comment would be worse than swallowing a double hyphen sideways.
The LLM is methodical, deliberate, unhurried — in a way that rewards the vigilance, attention and patience required to spot compulsions, such as the Rule-of-Three seen here.
It's important to catch and worth noting each instance of filler-transition that adds little information while working toward word-counts.
Undoubtedly, if we want to make our world safe for the emdash once again, we certainly should not overlook overused intensifiers posing in places where real work should be.
While such examples can be found in legitimate human-written texts, they can also be found in LLM generated texts, and this Reflexive false balance style sentence is suggesting we just can't always be sure.
Consider a person suffering generic example syndrome, where instead of naturally recalling a situation from actual memory, an example might be synthesized from some statistical category, seeming a bit out of place.
And do not forget sentence length clustering around a mean, which humans have a tendency to naturally avoid.
Frictionless surfaces might also hint the involvement of LLMs, where the final structure may be suspiciously clean of normal human quirks.
Of course, the ultimate question here isn't even being asked, and the emdash is not a registered trademark of LLMs. It's not about whether we should fear using them, or even their value — it's about this Symmetrical closing that despite being written by a human (maybe subhuman) has strong patterns of an LLM.
0. “Since I entered politics, I have chiefly had men's views confided to me privately. Some of the biggest men in the United States, in the field of commerce and manufacture, are afraid of something. They know that there is a power somewhere so organized, so subtle, so watchful, so interlocked, so complete, so pervasive, that they better not speak above their breath when they speak in condemnation of it.” "
The overemphasis by critics on punctuation and "this not that" is tiring. I don't really know how to describe LLM writing but it has the same qualities as flat soda. Some authors you feel as though you fall through their work with little effort; unable to stop until gravity slams you into their conclusion. LLM writing can feel like an uphill hike to a sewage treatment plant. That is, its easy to convince yourself to stop at any time. There's a positive quality to that. I hope AI labs don't find a way to remove it.
has anyone used/seen em dashes in file/dir names? just wondered — I have't.
It's certainly easier to read than a hyphen/minus...but i guess it may become problematic cross-platform (on windows/linux it isn't as straight forward as on macos).
Fedora with XFCE here, but last I checked on Windows, it worked exactly the same over there: With a Chinese IME installed, tap shift to switch input language, shift+hyphen produces a double em-dash: ——, likewise, shift+6 a double elipsis: ……
You encounter those a lot in simple, machine translated Chinese, along with literal translations of "not X but Y", "not just ... but also" and similar sentence constructs that are common in Chinese. I suspect that's how a disproportionate amount of that ended up in training material in the first place.
Typesetting aside, the em-dash as a grammatical construct is indeed pretty great.
Some people even take pride on being able to detect (often falsely) AI generated work.
I only ever see them with an AP-style spacing before and after.
in any case, i mostly agree with the author that ai has co-opted too many things to be concerned about it anymore. i remember being linked a list of "1,000 words not to use" because ai used them a lot (e.g. "delve", "absolutely", etc.) and the venn diagram with the 1,000 most common english words was practically a circle. you would be unintelligible if you rigorously avoided them all. now i just write without care of my ai-similarity score.
You could write naturally and then let an LLM translate avoiding the 1000 words, with verification harness or maybe a custom sampler.
Also, cosmetically I don't really like how em-dashes are used without spacing (e.g. "this—then that" instead of "this — then that"). Feels like we're trying to import publishing styles from newspapers and books heavy on serif font to places where monospace and sans-serif is the norm and it doesn't quite fit well.
> Em dashes can be used instead of brackets in a sentence — like this one for example — and you can just type your shit and keep going. Brackets are for people who raise their hand in meetings and say: “this might be a stupid question”. A parenthetical is whispering: “Sorry, don’t mind me, I’ll be quick”. Em dash kicks the door open and announces itself. Use brackets when you are embarrassed by your own thought.
Literally had me laughing out loud.
And then this bit:
> As for semicolons, semicolons are neat, they are like the coolest punctuation mark after the em dash — the fundamental indecision compressed into a single punctuation mark — you are neither starting a new sentence nor you are, exactly, continuing an existing one. You are just… piling your stuff together. I am always happy when I find a place for the little guy semicolon. But, again, most cases where you could use one are better off served by the em dash as well.
couldn't help but remind me of the following Kurt Vonnegut quote:
> Here is a lesson in creative writing. First rule: Do not use semicolons. They are transvestite hermaphrodites representing absolutely nothing. All they do is show you've been to college.
:D
unfortunately microsoft (specifically outlook) and libreoffice have chosen to replace them with a proper em dash even if you turn off autocorrect which is bonkers-levels of infuriating.
if i turn off something it /should be turned off/, not just be kinda turned off; y'know?
PS: if you want personalised advice from me on how to use punctuation correctly, you can now get it for $175/hour.
humans sometimes ...
ps: sasha, if you want me to read your future regurgitations you can now get that for €175/minute (limited availability, only skimread mode for now)
If I'm writing something, typically I want to share my thoughts with other people and get their feedback. It won't work if they dismiss my writing because of some superficial stylistic choice, without even trying to engage with the ideas that I'm trying to express.
When I first learned about em-dashes, I was told they were a mark of sophisticated, intellectual writing. But now people say they sound like AI. Is that really a bad thing?
It's strange to see sophistication become low-class just because AI uses it. It makes me think that form is really about who the author is, not the form itself.
In that sense, I'm not sure what it even means to be 'human-like.' Being human-like just feels like an illusion that people who want to believe in it keep insisting on
We're experiencing broad scale changes in the sentiment of style in written English. I don't know what the best approach to take here is necessarily, but the more I see language evolving over time (especially over the past two decades) I think I'd prefer to shift my habits in line with whatever the "modern standard" is in order to avoid falling into pitfalls, i.e. being accused of being an AI writer.
The old school of writing it in digital documents. Typesetting predates and is not limited to digital documents.
To my knowledge (and what I meant by "old school"), it was common to type a "--" dash on typewriters[0] because most of them lacked en- and em-dashes, which are obviously not digital documentation devices.
But yes, typesetters of old (old, old school!) had myriad glyph blocks available to them for pressing!
--
[0] - https://en.wikipedia.org/wiki/Typewriter#Typewriter_conventi...
The idea that a double-hyphen is "old school" is on par with the false notion that double-spacing between sentences is a product of the typewriter; one need only look at a typeset book from before the twentieth century to see the truth, or look at a decent image of the Declaration of Independence or other hand-written document from prior centuries to see the extra space left between sentences.
Many of the changes we see in written English style are the product of ignorance of the past beyond two generations or so. Avoiding em-dashes to keep up with the latest anti-AI fad is just reacting mindlessly to the stimulus of popular ignorance instead of deliberately situating one's style in the history of writing conventions.
Agreed that in terms of precedent, typesetters had moveable type dashes for their presses.
> the fifteenth century, and "em" and "en" were standard printing widths in moveable type. People who write CSS still recognize em as a width.
This is very interesting!
> We're experiencing broad scale changes in the sentiment of style in written English.
Lol no we're not. The only people who give a fuck about this em dash AI bullshit are losers on Tumblr and commenters on fanfiction sites who are obsessed with the idea that AI is the current Great Evil.
This is the orthographical equivalent of the New York Times running an article that soy-glazed avocado bouillon is the latest trend among Millennials because, like, one of the editors caught their college dropout son eating it one morning at the family table. Son said "duh dad everyone's doing it" to fuck with his dad, and next thing you know, a hundred think pieces on Mexican-Japanese breakfast being trendy.
I've never met an adult in my professional or personal life who actually gives a shit about this. We're talking a fraction of a percent of people who don't want you to use AI, and a fraction of that who waste their life trying to detect it.
Write the way you want. If people can read it, that's all that matters.
Apologies for the French. Blame Canada.
*Edit* Also the em dash is based. Using a double dash should be a prison sentence.
This is largely because most people are incapable of thinking critically and evaluating substance. So they regress to indirect markers of good writing, and the whole thing becomes a game of signaling.
The rot is so deep, many popular resources on writing focus more on style and presentation than focusing on the substance of writing. And even when best-intentioned, all this does is make poor writing a bit more palatable by giving it a veneer of polish.
Guides such as S&W focus entirely on style. Then we have the dearth of guides on composing a good sentence.
Of course, S&W is also filled with insanely stupid advice. "Don't split an infinitive" immediately comes to mind. Ask yourself which sounds more powerful, oh fellow geeks:
* To go boldly where no one has gone before
* To boldly go where no one has gone before
Split infinitive form is so much better. "Don't split an infinitive" is one more smooth-brained 19th century rule invented by Latin-obsessed weirdos who thought it was the perfect language, and that they should impose Romance grammar on Germanic languages. (See also "don't end a sentence with a preposition" and "don't start a sentence with a conjunction.")
I agree that post-COVID we have a glaring issue with the problem you highlight, but I don't think the focus on orthography is a symptom. Long ago, my dad criticized my résumé because I didn't put two sentences after periods despite me using a variable-spaced font. The man's nearly eighty. He's not exactly some 21st-century miseducation case.
One caveat I'll add at the end is that focusing on form isn't always bad. Form is one of the great joys of reading both poetry and prose. Word selection is often form rather than substance. The opening lines of Lolita are pure form. Nabokov could've easily just said "I'm a pedophile and Lolita is attractive to me." Instead, he chose words that have you overuse your tongue when you sub-vocalize, making the text feel uncomfortably erotic to a normal, non-pedo reader.
No, with AI it's because it's MUCH faster than evaluating substance, and fast ways to weed out AI slop are desparately needed, since it can be produced so quickly.
It's the thing it was trained to do. The entire decades-old architecture has exited for specifically that purpose. That second "L" in "LLM" is "language" for a reason.
The combination of both I usually find in text generated for me by LLM, and so far, it serves as a good signal for density of information and, inversely, not-quite-right ideas and their connections. (That's not to say it's pervasive, but below the surface, the quality of LLM text as a medium is often not great.)
There may be well-written text of non-trivial length generated by LLMs, but so far I haven't knowingly encountered any, with or without em-dashes. Human editing remains very necessary.
If I post some information, then I have had people accusing me of cribbing it off Wikipedia. When I have translated from some languages, people assume I'm using Google. Even though in one case, the language in question wasn't even available on Google Translate. (I went and checked.)
https://www.mcsweeneys.net/articles/im-comic-sans-asshole
Turn off smart punctuation.
Saves bytes.
(sorry for a bit of gaslighting)
Hell yes, and ROFL. I’m tired of em dashes being cited as evidence of AI.
There are actually four different "dash" type things in TeX: a hyphen, an en dash (for ranges like "10--20 pages"), em dash (---), and the minus symbol ($-$).
https://www.pbjstories.com/blog/spaces-around-em-dash
I beta read for a popular american author and they started using spaces too.
Of course I recommend do-not-adopt, think Roman-cart-wheel-axle-width reasons.
At least, that's how I learned it. Everyone seems to disagree and write em dashes with spaces nowadays.