- 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...
I think this is the real reason for its (pre-LLM) ubiquity - it's a really flexible, versatile bit of punctuation, whereas most of its siblings have a more specific meaning or purpose.
A stream of short sentences often feels lazy, reductive, and tedious. But it is more forgiving--you can't fail if you don't try. And it inflates word count, similar to verbose Java vs compact Perl.
The New York Times Manual of Style, for instance, only permits using an en dash as a minus sign. Joining nouns like Sarbanes-Oxley and even number ranges use hyphens instead.
The BBC News Style Guide and the Economist Style Guide similarly use hyphens in place of en dashes - though sometimes BBC seems to use en dashes in place of em dashes.
The University of Cambridge Style Guide simply says to avoid dashes altogether.
The AMA Manual of Style explicitly declares that one should hyphenate two or more nouns of equal importance. They however continue to use the en dash when the first word in the compound modifies more than one word (e.g. anti–Norwalk virus).
The US General Printing Office Style Guide also uses hyphens, so the Sarbanes-Oxley Act is hyphenated in documents published by Congress.
The Chicago Manual of Style nearly got rid of the en dash as well, but backtracked after backlash.
Ceasing the happy tradition of using the hyphen between equal-weighted nouns because of a hacker news comment is the moral equivalent of adopting expletives from the kids menu at a diner.
> 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.
(Not my intent to be pedantic.)
This is not the best source (I’m pressed for time) but it gives a decent explanation: https://editorsmanual.com/articles/en-dash/#elements-of-equa...
While there's no central authority, both Oxford and Cambridge University press style guidelines recommend that.
my favourite (and I hope LLMs never converge to use it)
I checked a message from my last chat with Claude
> Noted for this conversation — though worth flagging
and that's an em-dash (unicode 0x2014)
– Robert Bringhurst, The Elements of Typographic Style
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?
Konnichiwaー
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 pre–Websterburg Bakery era
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"
But there's also hyphen-minus: https://www.compart.com/en/unicode/U+002D, and a search for 'hyphen' returns a bunch of others: https://www.compart.com/en/unicode/search?q=hyphen.
- − (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...
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.
Imagine a discussion on how many em dashes should be used that mimics the discussions on story points.
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!
Seriously? You're claiming printers invented shorthand and abbreviations to make working with movable type easier? They were common in handwriting and predate the printing press. Look! Even the Wikipedia page for abbreviation has an illustration from scribal handwriting: https://en.wikipedia.org/wiki/Abbreviation.
> 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!
No, that's from handwriting, invented in the 8th century (https://en.wikipedia.org/wiki/Long_s). Go read the original handwritten copies of the US Constitution or Declaration of Independence. You don't have to read far until you find a long s:
https://en.wikipedia.org/wiki/United_States_Declaration_of_I... (necessary, 3rd line/1st line of cursive)
https://en.wikipedia.org/wiki/File:Constitution_of_the_Unite... (bessings, 2nd line)
In conclusion, everything in your comment is totally wrong.
The bar has to be set somewhere. How many people actually use this thing? We can't go introducing characters into standards just because six people with power want some claim to fame.
I for one wasn't even aware that Unicode characters could be so large. Did they have to make a special exception for this or are there other characters that are that wide?
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...
That's not what I said or what I intended to imply.
I think it's stupid because It sounds like the thing that I would assume almost nobody encounters. Like so few people that this is not needed. If more than 10,000, people have ever typed this thing out for its intended purpose, I stand fully corrected.
The line has to be drawn somewhere to prevent garbage from making its way into open standards.
How many people actually use this thing? 10 loud people with power using it does not justify putting it into the open standard.
This isn't some some language that thousands of people speak. To me this seems like some weird thing that some drunk academic came up with and told his friends about.
Also, this thing is freaking huge. How many characters are this wide? Should the bar be higher to introduce obviously weird characters?
Given that the term is explicitly mentioned in the Chicago Manual of Style, I suspect that this is much more of unicode trying to capture existing behavior, even if rare. And I think it's important to err on the side of capturing what people did, even if it's from an bygone era.
> How many people actually use this thing? 10 loud people with power using it does not justify putting it into the open standard.
It's definitely more than 10. But even if it's 10, I disagree with your assessment. People used this; it's worth a codepoint. There are far stranger code points out there.
And I've seen far weirder things in open standards. Have you read the C99 standard? Are you aware of the three allowed integer representations? Of which maybe two were in use in 1999, and one only in use in relic hardware that might exist in maybe tens of sites.
That assumption is incorrect. There are a lot of people doing academic research and of those people the vast majority of them encounter this situation a lot. This is like insisting that a strap wrench is stupid because most people don't ever need one.
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.
As other people have commented, it's partly a relic of an era where things were typeset by hand. Some journals may also require this in their style guide; I don't know of any offhand but the list of journals is long and their choice of style is varied.
> I dunno, do people actually read through a citation list, in a manner where the redundancy would be relevant?
Yes, they do. When you work on a Ph.D.-level dissertation, you end up having to learn your subfield pretty well: its major works, open questions, current major players, their ongoing arguments, etc. You start to notice citations cluster, and you can often find more than one opinion group. People you have to listen to (ie: your advisor / committee) will hand you articles with authors you don't know and who may come from labs that you don't know. One common way to start orienting yourself is to look at who they cite.
You pick up patterns: some people love to cite themselves and their friends, some people really try to give a detailed overview of the state of the field, and some people go out of their way to attack their rivals via citation. Sometimes the reference list has an author or work you don't recognize, and might have missed while glancing over the text. Mostly it's something mildly relevant and occasionally it's an exciting direction to look into.
The really long citation lists, where the triple emdash would make sense, mostly occur in fields outside science and engineering where knowledge is exchanged through chapters and books. Some fields expect a lot of detailed textual analysis which can involve a lot of careful citation of the same chapter or book over and over again.
While I've never worked in these fields, I've been friends and roommates with people in them, and I quickly learned that bibliographies are way more of a thing for them than I ever imagined.
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 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.
Counterpoint: the people who use tools the most should be the ones designign and creating them. All you have to do is continue not using it just like you did before it existed, but instead you insisted that its existence make sense to you as someone who doesn't need it and doesn't do what it's meant to help with.
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)
<s>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.</s> (Edit: Han unification was done when Unicode was still 16 bit, so there were real space constraints — so no hypocrisy actually)
Really, they needed to come up with the idea of variation selectors way earlier, and use them aggressively. And design something more consistent for character composition, and skip the precomposed characters (those really are glyphs representing a cluster; being able to treat the cluster as an atomic entity versus its components doesn't add actual semantic information to the text).
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!
Triple emdash cannot be used as a lubricant. Do not touch, inhale, or look at triple emdash. Do not taunt triple emdash.
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.
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.
> As for semicolons, semicolons are neat, they are like the coolest punctuation mark after the em dash
All that energy spent on punctuation, and still: comma splices. ):
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.
As a reader, I dislike seeing em dashes almost everywhere.
//! This is mission critical, don't remove...
https://www.nybooks.com/articles/1993/11/04/survival-of-the-...
I have always loved---and used---the em-dash, and after reading this essay in The Best American Essays 1994, sometime around 1995, I finally felt justified in my affection.
I have always loved—and used—the em-dash [...] ;)
They also like to use em-dashes excessively, especially in bulleted lists, to separate a topic from a comment in each list item.
Avoid those and you're fine. They aren't good writing style anyway. Used well, em-dashes are good!
I have a question to all em dash users reading this - why wouldn't I just use a regular dash in place of the normal full-length character? I have already done this several times in this very comment, and it's a very common convention on the internet. Outside publishing and other types of very high-brow writing, the fifty different variations of dashes with minor differences in length are pretty pointless, in my opinion. There's almost no scenarios where you need to make a distinction between them, what could possibly be misinterpreted because of dash length? So why not just use the regular, standard dash everywhere?
More importantly though, the em-dash-in-a-sentence construct is a way of thinking and the notation is of lesser importance than the way to express oneself. I dont think the current distrust is from the symbol: the LLMs love em dashes for the same reasons many of us do, and, perhaps better than most of us, they have no trouble using the correct symbol.
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.
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!"
And it really couldn't have been anything other than the em-dash because I've been deleting the account's comments when they were about a week old so there were no previous comments to draw any conclusions from.
I don't care though, I've been using em-dashes since the late '90s making me partly responsible (along with all the other em-dashers) for LLMs being so fond of them, so I'm not going to stop just because of them.
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.
Great accessibility.
huh?
Oh - you meant iOS. I don't see a need to differentiate dashes there either...
Sometimes adding spaces around em dashes is not about preference, but helpful due to the font, or the device, you're reading on. At some zoom-levels, for example, an em dash may bump into an italicized word unless there's a bit of extra space.
One advantage of my ESL english is that the text I produce doesn't look that much AI generated anyway…
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 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.
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.
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.
Some people even take pride on being able to detect (often falsely) AI generated work.
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.
Profanity for the sake of profanity is lame and immature.
oh no!
You could write naturally and then let an LLM translate avoiding the 1000 words, with verification harness or maybe a custom sampler.
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).
Typesetting aside, the em-dash as a grammatical construct is indeed pretty great.
I only ever see them with an AP-style spacing before and after.
I want to think it's the latter, but it's so exquisitely stupid I keep thinking it's "stupid on purpose".
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 --
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.
Endorsed. This post gets my vote. Two thumbs up vote.
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)
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
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.
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.
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.)
this is several sentences bound together in a rat king and furthers my theory that the use of more exotic punctuation like em dashes is the linguistic equivalent of a code smell. It may do the job, but it strongly indicates that your writing would benefit from rephrasing the entire sentence.
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.
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.”
Wilson went a bit farther than might have an LLM. But who knows, the future may bring new rules, perhaps even a rule of 6.
I'm annoyed that so many people limit their analysis to shallow things like em dashes and "delve," and don't seem to have the critical thinking necessary to analyze textual construction, let alone actual meaning. Instead it's all cutesy remarks about how Emily Dickinson must have been an AI.
> 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
Turn off smart punctuation.
Saves bytes.
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?
(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.
edit: Another comment recommended https://www.pbjstories.com/blog/spaces-around-em-dash which does an awesome job of diving into dash usage by style guide
[^1]: https://will-keleher.com/posts/your-dashes-suggest-which-mod...
Hell, overuse of emoji is more of a tell than the mere presence of emdashes. The problem here is really "AI detection" services that themselves use AI to perform the service. More than one person has fed pre-AI text to such a detector only to be told "this is 75% AI" or some shit. Students who didn't use AI have been near harassed by their teachers because the teacher used some kind of AI-based AI detector, and it hallucinated. People want to believe the computer because they think "oh the computer is objective" and throw out all logic afterwards. Unfortunately, the rise of AI has brought subjectivity to computers, in the worst possible way.
You see the same thing with the family who got accosted in California by multiple police vehicles coming out of a retail store, because somebody in New Jersey lost a dealer tag and entered a partial tag number in the report, which ended up being fed to Flock's search engine, which then provided the partial search results it was asked to. Everyone along the way trusted the computer as if it was the single source of truth, but that "source of truth" was tainted by incorrect input. Literally GIGO on a life-altering scale. Other people were already being watched. It was just probability, to the point where that family probably had equal probability to win the multi-state lottery jackpot.
Go watch the Computer Chronicles episode on AI. It's very prescient. The main issue is that currently, people treat AI as if the problems brought up in that episode have been resolved, when they haven't at all.
Use of AI is fine as long as there is human review along every step of the process. Anyone using AI as a shortcut to actually doing the work they're paid to do needs to be removed from whatever situation they're involved in at the earliest convenience. Otherwise it's just an expose on the fact that humans are somehow intrinsically lazy, despite the outliers of the hardworking person (which does exist, I've known people who would move mountains to make what you need done possible, and the rest are just pivoting to AI to fake it).
Fake it until you make it, I feel, is the core of modern "AI". People also forget that AI as a field is much older than modern generative AI.
People have been running from their origins for decades. To me, this desire to "take human fault" out of the picture via AI is no different than centuries of people insisting that humans are separate from and not part of nature, even though our bodies run under the same physical and biological rules that all of nature does, and that is unavoidable. Objective reality only exists as an underpinning. Nothing that lives on this planet can ever perceive that reality without involving their own personal bias, and so it exists with the same relevance as an Earth-style planet even five lightyears away with an active civilization (if such a thing exists). I think that value might even be generous. There could be such a civilization a single lightyear away, and we'd never be able to reach them without a generational vessel designed to support at least one generation of explorers dying, and relying on their children (or grandchildren) to see it home.