You cannot prevent AI-generated content from being created. Of course not. It is cheap, instant, and people are lazy. Some people will produce careful work with it. I hope I am one of them. Some will produce absolute slop they did not even bother to read.

Someone got AI to explain database columns, pasted the result into teaching material, and shipped it. No one stopped. No one checked. No one enforced a boundary between "the model said some words" and "this is now fit for publication." Author, editor, publisher...

I started the draft of this post by asking ChatGPT what Pim would say about the original tweet. It said this:

Funny, but this is not an AI problem. It is a process problem.

Someone generated text, pasted it into teaching material, and nobody did a real editorial pass before it shipped.

The model did what models do. The failure is that no one owned the last mile.

Then, when I told it that is most definitely not what I would say, it said that I would say:

Not bad. A decent impersonation. But my actual take is simpler: people do what they do, and the systems people create are what they are. If nobody enforces boundaries, this is what you get. AI did not break a good process here. It flowed straight through a bad one.

Even when I try to explain it, codex will have trouble replicating me. This whole post is drafted by AI, but I am the author, because I decide to publish it. Your boredom is my fault. The take is not that slop content funny, or slop content should not exist.

Tools and automation produce slop. That's just what they do. AI is not good enough yet to not produce slop.

And people will let it, because not producing slop is actually quite hard. It feels like determinism. They will produce and you will complain and nothing changes.

So my take is: protect your boundaries. Use AI for it. Quickly leave if it's slop and you don't feel like eating. But, let it go. Do not get hung up on other people's slop.

Original tweet

My quote tweet

ps.

Codex' final review:

  • Medium: src/blog/2026/2026-05-11-the-boundaries-of-ai.md:25 is the weakest paragraph. “Even when I explain it, codex will have trouble replicating me” shifts from the post’s point into tool-name commentary, and “Your boredom is my fault” is strong but a bit disconnected from the preceding setup. The argument gets blurrier right where it should tighten.