Post

@hern.bsky.social

I’m convinced, upon rereading, that this simply misunderstands how either or both of how LLMs or watermarking works daringfireball.net/2026/08/anth...

daringfireball.netAnthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of WritingIt’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance. The idea that anyth...

yes, this blog is v wrong, a shame because he's said some sensible & useful things in the past especially regarding apple app store policies & politics. ai watermarking interestingly invented by visiting researcher at OpenAI many years ago; co built it then sat on it as scared it would hit adoption

google has also been doing text watermarking for gemini for months and months, just never released a detector

@worklesshard.bsky.social

Think Google just open sourced their detector code? I will dig (if at all interested)

the detector code may be open sourced but SynthID relies on a private key

the EU AI Act's requirement in the guidelines that systems are interoperable, open, run locally etc, is the biggest issue with current designs, and one I haven't seen anyone pick up on

17/8/26, 1:15 pm0 reposts1 like

No replies yet.