A post on X last week:

"The fact that people can't even reply to posts without AI anymore says a lot more about them than they think."

The replies agreed. Nobody asked the obvious question: says what, exactly?

Here's what I think it says: nothing new.

People have been hollow online since the forum era. "So true!!!" didn't become meaningless when GPT launched. It was meaningless in 2009. Copy-pasted under blog posts. Typed by hand. Fully human. Fully empty.

What changed isn't the emptiness. What changed is the cost to produce it.

That's a different thing.

AI is reportedly writing and reading more of the internet than people are at this point. I haven't chased down the study, but it tracks with what everyone's noticing.

But the conclusion most people draw from that — that authenticity is dying — assumes there was a lot of it before. There wasn't. There was friction. Friction isn't the same as authenticity. It just made the emptiness more expensive to ship.

Autocorrect didn't make people bad texters. It made bad texters faster. The badness was already there.

The tell is whether anyone is actually home behind it. The tool doesn't matter.

Someone typing "this resonated with me 🙏" isn't more present than someone who generated it. They just did more work to say nothing. That's not a virtue.

It does cost something. It costs attention. It costs the willingness to say something specific, something you actually think, something that could be wrong. Most people online weren't paying that cost before AI, and they're not paying it now.

The gap was never intelligence. It was performance. It just wasn't visible when the performance required typing.

Generation got cheap. We never built good filters for the expensive stuff — actual thinking, actual specificity, actual stakes.

Slop has always existed. In dev work, in blog posts, in comments. The skill was always knowing where slop belongs and when to clean it up. Fast parallel experimentation? Slop is fine. Shipping to production without understanding it? That's the problem.

Same principle applies to replies.

Use the tool. Own what comes out.

The people performing outrage about AI replies are doing the same thing. Pattern-matching to a moment. Not thinking it through.

AI helped me research, structure, and edit this piece. The arguments, the examples, and the opinions are mine. So is whatever's wrong with them.