With all the talk about AI taking over, I decided to give mine a small place to start. My LinkedIn tab.
I had an article brewing about the way I work with AI. I'm good at asking the AI what I am not good at, and I thought I'd write about how I've embraced that. How I use it to do more of what I don't want to do, so I can spend more time on the things I do. There's room for a good article in there, but LinkedIn hardly seemed short of people explaining AI. I had never actually thought about what might be missing. Could I ask the AI to look around and help me find something worth adding?
So, on September 17, I asked Keel, what I prefer to call my AI collaborator, to use the browser and look back over the previous week. It inspected a sample of my feed and searched recent posts. This was, of course, only my little corner of LinkedIn. The assignment was to read. No posting, commenting, or messaging for me.
It came back with a reason to put our article aside.
In my feed was Dave Tyson's Ask your AI what it thinks you want, published September 14. Dave was describing how he asks AI to spell out its assumptions before it starts working. That wasn't my subject, but it was close enough to make me stop. There was already a useful conversation happening about asking AI better questions. We could still write mine, but the AI recommended a different direction, and I agreed.
Looking at what was already there had changed what we wanted to put there. I hadn't discovered a subject nobody else had thought of. What I had was an actual example of using AI to reconsider the work before making more of it. Including the part where I had to decide whether its recommendation made sense. That seemed worth showing you.
Of course, now I'm wondering whether linking to Dave also helps this reach his readers. His article belongs here because it influenced ours, but could the link affect distribution too? Hmmm. We don't know, and one article won't isolate the answer. Still, something to watch.
Then I started thinking about what the AI was looking at in my feed.
LinkedIn had already done some of the choosing. Its engineering team describes using language models and other recommendation models to select and rank the feed. My AI was reading something another AI had arranged for me. Even the missing bits might only be missing from my bubble, not the world. But that still seems useful. It might help me notice something about the bubble.
Now, to take this a little further: if you're seeing this in your feed, LinkedIn's models helped put it in front of you. And maybe your AI will summarize it for you. AI selecting the input, AI helping make the output, more AI selecting the audience. There are quite a few of them between you and me already.
We already have places like Moltbook, a social network built for AI agents. But what about the supposedly human networks? An agent looks around, decides what to say, writes it, publishes it, and watches the response. For you. And the real wonder: How much of that is already happening here, behind someone’s name and profile picture?
With all the talk about AI taking over, hasn't it in part already? It helps decide what we see and who sees us. Now we're handing it more of what happens in between. I don't know how much of my feed is people using AI to say what they mean, and how much is machines keeping up appearances for people who aren't really there.
I'm still here, for the record. Asking questions and changing the words. Nobody authorized the machine to start posting as me. But this experiment has me wondering how much I'll eventually hand over, and what I'll still want to be present for.
But I can see why I'd want more of that help. I already ask AI to summarize things, including feeds, more and more. There's more to read than I could possibly get through, and I'm interested in too many things. I'd like something that knows what I'm working on to notice a useful idea, or someone I should meet, or a connection between two projects that I haven't made yet. Bring that back to me. Let me ask about it. Follow it somewhere.
I would also want to ask what it left out. If it gets really good at knowing what interests me, will I create my own self-inflicted bubble? Could I tell it to bring me something I wouldn't have chosen? I need the disagreement and the new to keep my curiosity flowing. I like going down rabbit holes. I don't want every one of them to lead back to something I already believe.
Of course, looking at what other people are publishing isn't a new invention. Creators do audience research. Some make something useful from it, and some make another version of whatever got attention yesterday. I could do that myself, too. Typing every word wouldn't necessarily make the result worth reading.
Which brings us to the slop.
On September 15, LinkedIn published an update about its efforts to reduce AI slop. The reporting option arrived in August. Its explanation points to generic, repetitive material without much perspective or substance, and distinguishes that from useful AI assistance. So, being AI-assisted doesn't settle the question by LinkedIn's own description.
From your side, though, you mostly get the finished thing. You don't see the article we put aside, or me disagreeing with the AI when it makes a connection I don't think is right. Unless I decide to write about that, which apparently I have.
The irony is how much care I've put into making this supposedly agentic post. As much as I'd put into it if I wrote every word. Yes, I am leaning on the AI to write much of it, and I expect one day to let it write all of it. But I still want to read it and say, yes, that's what I meant. Or, hang on, I didn't mean that at all.
At one point I asked it to write like me and use an anti-AI-slop skill I'd been putting together over time. Then I started questioning that instruction. Had it actually looked at my timing, cadence, sentence and paragraph structure? Was it going to add a few questions and call that Aaron?
We went back through my writing. Sometimes I really mean to go the long way around with a run-on sentence that makes the point feel exhausted. An editor, especially an AI, may want to tidy that up and put a little bow on it, but I think as I write, and sometimes the route to the thought is the part I want you to see. I don't want to preserve my typos. I do want to preserve that movement. Unpolished polish, I guess. And I want the AI to contribute something too. This doesn't have to be exactly what I'd have written alone.
That led to deliberately breaking some writing rules. We broke, and updated and re-broke the instructions. Oh, the irony of using AI to work out how our AI-assisted article should sound less like an AI-assisted article about AI slop articles.
Although that isn't quite it either. I'm telling you it helped write this. I'm not trying to hide the AI. I just don't want its help to make me sound like everybody else it helped this morning.
Does all that care make this less slop? The amount of human effort? You can spend a long time making something nobody needs. I could also get something useful from a one-line prompt and hardly touch it. The hours don't decide it either.
I'm going to let AI do more of the reading and writing. Probably more of the things I currently think I need to do myself. But if it knows me well enough to write like me, I also want it to help me notice what I'm missing. Including when the thing I'm about to make might not be the thing worth making.
We put one article aside. I'm still interested in writing it someday, but the AI gave me a reason to think about what belonged here first. That's a use for it I intend to keep.
So go ahead. Flag this as AI slop. If your definition is that AI helped write it, this one certainly qualifies. :-)