Skip to content

Comment on How AI text watermarking works

Comments

In curious how does this work with tool calls or CLI scripts etc?

like if you have a long cli command or something will it still try to watermark it ?

Is there some way you can know which tokens are required to be syntactically correct vs not?

Yeah code in general seems like it wouldnt work

but is it even possibly to selectively apply it? I guess some kind of tag or indicator token that its code or not code?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.