This week, the internet is once again trying to figure out what counts as "human."
Let's get into it.
LinkedIn vs. the bots
LinkedIn has a new way to flag “AI slop,” which feels both inevitable and a little funny, given that LinkedIn is also one of the main places where AI slop goes to breed.
The product manager who came up with this feature most definitely deserves a raise.
Are the new models making things worse?
Opus 5, Claude’s newest model, launched this week and people are complaining that it feels worse than previous versions. Here's what's likely happening: when you use a model long enough, you unconsciously learn how it responds (what phrasing gets the best results, how much context to give it, when to push back). You build invisible muscle memory around it. When a new model drops, all of that calibration breaks. The model isn't necessarily worse. Your setup just doesn't fit it yet. It's like switching from an iPhone to an Android – the phone works fine, you just keep reaching for buttons that aren't where they used to be.
The internet had a lot to say about the Opus 5 release.
This is actually related to what we wrote a few weeks back: setup is everything. The people who feel like a new model is broken are often the ones who had the best setup on the old one. And the fix may be to recalibrate your prompts, your templates, and your workflows for how the new model thinks.
The strategic typo
But there's a broader thing happening, too. AI is moving fast, and our trust signals are moving even faster.
We saw a post recently about people deliberately putting typos into emails to signal that they weren't written by AI. Which is wild, because until about five minutes ago, typos were a sign of laziness or sloppiness. Now they're becoming a secret handshake: "Don't worry. I'm not a bot."
This came up in a conversation with a prospective client who was worried about sounding like everyone else – too polished, too generic, and over-optimized. In other words: like AI slop. It’s a completely reasonable fear, because nobody wants a vanilla brand voice.
Instead of cosplaying as a human by making typos and being messy, maybe the answer is to get sharper about what actually makes something feel human: specificity, taste, point of view, context, restraint, and the occasional sentence that sounds like a real person had to live through something before writing it.
The common thread (stay with us…)
The thread connecting all of these rambling thoughts… The LinkedIn flags, the model complaints, and the strategic typos, is that people are still figuring out how to use AI without losing what makes their work theirs. Whether it's a new model that needs recalibrating or a brand voice that's drifting toward generic, the answer is the same: don't use the default answer from your AI. Be deliberate about your setup, your voice, and what you actually want it to do.
And if you're not sure where to start, send us your burning questions. We'll answer them.
Stay curiuos (see what we did there!?),
Julia & Russell
