Top
Best
New

Posted by joozio 15 hours ago

I am definitely missing the pre-AI writing era(www.lesswrong.com)
198 points | 169 commentspage 3
thepasch 13 hours ago|
I never use an LLM to paraphrase my own voice as a matter of principle, but I’ve still been repeatedly accused of doing so because I happen to always have written structured posts, used “smart quotes,” and done that negative comparison thing (it’s genuinely not just fluff, it’s a genuinely useful way to— ah god damn it). Sigh.
gjm11 11 hours ago||
Right. The LLMs' quirks aren't bad in themselves, they're bad when they're in every damn paragraph. They're mostly things that in moderation actually improve writing, and that if you see them once (without the knowledge that they're things LLMs do) would rightly tend to make you think better of the author. And so, of course, in RLHF training they get rewarded, and unfortunately it's not so easy for an LLM to learn "it's good to do this thing a bit but not too much.

The structured thing you mention is the one that bugs me most. I genuinely think that most human writing would be improved by having more of the "signposts" that LLMs overuse. Headings, context-setting sentences, bullet points where appropriate, etc. I was doing "list of bullet points with boldfaced intro for each one" before the LLMs were. But because the LLMs are saturating their writing with it, we'll all learn to take it as a sign of glib superficiality and inauthenticity, and typical good human writing will start avoiding everything of that kind, and therefore get that little bit harder to read. Alas.

beej71 8 hours ago|||
I refuse to cater to the "em dashes are AI" crowd.

And I was just noticing that my home-built blog render pipeline produces dumb quotes and that was embarrassing to me. Needs to be fixed.

(Counterpoint, dumb quotes are 7-bit clean and paste nicely... Hmm.)

BeetleB 3 hours ago||
> I refuse to cater to the "em dashes are AI" crowd.

I wrote a plugin for my blog that converts all hyphens (surrounded by whitespace) into em-dashes.

https://blog.nawaz.org/posts/2025/Dec/a-proclamation-regardi...

Freak_NL 12 hours ago|||
I feel ya. I've never been accused of using an LLM, fortunately, but depending on the context I do use “smart quotes” (even in „Dutch” or »German«) and the em-dash obviously… (And that ellips fella there. It's just so simple to type with a compose key set up.)
shagie 5 hours ago||
I thought the guillemet was French rather than German and the other way around.

https://en.wikipedia.org/wiki/Quotation_mark#Summary_table

Freak_NL 4 hours ago||
German uses both kinds depending on the style and writer's preference. French has the guillemets the other way around.

(That Wikipedia table shows that too by the way.)

tomku 3 hours ago|||
It's absolutely shocking how many people think that inverting all the quality metrics that we've traditionally used "because LLMs" will lead to good things. Nothing about this will end well.
internet_points 12 hours ago||
Same here, I've always used em dashes and have been called out on negative comparisons – I didn't even know they were an LLM thing. Should I read more LLM to know what phraseology to avoid, or will doing that nudge me towards sounding more LLM? :-(
keiferski 13 hours ago||
I have been writing stuff for a long time; my first internet experience was posting on forums about a Gameboy Advance game. Then in other forums, for a philosophy degree, and professionally as a copywriter and technical writer. I’ve been meaning to write up a post of my thoughts on writing and AI, but there things I’ve been thinking recently are:

1. There was a lot of slop pre-AI. In fact I’d say the majority of published writing was bad, formulaic, and just written to manipulate your emotions. So in some sense, I don’t really think pre-AI slop had more value. It’s just cheaper to make now.

2. AI has prompted me to study more off-beat writers that followed the rules of language a little less frequently. This includes a lot of people from circa 1890-1970, when experimenting with form was really in vogue.

3. Which brings me to my third point, which is that no matter how much the AI actually knows about writing, the person prompting it is limited by their own education and knowledge of writers. You can’t say, “make me a post in the style of Burroughs” if you don’t know who Burroughs was, or what his writing style was. So in a sense there is an increased importance to being educated about writing itself. Without it you’re limited in your ability to use AIs to write stuff and in your awareness of how much your non-AI written work is influenced by AI writing.

CrzyLngPwd 8 hours ago||
I've been a Grammarly customer for quite some time, and I have tried the AI suggestions, but it always loses something and ends up with a whiny, apologetic tone.

AI always seems so verbose and wordy.

ChaitanyaSai 4 hours ago||
I am sorry but perhaps some use of AI or grammar-check would help? A lawn that's not overly manicured has its charm, but if it has one too many barren patches of clumps of overgrown grass, it doesn't appeal as much? This essay feels a bit like that.
amelius 13 hours ago||
Are there any good writing LLMs out there?

I get that the mainstream ones have been RLHF'd to death, but surely there must be others that are capable?

shaoner 13 hours ago|
https://hemingwayapp.com/ gives you advice about your writing.

This is called Hemingway because he was apparently good at communicating efficiently which made him a popular author.

amelius 11 hours ago|||
What happens if you take the output of a mainstream LLM and send it through this app? Would that solve the issue of the original article?
Kye 9 hours ago|||
This is an interface, not an LLM. Do they say which LLM they use? Many of these are interfaces to one of the big three model providers. Others run through OpenRouter to use one of the better open models, all of which have their own quirks.
joemazerino 3 hours ago||
I'd push back on the author and ask him really if his writing is getting worse or his standards have increased, leading to undue stress that might throw the flow state off.
bananaflag 13 hours ago||
Can't you just... not do this?

I never passed any AI writing as my own. I would feel utterly awful. Also, I love tweaking words until they sound perfect.

The number of people who just nonchalantly admit that AI writes their messages is honestly scaring me.

skywhopper 12 hours ago||
It’s largely a problem of how these tools are packaged, but while it’s certainly nice to have an LLM check your spelling, or review your grammar or style or usage, you should never allow them to actually edit your document directly.

First of all, they will make substantive changes you didn’t intend. The meaning will get changed, errors will be introduced. Tone will be off, and as the author says, your voice will disappear. There is no single “correct” way to write something. And voice and tone are conveyed with grammatical and usage variation. Don’t give that up to a robotic average.

Secondly, you will never improve, or even maintain, your own writing skills if you don’t actively engage with the suggested changes. You also won’t fully realize half the purpose of writing, which is to understand the topic better yourself. Doing the work of editing your piece will help you understand the subject even better. If you just let the machine “fix” your errors, you’ll become a worse writer and less of an expert over time.

nsxwolf 4 hours ago||
Once I think something is AI I just can’t read it anymore. It isn’t out of principle or anything, I just become so distracted by the idea that I can’t focus or derive any benefit or pleasure from continuing.
globular-toast 1 hour ago|
This writing is terrible. I can't read it. But are people really unable to write without wanting to put it into an LLM? I haven't done a single piece of natural language writing with an LLM. The thought has never even crossed my mind. Why would I? Surely to give the LLM context of whatever I want to write about would amount to, you know, writing it down? Just write that "prompt" in your blog and send it. No need for LLMs.
More comments...