John Gruber follows up on a number of issues around watermarking. On what to even do with detection tools:

There cannot exist a single detector for all LLMs. And even if you find out that it says it’s a match, that an email or blog post or Slack message was very likely generated by, say, Claude, what are you going to do? March into your colleague’s office and tell them you caught them?

Watermarking critiques are compelling because there are multiple points against it. Worse quality (maybe), false positives with quoted or lightly edited text (likely), centralized control of keys (yes).