I'm more bored of the AI slop debate than AI slop

AI slop is real. There’s plenty of bad AI-generated writing, and I’m not defending any of it.

What I’m bored of is where the conversation about it stops.

My feed is full of arguments about whether a sentence sounds like ChatGPT. Whether somebody used an em dash. Whether a particular post is “authentic”. Whether the machine has enough taste, soul or humanity to earn a paragraph. Some of that matters a little. Most of it is forensic analysis of the least interesting layer of a much bigger change.

I’ve made the substantive case elsewhere and won’t rebuild it here: slop didn’t arrive with ChatGPT, and the opposite of it was never human authorship; the value in writing was always the judgement it was hiding, not the sentences; that value is moving upstream, into evidence, argument and editorial judgement; and what compounds is the knowledge behind the work, not the individual asset.

What I haven’t said plainly is this: writers who spend their attention litigating the paragraph may be defending the wrong territory. The visible layer — does this sentence sound synthetic — is exactly the part that’s easiest to argue about and least connected to where the value actually sits now: which evidence a model should see, which sources deserve trust, when a second model should be asked to attack the first, what knowledge is worth keeping, where a human has to intervene. Those are editorial questions. They’re also considerably more interesting than another thread about em dashes.

Bad writing existed before ChatGPT, and it’ll exist after this debate exhausts itself too. Fine — criticise it when it matters. But I’d rather read someone testing what a second model catches that the first one missed, or figuring out how knowledge should move through an organisation, than one more post arguing about whether a paragraph sounds human.

The question I actually want to see argued is bigger than authenticity: what can we build now that we couldn’t realistically build before?