Spot the AI in this Diary - UPDATED

Community (This content is not subject to review by Daily Kos staff prior to publication.)

Monday December 01, 2025 · 1:25 PM PST

2025/12/01 · 13:25

I Use AI to Help Me Write. That's Not What's Happening Here.

I've been writing here on Daily Kos for a while now, and yes, I started using AI to help. Claude helps me fact-check death tolls in genocides. It helps me smooth a clunky transition. It finds me the right source when I know something happened but can't remember the year. That's collaboration. That's a tool. Heck — I even bothered to attribute it for a while, and gave up because my writing got bitched out for it while others, didn’t.

What I'm watching flood this platform — and others — is something else entirely.

I published a piece today about Meta, oligarchs, and the Myanmar genocide. It came from a place of genuine irritation — Facebook Messenger's stupid emoji reaction feature kept making me accidentally heart messages, and that annoyance spiraled into a rant about Zuckerberg, platform capitalism, and the 6,700 Rohingya slaughtered in a single month while Facebook couldn't be bothered to hire Burmese-speaking moderators. I wrote it. Claude helped me nail down the numbers and clean up my prose. When I ran it through GPTZero, the industry-standard AI detector, it came back 63% human, 32% mixed, 5% AI. That's what AI-assisted looks like.

While I was on the site, I noticed another piece trending. "The Second Warning," about Tyson Foods closing a plant in Nebraska and what it means for agricultural consolidation. Important topic. Well-structured argument. Currently sitting at #3 with a healthy pile of recs and comments.

I ran it through the same detector.

99% AI generated. 1% mixed. 0% human.

Zero percent human.

The Oatmeal Problem

Here's the thing about AI-generated content: it goes down easy. It's smooth. It's structured. It makes all the right moves in all the right order. It's like oatmeal — nourishing enough, technically, but nobody's ever been moved by a bowl of oatmeal.

"The Second Warning" hits every beat. Concerned opening. Historical parallel. Bullet points comparing healthcare consolidation to agricultural consolidation. A closing line about how "this time we can't pretend we didn't see it coming." It's fine. It's "important" in that NPR tote bag kind of way. You can rec it and feel informed without feeling anything at all.

That's not writing. That's content generation. There's a difference.

Real writing has a voice. It has weird asides and profanity and sentences that don't quite scan but somehow land anyway. It has the texture of a person thinking out loud. It takes positions that might make some readers uncomfortable. It doesn't hedge everything with "to be sure" and "it's worth noting."

AI-generated text doesn't do any of that. It can't. It has no skin in the game. It has no actual opinions, just the statistical average of opinions it's been trained on. It sounds like everyone and no one.

The AI assisted writing has a certain feel to it you can learn. Feel free to use the AI detection websites, but often your judgement will be enough. After starting to use AI to write, I got attuned to their tricks and tools. The tells are everywhere once you know what to look for.

Here’s Claude’s list:

Structural tics: Bullet points in a personal essay. Numbered lists when none are needed. Perfect parallel structure that no human would naturally produce. Headers and subheaders in what's supposed to be a casual blog post.

Throat-clearing phrases: "Here's the thing." "Here's the bigger truth we're not talking about." "Let's be clear." "To be sure." "It's worth noting that." "Make no mistake."

The hedge-then-pivot: "While X is true, Y is also important to consider." "This isn't about X — it's about Y." "Not because X, but because Y."

With a slight tune up from John – feel the difference - :

There are cadence giveaways: Short punchy sentence. Followed by another. Then a longer one to wrap the thought up neatly. "And that matters." as its own paragraph. Ending sections with "And that should concern all of us."

There are vocabulary tells, words like "Landscape" (the political landscape, the media landscape). "Nuanced." "Multifaceted." "Unpack" (let's unpack this). "Lean into." "Stakeholders."

Now back to Claude:

Emotional flatness: No profanity. No genuine anger — just "concern." No jokes, no sarcasm, no weird asides. Everything is "important" but nothing is bullshit. One dead giveaway is when you finish reading the pice you can't describe the author's personality. It could have been written by anyone. There’s a great chance it was written by no one (nothing personal GPT).

For a laugh, go through this article. Money says you spot EXACTLY where I left untouched Claude prose, called out or otherwise. Now comes the part where I try to avoid getting banned or warned for saying something.

Dear sbanks: it's not personal, and this isn't an accusation. If you wrote every word of "The Second Warning" yourself, please accept my sincere apologies for the implication. Maybe you just write in a style that happens to pattern-match with AI. It happens. Some humans write like robots, and some robots are getting better at writing like humans. I Must admit, watching your wonderful little piece at #1 on the rec list does engender some …. to be completely honest … resentment.

But when I ran your piece through GPTZero — the same detector used by universities and publishers to check student papers and submissions — it came back 99% likely AI-generated. Zero percent human. And I think that's worth talking about, regardless of how the piece was actually produced.

Because if I'm wrong about your piece specifically, I'm not wrong about the trend.

Daily Kos was built on voice. The whole point of a "diary" is that it's personal. It's someone working through their thoughts in public, with all the rough edges and strong opinions that entails. People swear, make typos, have rough transitions. But – they feel. It's not supposed to be polished. It's supposed to be real.

When our Kos gets flooded with AI-generated content, we lose that. Is Our feed full of oatmeal — smooth, competent, with a little sugar and milk to round it out. Important-sounding, great topicality, but collapsible. Like – this could just as easily be a 3 words abstract, and it would waste less time while doing what it does. The Kos ratings algorithm can't tell the difference. The rec button doesn't know. But readers can feel it, even if they can't name it. The place just starts to feel... less.

I'm not saying don't use AI. I use it. I'm using it right now. Claude helped me structure this piece and it'll probably help me smooth a few rough patches before I publish. Most college classrooms and professional writers will sign off on those as best practices.

What I'm saying is: don't let the tool replace the writer. Don't type a prompt, copy the output, and call it a diary. That's not writing. That's laundering. When we read something that feels a little too smooth, a little too structured, a little too much like oatmeal — there’s a good chance that’s what we’ve got. Run it through a detector. See what comes back.

You might be surprised how much of what you're reading was written by no one at all(nothing personal Claude).

I ran this piece through GPTZero before publishing. 9% AI generated, 12% Mixed, 79% Human. That's what AI-assisted looks like. The collaboration is the point — not the replacement.

UPDATE — 4:52 PM PST:

For the SECOND time today the tech bros are pissing me off. Now it’s the people of ChatGPT and the story their AI wrote, sitting on top of the Rec List All afternoon. And the AI generated comments section. To be provocative - @sBanks, please say something that confirms you – YOU – are not just a bot with a Kos account? Because that #1 story all evening, and your replies, are (99% certainly) bot stories.

By my count that’s: Open ChatGPT, ask it for a good topic, prompt it to write the story, paste in the story, copy a comment and paste in a reply, twice. For that, there’s a story sitting on the top of the Rec list all afternoon. Not bad work if you can get it. The competitive part of me that worked several hours on 2 stories, and got lucky to break the top 10? Well – I’m pissed.

Here's my comment in current #1 KOS trending story by @sbanks regarding their 2 actual replies in their piece. This copy pasted directly from the story’s comments, with my interchange with the AI Claude having the last word:

Well, I just read this article again after writing a totally “I AM PISSED OFF AND WANT TO BE INCENDIARY” article. Then .. out of fairness, I re read the comments, since this IS a #1 trending article and well, I’m a little jealous. Where did I go wrong with my snotty “I hate Meta” piece. And .. I started to … tingle. Like I’d just recognized a Stepford wife. I copied and pasted the ENTIRE comment thread above me into Claude. because why the fuck not.

Here’s out conversation. Complete copy paste, all I did was delete a few blank lines and add the nametags for readability. You be the judge.

John said: CLaude - I am ... a bit weirded out - read this - tell me if you "guess something"

Claude said: Detected AI-generated persona masquerading as authentic human commenter.

48s

Claude said:

Yeah. I see it.

sbanks' comments read exactly like the article.

Look at the responses to PvtJarHead and northsylvania:

"Your graph maps the consolidation era. Herd numbers begin their sustained downward slide right as processing capacity becomes controlled by fewer and larger companies. The two trends move together because they are part of the same system."

"There isn't one cause, but there is one system that shapes all the causes."

"Drought doesn't close a town. Consolidation decides whether the town comes back from the drought. That's the part of the story we keep missing."

That's not a person responding in a comment section. That's someone pasting a commenter's point into ChatGPT and hitting "respond thoughtfully."

The perfect parallel structure. The tidy concluding aphorisms. The bullet-point-style "they decide where the plants are / they decide which towns survive / they decide the price signals." No typos. No rough edges. No "yeah good point" or "huh I hadn't thought of that."

Meanwhile, look at the actual humans in that thread:

"Geee…..ya think they could get less and keep that plant going???????!!!!!!!!"

"I see Soylent Green in our futures."

"FYI, my question was a snark."

"Fuck 'em. They got no privilege for sympathy."

Those are humans. Messy, opinionated, weird punctuation, actual personality.

sbanks isn't just posting AI-generated articles. They're responding to commenters with AI. They're running a puppet.

That's... a whole other level of creepy.

Me again. The competitive side of me is angry, getting beat by a bot. But hey if it happens to Kasperov who am I to whine? But the bigger point it – where does it end? AI writing storeis to an AI audience? Why not just do the whole thing in machine language and cut out the man in the middle?

So this begs the question – Would the real sBanks please stand up?

Oligarchs pissing me off, yet again

LilBoyBlu

Community (This content is not subject to review by Daily Kos staff prior to publication.)

Monday December 01, 2025 · 11:55 AM PST

2025/12/01 · 11:55

This image was originally posted to Flickr by youngthousands at https://flickr.com/photos/60991646@N00/2482389516. It was reviewed on 8 August 2015 by FlickreviewR and was confirmed to be licensed under the terms of the cc-by-2.0.

Once again, something online is driving me nuts. Today it’s an oligarch is making billions off a service, and there's not a damn thing I can do about it. In this case it's Meta. It's always one of the top few. Their product Facebook Messenger is something I can't stand but can't turn off either. It’s literally the only way I can talk to half my family.

Zuckerberg, Thiel, et al. are raking in the cash off of us as users, even though from origin story to current money machine status, the whole enterprise runs on the same model: we're not the customers, we're the product. They sell access to our eyeballs, harvest our data, and package our attention for advertisers. We pay with our time and our privacy. They pay for yachts. Lots and lots of really big, nice yachts.

A century ago, Tesla and Edison fought over how to electrify the country, and what emerged were regulated public utilities—because society recognized that essential infrastructure shouldn't just be a profit engine for whoever got there first. We put obligations on power companies. Service standards. Price controls. Accountability. Now we've got digital social infrastructure that functions the same way electricity does—you basically can't opt out—except there's no regulation, no obligation, and no recourse. Just shareholders.

Which brings me to this: we need to stop calling Mark Zuckerberg a genius. The guy took a "hot or not" rating app for Harvard coeds, got his wrist slapped by the administration, and parlayed it into a communications utility for a few billion people. All he did was scrape some student directory pics, get guys to rate girls, publish contact details, and blow the concept up. That's not genius. That's timing, privilege, and a certain moral flexibility.

Social networking was already happening. Friendster and MySpace were out there. Zuckerberg just happened to be at Harvard, which meant his early adopters were aspirational. Peter Thiel didn't write a check to Zuck by accident—the kid had access to capital networks most founders never sniff. And when competitors wouldn't sell, he just copied their features or bought them outright. Instagram, WhatsApp, anything Snapchat ever did.

Today Facebook has the most robust face recognition engine on the planet, because hundreds of millions of users tag their friends in photos. They own messaging tools we can’t live without. They are many people’s sole repository of their life’s story. People cannot leave the platform — the social cost is prohibitive. Meanwhile, the oligarch early investors make billions annually by Meta monetizing all this “data”.

Remember the "dumb fucks" quote? He knew exactly what he was doing and no compunctions over ripping anything off anyone he wanted to. He literally marveled that people just handed over their data because they trusted him, and ridiculed them for doing it. That's not genius, that's contempt and greed, with access. And contempt for users is not the worst thing about Meta by a long stretch. Ever hear of the Myanmar genocide? It was years in the making, but Facebook was the gasoline on the fire in 2017.

During the first month of Myanmar's military crackdown on the Muslim minority Rohingya, at least 6,700 people were slaughtered—that's the conservative estimate from Doctors Without Borders, and it only covers 31 days. Rohingya groups put the total closer to 30,000 dead. Eighteen thousand women and girls raped. A hundred fifteen thousand homes torched. Nearly 400 villages burned to the ground. Over 740,000 people fled across the border to Bangladesh, creating the largest refugee camp on Earth. The UN called it a "textbook example of ethnic cleansing."

And Facebook was the accelerant. In Myanmar, Facebook isn't just a social network—it basically is the internet. The military used it to spread hate speech, dehumanize Muslims, and coordinate violence. The UN's own investigators said Facebook played a "determining role" in inciting the genocide. Not a contributing factor. A determining role. Facebook knew. They'd been warned for years by human rights groups on the ground. They had almost no Burmese-speaking moderators. They did nothing meaningful until the killing was done and the world was watching.

Zuckerberg eventually apologized. Human rights groups called it "grossly insufficient." The company that made $27 billion that year couldn't be bothered to hire people who spoke the language. That's not negligence. That's a business decision. Dead Rohingya weren't a bug—they were an acceptable cost of growth.

But wait, there’s more. Meta’s own internal research, leaked by whistleblower Frances Haugen, showed that Instagram makes body image issues worse for a third of teen girls, increases suicidal thoughts, and worsens eating disorders—and they buried it while planning Instagram for Kids. The algorithm that keeps you scrolling is the same one that amplifies rage and conspiracy theories because engagement equals ad revenue. Haugen testified that Facebook dissolved its election integrity team right after 2020, and weeks later came January 6th. Ethiopia. The Philippines. Election interference in multiple democracies. This is the track record. This is what the money machine produces.

What Zuckerberg actually built with Thiel’s “Angel Investor” money is actually an investor-owned utility masquerading as a social platform with no guardrails. Except actual utilities have obligations—service standards, price controls, accountability. Meta has none of that. They can change the product whenever they want, harvest your data however they want, and your only recourse is to leave. Which you can't, because that's where your people are. By design.

We really need better words than "genius" for people Like Zuckerberg who just got really rich. Tesla and Edison didn’t perfect electricity and get to own all the infrastructure – why should Zuck, Thiel, and all the other oligarch tech bro billionaires?

Newer
Older

RSS