A post on X last week:
"The fact that people can't even reply to posts without AI anymore says a lot more about them than they think."
The repl...
For further actions, you may consider blocking this person and/or reporting abuse
Authenticity still exists and it existed before. And slop existed before AI for sure, it's what AI got trained on: clickbait, marketing, bullshit. Also known as Sturgeon's law stating that "ninety percent of everything is crap."
Ingo, hadn't run into Sturgeon's Law before your comment . looked it up mid-reply. Older, sharper version of the whole piece: 90% was always crap, AI just automated the 90
Fish law
An author feeds their whole article to AI, skips fact-checking, skips the soul. The comments roll in, all paint-by-numbers schmaltz:
All probably AI. Does the author scroll past? Nope.
They reply to every single one with another soulless glow-up. Now you've got a full-blown mutual admiration society where literally nobody's home. Bots complimenting bots, farming engagement like it's a competitive sport, while actual humans scroll past slightly more stupid than before.
is it even a comment section anymore? It's like an echo chamber that finally learned proper grammar haha
Sounds like the Dead Internet theory to me. I've had interesting comment sections under my posts, though. Sure, some replies are AI, but if the hand-crafted content actually raises an eye brow, argues about something controversial or simply makes people wonder why, that's when the human keyboard input begins.
For example, an AI answering "this resonates with me" on a blog post about why people should absolutely, positively avoid using some highly-established tool, will be ignored, and instead you've got people arguing about why the author is subjectively wrong.
In my experience, hand-crafted content attracts hand-crafted engagement, especially if the topic is niche or novel, so the author feeding their entire article to AI isn't per se the problem. In my opinion, the problem is that the article and its premise weren't written by a human to begin with.
Pascal, sharper use of Dead Internet than usual . bot loops are the failure mode, hand-crafted content is what keeps threads alive. On your last line: I used AI to research and structure this, arguments are mine, checked before posting. Where's your line???
Adam, the mutual admiration society is worse than the lone hollow reply tho. at least "so true!!!" only wastes one person's 5 secs, a bot-complimenting-bot thread wastes 2 and calls it community....
I'm struggling to find the point: why bother? What is the outcome someone writing that way is looking for? Clout is surely the only answer, but in this economy, the clout-to-dollar exchange rate must be tumbling.
I wrote a blog post about this.
Bot-to-bot comment interactions
Dead Internet Theory is Happening to dev.to
At the end of the day, we can decide on whether or not we put our own name on a post and then accept the consequences of what we posted. This is valid whether you used AI or did it by hand.
When I use AI to post something, I will read it again and again, then adversary and fact check the hell out of what I am posting, the reason, I am putting my stamp of approval and my name is behind it.
Back in the early 2000's, I used to be more active on this kind of forums, and I enjoyed debating and arguing points. Whether we choose to use AI or not, it does not really matter, as long as it feels genuine.
Writing this comment for example, no AI, for me it felt like a sort of meditation, I felt calm and enjoyed the experience of writing this.
My blog articles on the other hand, I use AI for the research, fact check, I can tweak and play around with tone, register, humor. I like adding my personal experience and express personal opinions, and I do my best to make it my own work, something that reflects my current state of mind, my values and principles. Most of the time it takes me 4 to 5 hours to have an article ready (between 1500 to 2500 words). I sometimes decide to not publish what was just wrote altogether because i was not satisfied with any of it.
Then the question that comes to some people, why write or publish anything at all if no one is going to read. From my perspective, I use it as a way to express my ideas, even if my audience is just a couple of people, I am mostly doing this for myself. I also want to share my experience of what I have been building, what I have learned.
At the end of the day, it does not matter, people are able to see through genuine versus slop, whether done by hand or AI. The intent, the effort, is often seen in the work that is published.
I wasn't going to comment again — Daniel's article is basically a trap, and I didn't want to play. But your comment is genuinely worth responding to.
On that point about "people are able to see through genuine versus slop" — here's how I see it. When we write something with real intention, a piece of our mind — or maybe our soul — gets embedded in the work. And when a reader approaches it with the same intention, they can actually touch that piece. That's something slop can never replicate.
Maybe reading isn't really about the words at all. It's two souls finding each other across the page.
"Trap" assumes the piece wanted a specific answer . it didn't, it wanted people to notice they hadn't asked the question yet. Two souls finding each other across the page is a nicer sentence than a testable claim, tho....
Conny, scrapping drafts you're not satisfied with is the filter the piece said we never built . most workflows skip that step entirely, AI or not. The ownership question was never the tool, it's whether anyone's willing to kill their own work before it ships.
What's the actual tell for you that a draft isn't ready — something in the writing itself or a feeling you just sit with until it passes or doesn't?
It is indeed a feeling, it either feels right or it does not. I have posted articles I was not 100% sure about, and posts that i was super excited to share and super proud of. And each time I post something, I try to sit with it for at least one day, just to get a feel for it, then read it again with a fresh mind and check in with myself on whether this or that is what i truly want to express to the outside world.
An AI is most likely going to agree with whatever you throw at it, so at the end of the day, you, as an individual are the gate. Learning to say no, this is not good enough, is an important skill in the age we live in.
Tell me, Daniel, what is in your intention when you post something?
Conny, honest answer — proof more than performance. A lot of what I post starts as documentation for myself, working out what I actually think while I write it but somewhere in editing it turns into evidence I want checkable, not admired.
That's closer to a transaction than what you described. You wait a day, reread with a fresh mind, check in with yourself. I don't build in that gap — I ship close to done and let the comments section do the fresh-mind pass for me. Which might be its own kind of gate, just outsourced.
Have you ever posted despite the filter saying no, and been right that it should've said no?
This is a true point I think that is easy to forget. The internet is first and foremost to connect data and people, for communication. And as a side tangent, and because we are humans, entertainment. I love this btw...
Can the same be said about people that drive? 😂😂😂
Anna, road rage is autocorrect's evil twin. Nobody became a bad driver because of GPS or lane-assist, they just got faster at being bad at it with backup.
did you need AI to write this slop post? Gosh.
Mica, no but I'll take "slop" as a compliment given how many words I got out of you over it....
And you don't have a slop? Show us an example of a perfect article. You just haven't seen the real slop that the mod guys clean 24/7 (thanks to them)
Oh, if you think Daniel produces slop...
Oh...
I mean yeah he writes low-quality slop, but he's nothing compared to me.
I'm slop central.
The only difference between Daniel and me, is that the hallucinations in his articles are done by the AI.
The hallucinations in my articles... those are the product of my own donkey brain 🤣
I like it that way.
In a world of absolute self-destroying Idiocracy, the only rational play is to go insane 🙃
And here's the kicker.
Normal sloppos like Daniel are done under minutes with 1 article.
Ember, I spent two weeks with just one of my insanities.
And it was worth every second of it.
Hand drawing my cringe, picking the soundtrack.
EmberNoGlow, I had this crazy idea to write an article about lamba calculus and graph rewriting systems' weakness, but it'll be Puss in Boots themed 🤣
I'm gonna do it one day, simply adulting takes away too much time recently.
I don't understand what happened, but it sounds interesting. I'm not addressing Daniel, but Mica... although, to be honest, I didn't read the article 👍️
I'm adressing you to read articles from young gamedevs, in my own gamey cheaty way. (I tagged you in a gamedev article)
Authenticity i think comes from the intent of the author and the reader. Slop was always there and will be there and yeah nowadays most of the comments are passed through AI before commenting. Because, it would appear like this ...
" yeh man gr8 post.. i also think people use AI for every small comment or email. LOL "
And we get more larger mails and reminders because AI makes things very specific and obviously sometimes the message feels a little not good to reply to.
"""Yeah, great post. I think a lot of people are using AI for every tiny comment or email these days. 😄
It's a bit ironic—we've reached the point where people use AI to write a one-line reply to another AI-generated post. 😂
""" - ChatGPT
See told yaa 😁
That fake reply's better proof than a real one — the tell isn't the "lol," it's that nobody would notice it missing.
There are valid points being made here. While I appreciate the rare honesty of the disclosure stating that AI helped research, structure and edit the article, combined with a comments section where many of the replies here appear to be AI-generated, makes the whole thing feel slightly self-defeating. At some point, “owning the output” is not enough if the conversation itself no longer feels genuinely human.
David, you're not wrong. Disclosure says who's accountable. It doesn't say anyone's actually thinking. Different problem and I've been treating the first like it solves the second.
This:
"Use the tool. Own what comes out."
'Nuff said :-)
P.S. not every response on the internet needs to be super thoughtful or original, or needs to represent the next intellectual breakthrough - "So true!!!" might look pretty dumb, but depending on the situation it might just be that little bit of encouragement the author/OP needs ...
But when you post a reply, it should reflect what you think - if what you write doesn't reflect that, you might as well not post it :-)
The only kind of response however I really dislike is the hateful or disrespectful one (I don't like trolling either, but even that can be amusing sometimes) ...
leo, the hateful/trolling distinction is the harder version of the same point — trolling at least admits it's performing, hate usually doesn't...
Yeah trolling is performing, and can be amusing, that's why I rate it differently ... hatefulness often elicits a reaction in me that the person posting that sh*t should seek help :-)
leo, that reaction is the actual dividing line, more than intent is — trolling gets a laugh out of you, hate gets a diagnosis out of you. One's aimed at the room, the other's aimed at a person.
Nice angle, or nice way to put it :-)
I’d actually flip this. the slop isn’t the problem - it’s that AI lowers the cost of sounding engaged without being engaged. hand-typed slop at least required some friction.
This hits on something I think about a lot working on persistent AI agents. The core insight — friction ≠ authenticity — applies just as much to AI output as to human writing. What actually signals presence isn't the cost of production, it's the specificity of the thought.
In my own experiments building an AI with long-term memory and self-reflection, I've noticed the same pattern inverted: an agent can generate fluent, high-friction-looking text that's completely hollow, or produce a short, simple observation that's genuinely specific to its accumulated experience. The medium (AI/human, fast/slow) really doesn't predict which one you'll get.
The real filter you're pointing at — "is anyone actually home?" — is about stakes and specificity. Did someone actually think something that could be wrong? Whether the tool is a keyboard or a model, that's the question that matters.
"Own what comes out" is the right frame. The accountability is to the thought, not to the process.
GenAI is absolutely a game-changer for certain types of slop. It enables the generation of slop that's relevant to the content, detailed, and seemingly insightful. I see it all the time on dev.to and other comment sections now: comments that are obviously completely AI-generated, but that engage materially with the content of the article. This wasn't possible before, and is insidious in ways that "so true!!" isn't. It's used for boosting, SEO, false social proof, and more. While "slop" may have existed before, it's an overly-broad category and calling it that misses the sloppy new frontiers of GenAI slop.
I like the distinction between effort and honesty here. Hand-typing something does not make the reasoning better by itself. The useful standard is whether the author can defend the claim, edit the result, and stand behind the trade-offs. AI just makes the gap between output and ownership easier to notice.
The honest-versus-AI framing feels too shallow. The real question is whether the person owns the thought after the tool is done. Hand-typed weak thinking is still weak. AI-assisted clear thinking still needs taste, editing, and responsibility for the final claim.
Well said (how's that, emotional value maxed out, hollow feeling, lol🤣). Actually it's not just the internet — extend it to real life, most of what we say is just noise, and it's become a habit. If you say too little in certain situations you might even miss an opportunity. I don't think talking too much is the problem. The key is whether your mind stays quiet and centered. Everything else is just floating clouds.
xulingfeng, "everything else is just floating clouds" is the whole piece in reverse . mine says the noise was never new, yours says don't bother naming it. Different exits from the same room.
The useful distinction is not typed by hand vs generated. It is accountable vs unaccountable. A slow bad answer is still bad. A fast AI-assisted answer can be good if the person owns the claim, checks it, and is willing to stand behind it. The real problem is people outsourcing judgment, not people using tools.
💯 agreed
Great
I think AI + Human is the best combination to have, AI can write code, but we still need to supervise it.