Anthropic has let me down

Anthropic has let me down.

I used Claude mainly to fix my grammar, tighten what I write, and make it flow better. It worked well for a while. Lately it's been a struggle.

The bigger problem is this:

How Claude marks AI-generated content

They're putting watermarks in the text Claude generates. They call them imperceptible. You can't see them, but an algorithm can detect them. And those marks can stick around even after you edit the text.

I completely disagree with the "imperceptible" claim. Humans are extremely good at pattern matching. Show someone an old movie CGI today and they immediately know it's fake. That wasn't always the case. Much of that pattern matching is subconscious.

What no one notices today will be the tell tomorrow.

So who do you think is going to start checking for them?