I really don't give a F**K about the watermarking.

I really don't give a F**K about the watermarking.

Anthropic started marking Claude's text output on 2 August. EU AI Act, Article 50, applied globally. An invisible statistical mark, baked into token selection. Survives copy-paste. Dies under paraphrase.

Fine. Mark away.

I am borg'd. Claude is in the reordering of my pedalboard. It is in my KDP publishing packs. It is in my music theory practice. It is in the security architecture I get paid for. I did not sneak it in. I told everyone. Repeatedly. At length. In FIVE PARTS.

This is 1989 and spell checkers. There were people who said it was cheating. That real writers knew where the letters went. Those people were wrong then and they are wrong now, and I say that as someone who without spell check would be a forgotten dyslexic with a great many opinions and no readers.

Also worth saying plainly: the mark does not tell you Claude wrote it. It tells you Claude touched it. Ask it to tidy a paragraph a human agonised over and the human's work comes out marked. Paraphrase Claude's output lightly and the mark is gone. A test that yields false positives AND false negatives is not a detector. It is a compliance artefact. It exists because a regulation exists.

Which is fine. Regulations are how grown-ups do transparency.

I'd only point out that "was an AI involved" has never been the interesting question.

The interesting question is whether the thing is any good.

Mine are. Watermark them.