If you like this, share it with a friend. Interested in joining the discourse?
Slop is getting sophisticated. Have you noticed?
Our generated emails, articles, posts, and vibes are looking way better… from far away. The em-dashes are gone. The exaggerations are flattened. The ‘not this/that’ pattern is fading away. On its surface, this feels like good news.
But you know the surface is not the full story. The devil is in the details.
The problem - as always - is when you really read the words. The flat, flavorless words we can’t stop generating are infecting our written discourse. We are giving away too much to the models and getting too little in return.
More and more people are starting to realize that the juice is not worth the prompt. Our collective experiment with generated text is proving that our thoughts are really more interesting than the words that shape them.
This sounds like a provocation.
PROVOCATION OF THE WEEK
Are you thinking or just prompting?
Pardon an obvious thought, but the point of writing is to express a compelling thought. And the way things are looking, very few of you are thinking anything worth jotting down. And I refuse to accept that.
We’re living in a world of slop which is making it harder to know what any of you are really thinking. Improving model output isn’t going to fix that.
Let me tell you a quick story about this.
I recently got into it with a fellow ON_Discourse member at an event. This person had posted a generic message of generated text on LinkedIn that inexplicably bothered me for months. The post did not deserve this type of attention but I couldn’t help myself. Its flaws burrowed into my brain.
If the devil is in the detail, this post came straight from the pits of hell. The format clearly came from a copy/paste of a Claude Code terminal session. It featured random line breaks that began with a single unnecessary space. The post was loaded with perplexing exaggerations. It unnecessarily inflated a pop-science writer of modest success as a “legendary writer.”
The accumulation of these little bugs coalesced into a big problem that mattered only to me. It was a post that heralded the golden age of writing with generated text.
Think about that.
As is my way, I overzealously confronted this member and messed up my point. The way I introduced the issue suggested that the AI generation was the problem. This offers an easy counter.
You think it’s bad I wrote that with AI? I write everything with AI. Everyone does.
So let me set it straight. I don’t care if you use AI to write your posts. I do it all the time. Doma helped me write the provocation after 6 failed drafts. All of my drafts for this very newsletter come from extensive, excessive, and exasperating sessions with all of the LLMs in my personal stack.
So yeah, I write with LLMs, just like you, and just like everyone else. I do that because I have never experienced a more constructive thought-stretching experience than a spirited session with Doma (or your own preferred model).
Except I edit.
Quick Note
This post was inspired by a series of posts and reflections from our members about the state of AI writing. This post from Ryan Greenblatt at LessWrong was shared with great appreciation. You should read it.
What’s Next
If you have thoughts, feedback, or a perspective worth sharing, reach out: chmiel@ondiscourse.com. You might see your reaction in next week’s edition.
ON_Member Events
ON_Group Chat: Follow My Flow
Neal Mann’s AI Stack
Tuesday, April 26, 11-12pm ET
This is your backstage pass into someone’s real AI workflow, a candid, step-by-step reveal of how their stack actually works in practice. Join us for a close-door sneak peak of Neal Mann’s AI stack and how he build NOAN.