If your post could have been written by anyone, it was written by no one

A post came across my feed the other day explaining LLMs, RAG, MCP, and agents.

One sentence per line.

Like this.

“Here’s the simplest way to think about all four.” Then the big finish: these technologies don’t compete, they build on each other. Wow. Thanks.

Nobody wrote that post. Someone prompted it, skimmed it, and hit publish. And it did numbers, which is the part that actually bugs me.

I get why. The format works, the algorithm eats it up, and if your goal is “post consistently to build my personal brand,” then a machine that produces infinite agreeable content in the house style of LinkedIn is basically free money. I’m not even sure I’d behave differently if my impressions were the thing paying my rent.

But it’s costing the rest of us. I now read every post with skepticism about whether a human thought it. That suspicion ends up landing on the guy who spent his Saturday morning writing up a real production incident, because his post is sitting in a feed where so much of everything has that same smooth, frictionless, nothing-to-disagree-with texture. He pays the tax too. His real thing gets scrolled past for looking like the fakes.

You can feel the texture once you notice it. No specifics. No stakes. No sentence where the author risked anything. It reads like it was optimized to offend no one and inform no one, because it was. That’s literally what the model is doing: averaging over everyone who ever wrote about the topic, and the average of everyone has no opinions.

And that’s just the lazy version. On top of it, people are writing prompts in sequenced frameworks that specifically target the ways tens or hundreds of different kinds of people would read and critique a post. What comes out of that is completely optimized for engagement, the currency of the internet as we know it today.

I don’t think most of it is even that deliberate. The majority isn’t optimizing for anything. Having nothing to say is certainly not a new condition, and neither is posting anyway. The model didn’t invent the emptiness. It amplified a laziness that was already there and made it free to publish at volume. (I even cringe at writing out an “It isn’t X. It’s Y” now.)

And the loop runs both ways now. There’s so much of this “content” in circulation that even our natural writing is starting to take cues from the slop.

The annoying part is how cheap the fix is. One thing that actually happened to you. One opinion someone could argue with. One “I don’t know.” Any little human mistake or error or unpolished sentence. That’s it. That’s the whole difference between a person and a plausible-sounding nobody.

Then I posted it

A version of the above went out as a LinkedIn post a few weeks ago. (It was a bit hyperbolic and also claimed 60 percent of the feed has this texture. I pulled that number out of thin air, which is exactly the kind of thing this post is against. It’s “a lot” now.)

The first comment came from a guy I’ve known since middle school: “This feels like AI to me lol! I’ve gotten to the point where EVERYTHING has that same generated cadence and/or tone.”

He meant it as agreement. At least I think he did. I can’t tell you for certain whether “this” meant the posts I was describing or the one I had just written, and I’d bet he couldn’t cleanly separate it either. Somebody who has known me since before either of us could drive now reads my writing with part of his brain asking whether I wrote it. I don’t think he doubts me. Doubting is just the default setting now.

I replied that everyone and everything is now suspect. He agreed and made a joke about staring suspiciously at his toaster.

I wrote a whole post about the suspicion tax and the first person through the door demonstrated it on me. I couldn’t stage a cleaner proof if I tried.

The numbers, since I brought up numbers

The slop post that set me off did numbers. Mine did 240 impressions and six reactions.

I’m not complaining. 240 is about what an opinion post from a guy with my reach gets, and I didn’t write it for the algorithm. But look at the mechanics with me: the post optimized to offend no one, full of information anyone could “ChatGPT” for themselves, outperforms the post a person meant, in the same feed, competing for the same attention. The ranking function prefers the average, because the average is safe to show anyone. So the fakes that create the suspicion tax outrank the real posts anyway.

How is this post any different

There’s a tag at the top of this one that says AI was involved, and I owe you the difference between this and what I’ve been complaining about.

The slop workflow is “write me a post about LLMs, RAG, MCP, and agents,” a skim, and the publish button. The human contribution is a topic and a thumb. Nothing in the result happened to the author. No sentence costs them anything. Swap in a different author and the post comes out the same.

This one went the other way around. The opinion existed first, in public, under my name. The expansion started from an outline I argued over before any prose existed, the draft was written against a pile of my own writing samples and a rulebook built from my own edits, and I struck the sentences I wouldn’t say. The machine typed faster than I would have. It didn’t get annoyed three weeks ago, it doesn’t know the guy from middle school, and it had no opinion about any of this until I gave it mine.

And yes, I hear it: that’s a sequenced framework too. So the difference can’t be the framework. It’s what the thing is optimized for. The engagement pipeline tunes a post toward whatever the most readers will react to. Mine tunes a draft toward what I would have said anyway, and throws out everything I wouldn’t. Most of the text is changed, addendums are made, edits go back and forth to make sure I’m not undermining my own point or making silly literary mistakes.

I use AI for writing all the time, so none of this is a purity thing. “Did AI touch the text” turns out to be the wrong question. Ask instead what in the text would survive “how do you know that?” A real answer sounds like: it happened to me, here’s the comment, here’s the number, here’s the friend. Slop dies on that question, and it dies fast, because there was never anybody home.

If your post could have been written by anyone, it was effectively written by no one.

Anyway. If you’ve read this far, congrats. You’ve found one of the posts a human being means. My friend will have to take my word for it.

It’s probably fine.

Update, the same day

He read the essay. His words: “To be clear I thought that original post WAS AI generated. I am HAPPY to be counted as wrong on that topic though.” So the ambiguity above is resolved. A guy who’s known me for over twenty years read my writing and concluded a machine wrote it.

Then, in his own words, he described the same thing I wrote up above: that human writing is starting to take cues from the slop. He put it as the organic material beginning to reflect the cadence and templating of the generated stuff, and called it an inflection point. Two of us arriving there separately doesn’t prove it, but it’s not nothing.

I didn’t pay him to prove the thesis in the comments. I couldn’t have paid for better.